Home

D3.3: Data management in euroFOT

Contents

1. S EF 408 IP 10 1 1 10 D7 offline zuletzt online 10 07 2010 20 05 05 Kilometerstand 9767 9 Km interne Festplatte 20 85 GB frei externe Festplatte 5 68 GB frei Position S EF 416 IP 10 1 1 15 D12 offline Kilometerstand 23372 4 Km interne Festplatte 21 16 GB frei externe Festplatte 349 99 GB frei Position S EF 411 IP 10 1 1 11 D8 online zuletzt online 12 07 2010 15 39 58 Kilometerstand 21578 8 Km interne Festplatte 20 75 GB frei externe Festplatte 387 84 GB frei Position S EF 417 IP 10 1 1 16 D13 offline zuletzt online 12 07 2010 10 55 45 Kilometerstand 5831 Km interne Festplatte 21 06 GB frei externe Festplatte 412 52 GB frei Position S EF 405 IP 10 1 1 8 D4 offline zuletzt online 12 07 2010 08 55 40 Kilometerstand 11619 4 Km interne Festplatte 21 04 GB frei externe Festplatte 411 53 GB frei Position S EF 410 IP 10 1 1 12 D9 offline zuletzt online 12 07 2010 10 20 07 Kilometerstand 18425 8 Km interne Festplatte 21 21 GB frei externe Festplatte 387 76 GB frei Position S EF 418 IP 10 1 1 17 D14 offline zuletzt online 09 07 2010 16 40 49 Kilometerstand 19397 7 Km interne Festplatte 21 5 GB frei externe Festplatte 434 97 GB frei Position S EF 406 IP 10 1 1 7 D5 offline interne Festplatte 21 28 GB frei externe Festplatte 207 41 GB frei Position http
2. Rsync SCP SDE SQL SSH TRG Use Case VPN XML Xvid Deliverable D3 3 The file storage area on the CMS holding any data not stored in the database Graphical User Interface Video codec used to compress video An open source relational database management system A commercial object relational database management system A software application for Unix systems which synchronises files and directories from one location to another while minimizing data transfer using delta encoding when appropriate Secure Copy Protocol a secure version of the Unix remote copy rcp command See SSH Standard Development Environment Structured Query Language a database computer language designed for managing data in relational database management systems RDBMS Secure SHell a security protocol for logging into a remote server SSH provides an encrypted session for transferring files and executing server programs Also serving as a secure client server connection for applications such as database access and e mail SSH supports a variety of authentication methods Tool Responsible Group A use case is a narrative description of the domain process It describes story or a case of how an external entity actor initiates events and how the system responds to them Thus it specifies an interaction between an actor and the system describing the sequence of transactions they undertake to achieve system functionality
3. SD amp SDHC compatible USB 2 0 High Speed configured as MSC Configuration application for logging parameterization and data conversion Supports DBC databases for signal selection Firmware upgrade through GPRS Led functional indicators Some specific or unique features that this datalogger is providing to euroFOT are Contact KL15 behaviour emulation with very low consumption lt 0 1 uA when KL15 off Python scripting enabled can sending traffic analysis possibility to implement other services according to ISO 14229 sample Server Application Internet based solution for remote fleet management tracking logging datalogger configuration provided Deliverable D3 3 lt Version 1 0 gt 15 Data management in euroFOT euroFOT 01 08 2011 e Data output format compatible with vector tools Basically a modem with GPS and GPRS capabilities was integrated within the Datalogger but around it a practically whole brand new architecture was built up The storage capacity was improved with the SD SDHC compatible module The modifications include a new automotive cable and modification of the server tool as well as configuration tool 2 2 1 3 Final implementation to fulfil euroFOT requirements Before the euroFOT project started CTAG had already Datalogger v1 0 To fulfil the euroFOT requirements a number of modifications and new features have had to be made These modifications can be divided in HW and SW upgrades HW archi
4. t 91 2 FTP S UrbanPredecessorSLCCAllowed FLE Bemis L ONU 44 44 45 CC SL_State V S MotorwayCc Pick Papa elie CC SL_Speed V S RuralCC Pick g eae 9 agg Vehicle_Speed ConversionTimestamp 29 Jul 2011 22 19 03 RR ns bance ConversionVersion 1 mn S MotorwaySL Pick es F F Map_SpeedLimit H DriverID 2 S RuralSl Pick Ea Duration 3491 S UrbanSL Pick M End 06 Jul 2011 14 42 56 S 30KMHNOSLCC Pick i ExperimentalPhase 2 S 30KMHSL Pick M Length 59 3 S 30KMHCC Pick M LoggerID 23103 S SOKMHNOSLCC Pick i PreProcessingTimestamp 03 Aug 2011 19 52 14 E CCSLSpeedChange 1503 1506 A PreProcessingVersion 2 E CCSLSpeedChange 1513 1513 ProcessingTimestamp 29 Aug 2011 14 59 05 ProcessingVersion 14 E CCSLSpeedChange 1519 1521 TripID 0 E CCSLSpeedChange 1533 1533 rip S RuralCC 1572 1777 TripLongID 2310311070611444501 E CCSLSpeedChange 1763 1766 TripNoInRecord 1 E CCSLSpeedChange 1770 1770 E E CCSLStateChange 1777 1777 Lens pe con 1 E CCSLSpeedChange 1863 1865 VehicleID 3 E CCSLStateChange 1863 1863 VideoFilesFolder TRAJ C3 110706 134308 E CCSLStateChange 1872 1872 gt 7 E CCSLSpeedChange 1873 1873 E CCSLStateChange 1873 1873 Ra n PreviousSituationsDele true MeanSpeed 61 1 E CCSLSpeedChange 1926 1928 MaxSpeed 655 E CCSLSpeedChange 1946 1946 PretMotorway 0 00086 E CCSLSpeedChenge 1973 1974 PretUrban 0 37 E CCSLSpeedChange 2012 2012 A E CCSLSpeedChange 2017 2018 PrctRural 0 5 E CCSLSpeedChange 2025 2025 isCC true E CCSL
5. Virtual Private Network a private network that is configured within a public network a carrier s network or the Internet in order to take advantage of the economies of scale and management facilities of large networks EXtensible Markup Language an open standard for describing data from the W3C World Wide Web Consortium It is used for defining data elements on a Web page and business to business documents XML uses a similar tag structure as HTML however whereas HTML defines how elements are displayed XML defines what those elements contain An open source video codec used to compress video lt Version 1 0 gt D Data management in euroFOT
6. localhost bilder D6capture jpg gt Euro FOT Status Seite S EF 412 IP 10 1 1 13 D10 offline zuletzt online 12 07 2010 09 08 12 Kilometerstand 10688 5 Km interne Festplatte 20 58 GB frei externe Festplatte 407 43 GB frei Position S EF 419 IP 10 1 1 18 D15 offline zuletzt online 12 07 2010 10 16 06 Kilometerstand 8434 8 Km interne Festplatte 20 82 GB frei externe Festplatte 390 7 GB frei Position Aktualisierungsverwalt ff Bildschirmfoto speichern amp Bildschirmfoto speichern Figure 24 Web based monitoring tool at Daimler This figure is adapted from D5 2 2 2 5 Italian VMC Unlike the other VMCs the Italian VMC does not log any CAN data and therefore does not use any sophisticated data acquisition system The data collection in the Italian VMC relies only on questionnaires The drivers provided their feedback by answering specific questionnaires or filling in forms related to specific events that happened during the FOT The subjective data collection is outside the scope of this deliverable However how the questionnaires data is stored and managed will be discussed briefly in the next chapter Deliverable D3 3 Data management in euroFOT lt Version 1 0 gt 45 euroFOT 01 08 2011 3 Database and Data Storage This chapter describes the work of WP3400 towards achieving objectives 2 and 4 of SP3 see Chapter 1 Section 3
7. BMW Cars Trucks Cars Trucks Cars Cars Cars Cars Cars CAN extra CAN vat extra sensor s Based on Table 1 a set of hypotheses for each function was formulated by SP2 From the hypotheses SP2 suggested which performance indicators controlled factors and variable factors to be collected during FOT study Detailed functions specification and hypotheses that are evaluated in euroFOT are described in Deliverable D2 1 Dozza Kessler Petersson Rakic Tattegrain Veste amp Heinig 2010 The set of hypotheses as well as the suggested performance indicators were used by SP4 to determine the necessary performance indicators situational variables events and experimental design These performance indicators situational variables and events were then used by SP4 to derive a list of measures needed to answer the hypotheses The list of measures was then used by SP4 in collaboration with SP3 to derive a list of signals suggested for collection For simple cases the signals are the same as the measures While for other cases a measure can be a function of one or more signals The list of signals was then used by SP3 to find out availability of the signals from the systems on board the vehicles When necessary external sensors were considered to get those signals that for some reasons are not available from the vehicle systems The complete lists of performance Deliverable D3 3 lt Version 1 0 gt 8 Data management in euroFOT euroFOT
8. Control Units and Sensors L KISSte Power management Comment Network attached storage NAS device used as the main component of the DAS This component has 4 hard disks each 1 TB Central gateway Driver camera Cameras to the front rear view and centre stack Additional Video Server Sensors External Server LS tes Data collection UMTS data transmission for logger status only Figure 19 Communication of the data logger components Deliverable D3 3 Data management in euroFOT This figure is from D2 2 lt Version 1 0 gt 41 euroFOT 01 08 2011 All DAS components are connected to the vehicle power network via the configurable intelligent system control KISSte see Figure 19 This KISSte is also connected to the K CAN The functions of the KISSte are e to start the DAS components when it recognises a bus wake up signal e to shut down all DAS components when it recognises that the bus falls asleep e to measure the vehicle battery status When the SOC value of the battery is too low the KISSte will prevent the DAS components from starting The function of the ZGW is to transfer the signals from the vehicle busses and the extra radar sensors to the NAS via Ethernet The data is then stored by the data acquisition software in the NAS The video data JPEGs is transferred to the NAS by the video server via internet The status of the data logging components is transf
9. J P fl F PEA Preparing ssssssss ETETETT j susseusese T CL jus Tai oo a C a kse ee l E a gaa i i ka P aa L LL L L CR F 5 d a L E _ es E a E f _ E Driving E Jee Peete refer nt fee L Figure 3 FOT chain introduced in FESTA SP3 role is marked with red curve Deliverable D3 3 lt Version 1 0 gt 3 Data management in euroFOT euroFOT 01 08 2011 7 SP 3 TT Installation ae strategies Type of data arralysts toot us es SP 6 Re quality check SP 2 Data st and Analysis Validation MeasuresMatrix feasibility from EMs echnical view constraints supportin purchasing amp maintenance SP gt SP 4 FOT Execution Performance Indicator and Experimental Protocol Figure 4 Collaboration between SP3 and other subprojects SPs in euroFOT 1 1 Vehicle Management Centre VMC The field operational test FOT in euroFOT is managed in four vehicle management centres VMCs VMC is an entity which can be seen as the overall organisation responsible for management and logistics of FOT The four VMCs and the partners involved in the VMCs are see also Figure 5 e French VMC CEESAR and IFSTTAR with support from Renault CEESAR is the partner that
10. Separating the handling of incoming disks data from pre processing stage has been a success Especially since all steps in the pre processing were not finished before data collection started The following could have been done better 21 We spent a lot of time and effort on gathering requirements for database and data Storage in the beginning of the project It might have been more effective to use a more iterative process Since in practice some of the requirements were changed and the solution was adapted however the requirements specification was not updated to include those changes Review on pre processing upload database and data storage 22 The whole chain of data pre processing upload and analysis should be tested if possible until Pl calculation before the FOT Some problems with signals e g strange values were realized very late in the whole process 23 The data pre processing and upload procedures should be flexible and allow changes after the FOT has started This is necessary because new problems might occur that did not exist in the pilot 24 Evaluate at an early stage all the software needed At the Swedish VMC some software packages decoding CAN NAVTEQ map data needed Windows as operating system causing a more complex automated flow on Linux based high performance computing HPC environment 25 Overall the cluster environment using MATLAB parallel toolbox has worked very well The drawback has been that m
11. desires Modules communicate with each other through events which can be listened to and generated by the modules A core module is responsible to 1 enable creation of new modules 2 offer basic data replay functionalities and 3 save and load FOTware configurations The other modules enable the other functionalities which are coming from the functional requirements described in Chapter 3 Each module is presented in a new window and can be more or less complex depending on the specific functionalities which it implements A possible configuration of FOTware is shown on Figure 67 Deliverable D3 3 lt Version 1 0 gt 98 Data management in euroFOT euroFOT 01 08 2011 File Edt View insert Tools Desktop Window Help DSd lh AA9984 4208 sO sPeyf 0T Video AL OO rod e Current trix 0002 01 03 05 57 42 Current tre 555 30 5 Paytock speed 000 Event type Event ID us E 2125607 Videc press key 1 5 to select M 729 OT Video press key 1 5 to select np timestamps omAlitude_ GPS omAmbientTemp omDayfight om ixTime_GPS omGPSTime Quality cmGPS Quality omHDOP_GPS Good I Bad Driver View Good Bad Feet View Good Bad Back View Good Bad Notappic Side View Right or left hand traffic Right hand Left hand t Unknown Good Bad Notapp c Armotation Name Current Foward View Gost Drvar View Goos Faai View Goos Back View Bad Si
12. e tables containing aggregated data The named files are added to the version control program Tortoise tortoisesvn tigris org 2011 Here backup routines are run at WIVW on a daily basis Furthermore the weekly backup for the internal hard drives of the server also covers the above mentioned files Description of Database Structure The overall data structure used for the German 2 OC is displayed in Figure 48 The data can be separated into subjective data collected via questionnaires and objective data logged in the vehicles The subjective data is stored in an Access db From the objective data an overview table that lists relevant information per trip is created The overview table is integrated into the Access db Through that a direct link between objective and subjective data is possible Paper questionnaires Online questonnaires DAJ DAJ BMW Baseline BMW BMW DAI Baseine erobdeDevice bu nDevice mob eDevice buikiinDevice z z t Y z z Y Y kd z Y Y Y Driver Driver Driver Driver Driver Driver Driver Driver Driver Driver Driver Driver Driver Driver Driver Orivor Driver Driver 1 2 1 2 1 2 1 2 1 2 1 2 Subjective data can Objective data om Access db Relation subjective Overview tabel s objective Figure 48 General layout of German 2 OC database Structure of objective data At the German 2 OC a defined folder structure is used that contains one mat file per trip The used structure
13. its outputs and a script which computes the Deliverable D3 3 lt Version 1 0 gt 92 Data management in euroFOT euroFOT 01 08 2011 outputs from the inputs The analyst can define the processes by using the explore struct shown on the left panel of the experiment builder The enrichment manager enriches and refreshes the trips by building a dependency graph Figure 60 from the processes definition and by running the necessary scripts The scripts are managed in a version control system VCS Local execution of refresher on currently loaded trip allows the analyst to preview the effects of a newly implemented modification on the trip data model The act of saving the modified data model back to the database is done by the SQL interface which updates metadata tables and prepares data tables to receive newly computed attributes ih Experiment Builder SIECLE LI GD Data Objects Events CCSLStateChange 1 Signal Tables i Signals i Data Sources fi Data Source Classification ai Events Tables Ei Events D CRETE ea Generator Inputs ef TimelnTrip i i CC SL State BE T Refresher Inputs A A Attributes 7 H Sub Events TripDataModel Events 1 3 Mame Description Table Persistence Generator Script File Refresher Script File Event Instance Input Variable CCSLStateChange CC SL_StateChange CCSLStateChangeTable j ALY_GenerateCCSlLStateChangeEvents m ALY _RetreshCCsLStateChangeEv
14. length DOUBLE comment VARCHAR 512 columnname _1 SituationsCalculationsAttrs DOUBLE columnname_2 SituationsCalculationsAttrs DOUBLE columnname_N SituationsCalculationsAttrs DOUBLE Vid INT situationtypaid INT situationcalculationid INT 11 Vid INT 11 name VARCHAR 45 tablenameid INT scriptfile VARCHAR 128 occurencesvariable VARCHAR 128 description VARCHAR 512 comment VARCHAR 512 Vid INT situationcalculationid INT 11 name VARCHAR 45 type VARCHAR 45 variable VARCHAR 85 composition BOOLEAN Figure 29 Database schema at German 1 OC and French VMC This schema is split into two to make it readable see the next two figures lt Version 1 0 gt 52 Vid INT situationcalculationid INT 11 name VARCHAR 45 columnname VARCHAR 45 type VARCHAR 45 variable VARCHAR 85 unit VARCHAR 150 display BOOLEAN euroFOT 01 08 2011 Eurofot Database Schema Version 2 2 21 3 2011 m F vehicleID INT id INT annotation VARCHAR 512 columnname VARCHAR 45 name VARCHAR 45 unit VARCHAR 45 display TINYINT description VARCHAR 512 type VARCHAR 45 name VARCHAR 45 unit VARCHAR 45 Q display TINYINT description VARCHAR S type VARCHAR 45 m description VARCHAR 512 type VARCHAR 45 description VARCHAR
15. 01 08 2011 indicators and measures are described in Deliverable D4 1 Kircher Heinig amp Brouwer 2009 and the experimental design is described in Deliverable D4 2 Jamson et al 2009 The list of signals suggested for collection includes longitudinal driving behaviour such as speed mileage longitudinal translational acceleration lateral translational acceleration break force brake pedal lateral driving behaviour such as steering wheel angle lateral position of the vehicle yaw rate direction indicator lane marking type lane width number of lanes data from radar and other vehicle sensors such as distance to vehicle ahead wiper signal headlight activation high beam activation fog light activation selected active gear driver based system such as activation and setting of certain safety functions warning activation other vehicle system such as airbag status ESV activation ESV intervention ABS intervention engine information such as oil temperature engine speed and fuel consumption video data such as video of driver s face forward view driver s feet and cabin view eye movement data such as gaze direction and eye closure GPS data such as longitudinal and latitude positions speed direction of travel time and date weather data such as presence of rain outside temperature lighting condition times of sunset and sunrise geographical map based data derived measures such as jerk time headw
16. 1 explains the working method of WP3400 and Section 3 2 describes the database and storage solutions used in euroFOT All kind of data collected during the FOT experiment that are relevant for analysis are stored in the database or data storage server at each VMC This includes data from in vehicle data acquisition systems but also data from other sources for instance participant related data and subjective data 3 1 Method Procedure Used The common process used to select and finally implemented the database and data storage solution in euroFOT is shown in Figure 25 from the first step gathering all requirements until implementation and test 2 General Database Implementation VMC GE ENS OL Figure 25 The common process for database and data storage The requirements Task 1 in Figure 25 were gathered from the different VMCs in the beginning of the project using common templates with the aim to define in a unified way the data storage and processing chains used in different VMCs The requirements specification supports the description of agreed common definitions and requirements and also allows the description of VMC specific requirements An attempt to make a general database structure Task 2 in Figure 25 was made but this proved to be not efficient to be implemented in the different VMCs Therefore each VMC has had to make modifications to the structure to suits their specific need Similarly for Task 3 in Fi
17. 23 52 20 April 2011 18 26 48 Idle 3 23120 Ford 80 226 173 131 49202 4 2 21 April 2011 07 57 52 21 April 2011 07 58 06 Idle 4 23121 80 22617317651800 42 pril 2011 09 39 46 21 April 2011 09 40 43 dle 5 Fo 23122 Ford 8022617314260950 42 20 April 2011 19 58 02 20 April 2011 19 58 17 Idle 6 D 23123 Ford 8022617313644556 42 21 April 2011 09 40 56 21 April 2011 09 39 37 Synchronizing 7 8 23125 MAN 80 226 173 81 60258 40 21 March 2011 14 09 37 21 March 2011 15 05 59 Idle 8 23128 MAN 80 226 173 91 34443 40 21 April 2011 09 25 08 21 Aprl 2011 09 28 14 Idle 9 23129 MAN 80 226 173 63 39997 35 17 March 2011 15 42 02 17 March 2011 15 42 23 Idle 10 f 23171 MAN 80 226 173 75 34034 40 11 March 2011 10 28 43 11 March 2011 10 31 12 Idle 11 E 23133 MAN 80 226 173 70 39076 40 21 April 2011 09 34 28 21 April 2011 09 39 00 Idle 12 amp 23134 MAN 80 226 173 69 52403 40 21 April 2011 09 38 32 21 Aprl 2011 09 38 56 Idle kal frmConrersionMa Deliverable D3 3 omversionManacer ui Server Modules ai Avafable Dataloggers Figure 10 Web based monitoring tool used by German 1 OC and French VMC Power Management In order not to drain the battery a reliable power management should be performed In this sense it was necessary to adapt the logging system to work only with KL15 being suitable for series vehicles In the former version of the CAN DAS this possibility was not considered as it was prepared for R amp D activities An appropriate mana
18. A network standard for communication between control units in a vehicle Data Acquisition System The in vehicle hardware and software that constitutes a complete logging system The period from the ignition is turned on until the ignition is turned off It this document the word trip is also used Electromagnetic Compatibility Hardware and or software for tracking the movements of the drivers head and eyes Hard Disc Drive Global Positioning System General Packet Radio Service A packet oriented mobile data service A network standard for communication between control units in a vehicle Local Interconnect Network A network standard for communication between control units in a vehicle Lower communication speed than CAN An object detection system that uses electromagnetic waves to identify the range altitude direction or speed of both moving and fixed object Real Time Clock In this document it denotes the time and calendar circuit in a PC or similar computer A source of signals e g the CAN bus a video camera or an eye tracker Vehicle Communication Interface A piece of hardware for accessing CAN LIN and J1708 vehicle networks Secure Sockets Layer A protocol that provides security and data integrity for communications over networks lt Version 1 0 gt C Data management in euroFOT euroFOT 01 08 2011 Database storage and analysis tool related File Storage GUI H 264 MySQL Oracle
19. Additionally it would be really difficult and time consuming to gather all data under a certain hard disk in order to upload them to the database Finally we should be able to prevent a bad usage of the API functions that implement the interaction with the database which would create an inconsistent database So a lot of thought and experimenting needed to create a fast reliable consistent and flexible relational database Successes The careful design of the database schema was an urging and important work During the process of data storage the upload is going on trip per trip which includes original signals derived metrics and aggregated data related to each trip All meta information is also stored in the database This includes for instance the description of experimental or aggregated data as well as the description of the complete data reduction process and its history As a result although original trip files and video files are kept in a file based storage the database is self documented and hence contains all the information necessary for data analysis The Deliverable D3 3 lt Version 1 0 gt 57 Data management in euroFOT euroFOT 01 08 2011 interaction with the database is taking place in two phases Meta data describes all the objects in trip data as well as the relationships between them and trip data corresponds to the original data as well as the data which is derived from it during data reduction So using the meta data
20. E Marked and M A N EMC requirements most restrictive ones compliant Real driving tests About 10 000 Km were done by CTAG Intensive drives with several vehicles passenger cars and trucks were carried out analyzing the performance of the DAS under those circumstances including some vehicles with video logging and additional external sensors Web based monitoring tool A tool Figure 10 to monitor the status of the vehicles and the DAS is also built as part of the data management framework used in German 1 OC and French VMC described in Section 3 2 1 The tool is called Fleet Manager Besides an overview of all managed DAS on the server site and their current operation state additional processes of the framework have been integrated and can be directly accesses by using this GUI By selection of one DAS detailed state information can be displayed e g the current firmware version on the DAS the content and state of the FLASH storage the current GPS position etc Each DAS needs to be registered at the server before starting uploading data to the server When a possible issue is identified or attention is needed an email is sent to the quality management team 2 2 1 5 Experiences and lessons learned In the selection and adaptation of a suitable DAS for German 1 OC some interesting experiences have been obtained These experiences for CAN only loggers can be Summarised as follows e Reliability is a key factor in the DAS s
21. O a E E E A 33 2 2 3 3 Final implementation to fulfil euroFOT requirements ssseeeeeeeeeeeesssssseeeeeee 33 22 54 Vests and quality check performed sa 2 5 en a 37 2 2 3 5 Experiences and lessons learned 39 Drs FR STAN On ne 40 22A AL COCO on de a ee no ee tee a eo ee 40 M COS SOON a 40 2 2 4 3 Final implementation to fulfil euroFOT requirements o on 4 2 2 4 4 Test and quality check performed cc eececceccccecceceesesseeceeeeeeeaaaeseeeeeees 44 223 121 L 1 2 CR 45 SM D AA DAS AO DIT O eeue a hye uacans ee haneaauuacanece eee 46 3 1 IVICTHOC Procedure DS a a 46 32 Database and Storage Solutions Used in euroFOT sssnoneessssssssoeerssssssssseeerssss 48 321 German l OC amp Fronch VVC ace E 49 DE M PaSa SOO MR a de E E E dore 49 S22 Upload and nchment OLIS seeiis 56 Pea Eo NC AIC a a ad eee do cette ae 57 Jd N a VME E E E E E E E E E seam E E EE 60 322l Data storage SOON 22 css E ee ed senc tete 60 3 2 2 2 Upload and CRTICNIMENTEOOIS descendance etes 73 Dee T0 5 e E E E E E iets 74 Data management in euroFOT euroFOT 01 08 2011 Po CAO S 79 3 234 Datastorage SOLAR iia E EEE i 79 3 2 3 2 Upload and enrichment tools seccvcncresctvincsntvndsouncsreserasenatenarwsonsdexseeasreserneowesioxs 82 D Leom OA E S S 83 Jaa NM AAE ustvestece ceaceeeeeanes 84 3 2 4 1 Data storage and upload solution 84 eZ LE SSO RS AO a anes 84 4 Data analysis AOL ares E ne ne
22. SP5 47 lessons learned from this project from a data management point of view are highlighted Using common templates for gathering requirements and having specialised focused groups to enable joint efforts in performing certain tasks have been found fruitful Compact flash cards SD cards failing connectors and loose cameras are the common DAS hardware issues CF and SD cards with industrial grade are recommended Making sure that all personnel involved in installation are well trained to do the job is a must Furthermore giving extra attention to the installation of connectors and cameras is recommended On the data handling procedure separating the handling of incoming disks data from pre processing stage has been a success Considering the cost for making database backups if the raw or Deliverable D3 3 lt Version 1 0 gt 118 Data management in euroFOT euroFOT 01 08 2011 pre processed data are stored then there might not be a need for taking backups of the continuous data tables As described above in addition to storing non video data in the standard relational database some partners also stored these data as MATLAB files to allow analysis both through database and or directly using MATLAB The use of special features like partitioning and compression of relational database e g Oracle has not been thoroughly investigated in this project The effectiveness of the database and storage solutions we chose is yet to be tested to t
23. Start of evasive maneuver Edit TIME here End of evasive maneuver Edit TIME here Physiological impairment not including Drowsiness omen aea Hands on wheel No hands one hand Two hands Unknown Seat Belt eee See e Visual obstructions No obstruction Rain snow fog smoke sand dust Reflected glare Headlights Curve or hil Burlding billboard or other roadway infrastructur Trees crops vegetation Moving vehicle with or without load Parked vehicie Driver attention allocation Attention towards the cntical event Gaze direction RHT Driver activity Not distracted Passenger related secondary task Adut pass Adut pass Child in ad Child in re Talking singing no passenger apparent Talking singingidancing internal distraction not vehicle or passenger Reading A Wireless device Takingstenng Vehicle related secondary task Adjusting climate control Dining Eating win Eating win nana Dan o Smoking Reaching 1 Lighting ci Smoking c Extinguish Daydreaming Lostinthought Looked but did not see External distraction Looking at traffic other than critical event Personal hygiene Combing brushingfixing hair Navigation 4 y 1 Looking at map Miewing naviga Interacting with Driver activity RHT Checking centre rear wew mirror Driver activity LHT Checking centre rearview mirror Figure 67 Screenshot from a two scre
24. The Trip visualisation and Annotation tool and 3 Trip data model front end tool also Known as Experiment Builder Using the tools the analysts can e Search for trios corresponding to a specific set of filters get information about processing progress warnings and errors e View both raw un synchronised data just uploaded from the DAS and synchronised data extracted from the database e Modify any editable authorised attributes especially in relation with events and upload modified trip to the database e Prototype and test modifications of algorithms or data model and push them in the production environment Trips Explorer The Trips Explorer is a very powerful tool to search for specific trips using filters and access the corresponding data or test some algorithms on them Filters are based on standard status information for each trip configuration date origin error or warning but also on any kind of data which has been declared in the Trip Data Model and created by the user during the analysis It is therefore possible to search for trips where general attributes meet certain requirements or which contain a specific kind of event where an event attribute has such value Filters are defined using the GUI pictured below whose controls are entirely based on the data model Filters can be saved and imported Search is performed either by querying the database server or using a simpler local file based database whi
25. algorithms Execution using Petri Nets f Quality Check A a ooo Re usage of processing chain Pre and Post Processing og 20 D Y gt 9 a gt Raw d Datab Storage of raw Using pre processing ew ate atabase and processed data and interim results Figure 28 Framework Deliverable D3 3 lt Version 1 0 gt 50 Data management in euroFOT euroFOT 01 08 2011 To prevent algorithms to become more complex by respecting different sample times a framework Figure 28 encapsulating the algorithm execution ensures homogenous time samples during pre and post processing The operation of the framework is completely based on a version controlled configuration system all data artefacts their definition relationships and user operations or algorithms which create or alter them are stored in what is called Trip Data Model Data and is stored in two different ways file storage and relational database Table 12 Depending on type of data and configuration the same information can be stored either ina file in a relational database or in both Backups are also implemented at both VMCs and for both types of storage Table 12 Data storage at German 1 OC and French VMC Meta Data Video only Raw Data MATLAB Pre Processed TripDataModel French VMC Raw processed File Version Binary files Raw binary Same Cleaned up Pre Storage controlled h 264 as acquisition information Synchronized processed struc
26. all FOT data is available Whenever running a routine it is checked for which drivers data is available in all three conditions and which of those have no results table yet The analysis is only performed on the data of those drivers that are new Routines have already been implemented that extract indicators for the following cases e Per trip overview table e Per event e Per input to navigation system e Per situational category combination of categories defined by WP6400 4 2 3 2 Tests performed and quality assurance The tool was tested by the tool responsible person at the pilot stage The initial version was developed at BMW and then it was improved by the IZVW There was very close interaction between users and developers and the errors and bugs were corrected immediately since there are few users MATLAB scripts are written and executed on demand Deliverable D3 3 lt Version 1 0 gt 110 Data management in euroFOT euroFOT 01 08 2011 4 2 3 3 Lessons learned Difficulties and Solutions Data handling especially of video data should be tested with a large amount of data already in the pilots In the beginning also the Daimler video data was stored as single frames instead of video format Although there were a large number of separate files for each trip the handling of the videos e g copying was no problem with the pilot data about 50 hours 3000km of data logging After converting the first months of FOT data it
27. annotations in time Further the user can see the current status of the annotation work in a progress bar at the top of each window An overall progress bar is also presented at the top of the browser window Finally a check box in the browser window enables the user to write the annotations When closing the annotation module the user is asked if the annotations need to be saved If the user consents annotations will be saved in a MATLAB file This file contains all information needed to relate the annotations to the specific trio and event The file has a unique name indicating the user the time of annotation and the related trip e g dozza_A_20 14045 11 Aug 2010 13 22 36 mat can be interpreted as user dozza made an annotation on event 20 in trip 14045 on the 11 of August 2010 at 13 22 36 Annotations files are uploaded to the database by a super user with rights to write in the database typically not the annotator who is not likely to have such rights Foward View Event severity Pre event manouvre RHT Physiological impairment not including Driver activity Good Bad Normal driving Going straight constant speed Drowsiness Not distracted Conditions Driver View Te rash deformation extent Passenger related secondary task J Good Bad Sy ous Pw LL Pre event manouvre LHT Feet View ER So PR S on cent Primary task load Adult pass Adult pass Chuid in ad _ Child in re Bad MAIS 7 Talking singing no passenger appare
28. based on Sub Version For the development workflow a strategy on how to handle different development paths in the version management system in parallel was defined Deliverable D3 3 lt Version 1 0 gt 94 Data management in euroFOT euroFOT 01 08 2011 Log Messages D Ablage euroFOT euroFOT_trunk BAX From 05 05 2011 To 06 06 2011 v PI Revision Actions Author Date Message 560 ghaouty 10 33 03 Montag 6 Juni 2011 check if the trip shall be ignored 559 ghaouty 10 26 20 Montag 6 Juni 2011 change access for Uploadprocessor server 558 ghaouty 10 21 57 Montag 6 Juni 2011 new script to create Ignore attribute into according directories 556 ghaouty 16 16 51 Dienstag 24 Mai 2011 Update Fleetmanager 555 ghaouty 15 48 12 Dienstag 24 Mai 2011 changed the references of the project settings use CMS libs and GUI components instead of other interfaces 554 ghaouty 15 46 02 Dienstag 24 Mai 2011 change the selection mechanism For the logger state File 553 ghaouty 15 39 32 Dienstag 24 Mai 2011 added export Functionality for logger state list 552 i kuefen 14 14 21 Dienstag 24 Mai2011 added changes from release 1 1 0 to trunk 551 kuefen 14 03 46 Dienstag 24 Mai 2011 updated latest trunk version 544 ghaouty 18 57 11 Montag 23 Mai 2011 created new interim deliverable as version missmatch within releases were detected 543 ghaouty 18 44 55 Montag 23 Mai 2011 adapted qui of fleet manager 542 ghaouty 18 31 55 Montag 23 Mai 20
29. but maintains their objective data as MATLAB files In the Swedish VMC the CAN data are stored in an Oracle database and as MATLAB files in the file server where the videos are also kept Having only questionnaire data the Italian VMC uses a MySQL database The base software for data analysis used in the different VMCs is detailed in this deliverable While the look and feel of the software is different in the different VMCs the functionalities are very similar At the time of writing the tools have been used to extract visualise data from multiple sources and do some simple analysis in the Swedish VMC the tool has also been used to annotate the video data for several months Furthermore how we came to choose those different solutions as well as the general quality assurance throughout the whole data management chain adopted in this project are all summarised in this deliverable The quality checks are done at many stages for example before and after installation of DAS during FOT operation through the web monitoring tool during pre processing of the data as well as when the data are uploaded The data quality is measured per sample per measure and per trip As part of quality assurance the installation strategies software for installation verification web based DAS vehicle monitoring tools procedures of handling problems during data collection procedures of data handling from vehicle to VMC centres are created and agreed with SP2 and
30. codec e Videologger must allow synchronization with CAN logger with a better than an image 40ms precision e Videologger must have at least 4 video channels Deliverable D3 3 lt Version 1 0 gt 23 Data management in euroFOT euroFOT 01 08 2011 2 2 2 2 Chosen solution Several options were considered including the eventuality to use different dedicated dataloggers for each instrumentation level However CEESAR finally designed a modular system based around a single datalogger model a CTAG Datalogger Il same as the DAS used by German 1 OC see Section 2 2 1 2 This modular system can be as simple as the CAN only and GPS datalogger but can also support radar videologging eyetracking and lateral position measurement This modular approach has two main advantages e Flexibility the level of instrumentation can be adapted car by car e Compatibility All different levels of instrumentation produce comparable data Hence there s no need to harmonise the datasets or convert files produced by different data acquisition systems It is also quite cost effective in its most complete implementation it is not significantly more expensive than more integrated data acquisition systems which implement the same functionalities but which can t exist in a simpler version It finally proved feasible after screening of a lot of suppliers and long negotiations to within budget constraints reach both goals e Fitting th
31. easy to use graphical user interface modules 42 The development process of data analysis tool and improvements on the data format and structure was an iterative process While using the analysis tools the data formats that were not consistent with the design and defined structure were also found and corrected 43 The modular approach helped using the same modules in different use cases and or on different machines The following could have been done better 44 The modular approach of the software should be extended Especially the interfaces could be more standardised and could offer additional features like a command line interface As well the integration of additional external tools beyond MATLAB could be very useful Review on technical methods and tools used 45 The operating system had a relatively slow data transfer rate for networking operations Given that the data viewing for the annotation was performed through a network link it tended to be a lengthy process which in turn makes the annotators waiting to load the video for few minutes 46 Object oriented programming in MATLAB has a steep learning curve even for the 47 experienced programmer it took time to change the syntax from the usual structured programming Object oriented programming in MATLAB might be too complex for small GUIs and debugging could be difficult since the errors are not shown in the corresponding line It is apparent from the desig
32. ee eee 85 4 1 INICIO D PROCESS eee soiasaneeceeeeeesaee 85 4 2 Base of Data Analysis Tools Used in euroFOT 1 0 0 ccccccccccessseseeeceeeeeeeaeeeeees 86 a2 German OC and French VMC ns ae date esse teens 86 ddA WIM VC 8 AS RUE LENOIR cu scuaistacali nar suamadeaeet oe seacalalenncuceinupsauedssecee moe scelasesseueetins 87 4 2 1 2 Tests performed and quality assurance 94 My Ne SSO CAE A aaseeceeeuessezssatectectueane A saneecets 97 BD SW SR 98 A A LCL 1 02 LE GT 8 14 LENOIR AAE arand 98 4 2 2 2 Tests performed and quality assurance 106 BD ICS SOUS TOA deene EEEE T 107 MD A On a E A E E EAA AET S 108 ZA Tp Li 8 AS LE LE Osa e a e a a Hate 108 4 2 3 2 Tests performed and quality assurance 110 koo Eeo ea a a A 111 LE NM ee e E E E E E 111 4 3 K D e A E A A E E EE E EAA E E 111 5 Summary of Lessons Learned and Recommendations essosssooeenesssssssseeressssssseeeeees 113 5 1 Data Acquisition SYSLCUIS SR dede ee iles 113 5 2 Databases amp Data Storage Rd ee do 115 5 3 Data Analysis Ti OOS ecs E stem oaicastecuem 116 COCO sr a E E E A E 118 A RAON WC GO OOS a AA A 120 ROIS E A E secs 121 ADS LRO NN OS ae de de a el M PRINS KZ GO e E E E EE S B Deliverable D3 3 lt Version 1 0 gt iv Data management in euroFOT euroFOT 01 08 2011 List of Tables Table 1 Safety functions tested in euroFOT annann EENAA EEKEREN Nai 8 Tabe 2 TEC ODA lh UY 2009 a E T 8 Table 3 The main features of the data acquisition sys
33. field vehicle until the data is ina final MATLAB mat format to be uploaded at the VMC centre e Uploading as the step of reading the data from the transfer disk s and uploading the non video data to the Oracle database and uploading video files and mat files in appropriate folders in the fileserver A more detailed description of each part will be described later on in this section In order to optimize the overall data pre processing time it was important to reduce the number of time data is read from hard disk and pre processed before finally being uploaded to the database and data storage at the VMC centre Since there are some processes that can only be done at OEM for example decoding and extracting CAN messages it was considered more efficient to do as many pre processing steps at the OEM side as possible The implementation of the needed pre processing scripts module was jointly done by the Swedish VMC partners led by Chalmers During the pre processing step the non video data from USB hard disk is converted to a final MATLAB mat format before being uploaded to the Oracle database These MATLAB files will also be uploaded as files to the fileserver This is done to allow analysis stage to be done through and benefit from the strength of relational database in terms of access rights concurrency consistency integrity etc as well as the easiness of flat files operation for certain types of analysis The chosen database engi
34. for the pre processing at the OEMs large multi core clusters less time has been needed to spend on optimization In the pre processing the following gives an indication of the relative time between the different tasks and as absolute values Tasks Seconds per hour collected data workstation Copying of data from transfer disk to server Decrypting data Reading data from raw format into MATLAB Decoding CAN and do minor processing Perform all filtering resampling quality measures derived measures and events for everything except eye tracker Perform filtering and derived measure calculation for eye tracker Perform Kalman filtering position enhancement Map matching Unpack and convert video to X 264 Copy video data from transfer disk to server eSATA with RAIDO Load oDBdata into workspace Upload trip information to DB Upload event information to DB will differ significantly between trips Deliverable D3 3 lt Version 1 0 gt 77 Data management in euroFOT euroFOT 01 08 2011 Tasks Seconds per hour collected data workstation Create text files with data for SQLidr upload Uploading of data text files to DB with SQLidr All times are for one single process For fields marked yellow we have 8 parallel processes that is the processing time should be divided by three For fields marked For fields marked For fields marked For fields marked we have 6 parallel processes we have 12 parallel proc
35. for video playback However through the use of ffdshow a sufficiently fast solution could be achieved to play at least 6 video channels in parallel on a reasonably modern computer Although the FOTware is same between the partners in the Swedish VMC there are still differences that depend on the location from where the FOTware is used Thus configuration for a specific location was used which enables having 3 different setups instead of changing the root paths for FOTware location of video on the filesystem and the database settings Future Recommendations For future development the modular approach is strongly recommended MATLAB s COM object interface allows other application to control and interact with all scripts running in MATLAB thus development can also take place outside of MATLAB 4 2 3 German 2 OC In German 2 OC the data is stored as mat files MATLAB files in file management system no relational database Visualisation of video data and some time series data is done through a GUI built in MATLAB For some deeper analysis MATLAB scripts are written and executed on demand 4 2 3 1 Implementation A MATLAB GUI has been implemented that allows viewing the video together with selected variables of time series data The function of the program is tailored to the data folder structure used in the German 2 OC Based on the naming and structure conventions it selects automatically the video on the external hard drive that bel
36. gt 54 Data management in euroFOT euroFOT 01 08 2011 The following scheme presents the used data types and the different processing steps after data has been uploaded on the server As a first step the uploaded data is converted to a MATLAB file in order to start the following processing steps Afterwards these data is enriched with data from the digital map and the needed events Pls for data analysis These data is stored on a file server per trip basis AS soon as data processing is accomplished the data is uploaded on the database For this the processed MATLAB files are translated by means of the database schema and stored and stored on the SQL database The data stored in the database has different sampling rates in order to reduce the amount of data to be stored on the database The following sampling rates have been defined based on requirement for data analysis e 1Hz Status information turn indicator system status etc Map data road type speed limit etc e 10Hz vehicle dynamics speed longitudinal acceleration etc Incoming datafrom DAS encrypted and compressed MatLab Files Decompressed MatLab Files Pre processed enriched with map data MatLab Files Processed anemia enriched with events Pls etc sche 1 into DB SQL data base File Server Trip based Figure 32 General layout of data processing at German 1 OC and French VMC Deliverable D3 3 lt Version 1 0 gt 55 Data ma
37. in the process of selecting suitable DAS are 1 functions amp hypotheses to be tested performance indicators Pls study design measures sensors allowed needed ethical and legal issues as shown in Figure 3 FESTA consortium 2008 The safety functions tested by the different VMCs are shown in Table 1 2 number and type of FOT vehicles car truck considered to be equipped length of data collection considered and restriction on type of data considered to be collected if any A requirement or constraint to consider from the beginning is the type of DAS to be used by each VMC 1 CAN only 2 CAN video 3 CAN video extra sensor s However the CAN video is finally changed to CAN extra sensor s During the first year of the project types of vehicle to be equipped cars trucks how many for each type and the length of data collection period were revised to reflect the situation at that time the FOT plan in July 2009 is shown in Table 2 and the lower part of Figure 5 Deliverable D3 3 lt Version 1 0 gt 7 Data management in euroFOT euroFOT 01 08 2011 Table 1 Safety functions tested in euroFOT Function FORD MAN VW MAN vw VOLVO vcc CRE RENAULT BMW Forward Collision LL V and Brake Support Speed Limiter and Cruise Control Blind Spot mw System lt S Interaction a oli CARE ME D D 1 at PE CS ES ES in a A Table 2 The FOT plan in July 2009 Type of DAS FORD MAN VW VOLVO RENAULT
38. individual vehicle The flow chart shown here explains the main flow of the pre processing Deliverable D3 3 lt Version 1 0 gt 71 Data management in euroFOT euroFOT Pre processing in the Swedish VMC in euroFOT 01 08 2011 This flowchart shows the Swedish VMC s data pre processing steps from when data is available at OEM until it is delievered at SAFER Chalmers for upload 1 Read data and eye tracker processing 2 Naming fix time and high freq derived measures 3 Filter and resample per measure 2 w 3 O 5 Calculate derived measures and events Save to storage Deliverable D3 3 Load vehicle unique configuration file oPreProcCfg and create output directories if necessary Disk plugged into USB port in workstation and Per trip processing starts Estimate space needed and check free space on server and transfer disks Matlab instance intiates processing Change measure names according to configuration file 1 name harmonization Handling NaN Error values for some measures specifically for CAN Fix timestamps per data source Use linear regression to avoid clock drift Apply anti aliasing filter Based on oPreProCfg Apply resampling for each individual measure Crucial step Based on oPreProCfg Add to oDBdata Decrypt data extract all data sources and save in intermediate oDataSet mat format Calculate pre resample derived measures e
39. management in euroFOT euroFOT 01 08 2011 Figure 38 Part 1 of the database structure used in the Swedish VMC Deliverable D3 3 lt Version 1 0 gt 64 Data management in euroFOT euroFOT 01 08 2011 Figure 39 Part 2 of the database structure used in the Swedish VMC Deliverable D3 3 lt Version 1 0 gt 65 Data management in euroFOT euroFOT 01 08 2011 Figure 40 Part 3 of the database structure used in the Swedish VMC Deliverable D3 3 lt Version 1 0 gt 66 Data management in euroFOT euroFOT 01 08 2011 LE Figure 41 Part 4 of the database structure used in the Swedish VMC Deliverable D3 3 lt Version 1 0 gt 67 Data management in euroFOT euroFOT 01 08 2011 Figure 42 Part 5 of the database structure used in the Swedish VMC Deliverable D3 3 lt Version 1 0 gt 68 Data management in euroFOT euroFOT 01 08 2011 Figure 43 Part 6 of the database structure used in the Swedish VMC Deliverable D3 3 lt Version 1 0 gt 69 Data management in euroFOT euroFOT 01 08 2011 The MATLAB structure that describes the data oDBdata mat structure The oDBdata is the format of data after pre processing and before uploading to the database Figure 44 is an overview of the structure Note that oDBdata is the format that all functions in analysis are and will be using this is what the queries to the database will be returned as oSQLinfo oSegmentinfo 1 N voMeasureData 1 N oDBda
40. manages the French VMC database e German VMC is divided into two operation centres OCs o Operation Centre 1 VW AUDI FORD MAN and IKA The data management in this OC is taken care by IKA o Operation Centre 2 Daimler BMW and IZVW The partner that manages the data management in this OC is IZVW e Italian VMC CRF and POLI The data management for this VMC is conducted by POLI e Swedish VMC VOLVO VCC and Chalmers The Swedish VMC database is hosted at and managed by Chalmers A detailed description of each of the vehicle management centres is available from Deliverable D5 1 Flament Cavallo amp Hagleitner 2009 The participating partners in SP3 are Chalmers IKA Volvo VCC CEESAR CTAG ICCS and Bosch The German 2 OC and Italian VMC are not represented in SP3 i e Daimler BMW IZVW CRF and POLI are not member of SP3 Therefore only a summary of the data management solutions used by German 2 OC and Italian VMC is presented here more detailed descriptions of their work in terms of data management are available in other euroFOT deliverables Deliverable D3 3 lt Version 1 0 gt 4 Data management in euroFOT French VMC CEESAR Renault IFSTTAR 40 cars 13 months CAN Extra sensors 35 cars amp CAN Video Extra sensors 5 Cars Deliverable D3 3 Data management in euroFOT euroFOT 01 08 2011 Italian VMC 140 cars 100 trucks 30 cars 300 cars 12 months 9 12 months 3 4x3
41. really useful Therefore it has been decided that if that proved to be feasible within budget constraint all vehicles would be fitted with a headway measuring device and that the five vehicles with higher instrumentation level would be equipped with lateral position measuring device and an eye tracker Generally the French VMC requirements are similar to the requirements of German 1 OC see Section 2 2 1 1 However French VMC also has some additional or specific requirements The main additional or specific French VMC requirements are e The two levels of instrumentation should be compatible i e the richer instrumentation should be an enhanced version of the lower level of instrumentation relying on the same central data logger e All data CAN or non CAN data should be logged into the same data format as CAN data in the logger i e external sensors that produce CAN signal s or those that can be interfaced to CAN with a gateway are favoured e Videologger must have enough capacity to store data for two weeks This is because of the experimental design in the French VMC the five cars with richer instrumentation were planned to be rotated among drivers and these cars were planned to be lent in a two weeks based period to each of the drivers both during baseline and test phases e Videologger must be embeddable autonomous and affordable e Videologger must generate efficient and easy to use files i e use a modern standard
42. reduce energy consumption also affected the storage and upload procedure Therefore the server tool was adapted to enable both data collection and upload of stored data on the SD card simultaneously e Script possibilities implemented Phyton based The deactivation process of the data logger after the engine is switched off was additionally integrated into the available Phyton scripts e Modifications on Configuration tool The modification of the energy consumption also affected the configuration software Hereby the function of using the KL 15 possibility was implemented and integrated as an operation mode in the configuration software e Configuration file stored in Flash memory in case of SD card corruptions the data logger could only be approached after replacing the new SD card Hence it was decided to store the configuration file in the internal Flash of the data logger in order to have the data logger still accessible for diagnostic operations In addition to these some extra modifications were done to achieve the currents status Some of them triggered by German 1 OC or French VMC and some more decided by CTAG in order to improve and optimize CTAG Datalogger II capabilities gaining more reliability Serial port logging capabilities for video synchronization SPI connection availability for future upgrades Supervisor included in order to avoid hang ups due to some ignition voltage signal drops Possibility to work as HTTP RS
43. relates each trip to a condition and a driver ID A similar structure is used for storing the video data on the external hard drives The used structure allows to link for each trip the mat file stored on the server to the video stored on an external hard drive Deliverable D3 3 lt Version 1 0 gt 80 Data management in euroFOT euroFOT 01 08 2011 Data logging in vehicle Video as video or single frames Driving data as mat Files temal hard drives lt Defined folder structure DAL DAJ BMW BMW DAI Basebne BMW Baseine motieDevice tuik nDevios mobdeDevice buldnDevice External hard disc Y M M M Y Y Y Y Y Y 4 Y Y Y z z Y Driver Driver Driver Orivor Driver Driver Driver Driver Driver Driver Driver Driver Driver Driver Driver Orivor Oriver Driver 1 2 1 2 1 2 1 2 1 2 1 2 Drive X Drive X Drive X Drive X Y lt gt s pictures pictures pictures v CAN y Figure 49 Adapted folder structure required for CAN data stored as mat files The used structure supports a division of the data into is conditions Each driver ID can occur in each condition once The mat files in the different conditions can differ in number of variables The following structure is used in the German 2 OC DAI DAI l BMW BMW DAI Baseline mobileDevice buildinDevice BMW Baseline mobileDevice buildinDevice X v y 7 x 7 7
44. required infrastructure and services for data storage 3 Develop analysis tools and general quality assurance 4 Support SP2 and SP5 in installation and maintenance strategies see also Figure 4 These objectives of SP3 are mapped into six work packages within SP3 Figure 1 The interfaces between the three main work packages are shown in Figure 2 MARIE Management WW iseielem Provide the best suited data acquisition system DAS for euroFOT ikte Provide the required infrastructure and services for data storage wi Develop analysis tools and general quality assurance ken Perform small scale testing and validation ki Interaction Other SPs in euroFOT Figure 1 SP3 objectives and work packages Database Upload Search Interface Interface WP 3300 wP3400 WP 3500 Data Data Data Acquisition Storage Analysis cee Data Server Workstations Figure 2 Data management chain considered in euroFOT SP3 Deliverable D3 3 lt Version 1 0 gt 2 Data management in euroFOT euroFOT 01 08 2011 The role of SP3 within the FOT chain introduced in FESTA is shown in Figure 3 In order to achieve the objectives of SP3 collaboration with other subprojects within euroFOT was essential see Figure 4 Within euroFOT functions and hypotheses are the main focus of SP2 while performance indicators study design and measu
45. required unsuitable and specific cameras and used proprietary file formats or poor compression methods such as Deliverable D3 3 lt Version 1 0 gt 26 Data management in euroFOT euroFOT 01 08 2011 MJPEG None of the systems reviewed allowed precise synchronization through RS232 which was the only available port on the CTAG datalogger lt was therefore decided that CEESAR would develop its own video logger The result is based on Axis Communications Ethernet video loggers which have the following advantages e They use the very efficient H 264 compression e They accept composite video inputs which allows using any standard analogical camera e Their network capability allows using any combination of encoders on the same network to use as many video channels as needed Currently a single Q7404 is used which allows for the 4 channels which were required Video logging itself is ensured by a NEXCOM VTC3300 VTC 3300 2007 embedded PC running Windows XP and an application programmed using Axis Communications SDK The cameras which are used are standard analogical video surveillance cameras with the following characteristics e 1 3 CCD Sony Black amp White SuperHAD Exview Image Sensor e Interchangeable lens f 2 aperture e 3cm wide enclosure embedding all electronics Those were chosen after an extensive review for their great efficiency vs price ratio Different cameras were tested in various driving and illumi
46. requirements 1 It must be possible to order the DAS with the possibility to send out CAN frames physically disabled on three of the four buses 2 The DAS must support being powered from both 12V cars and 24V trucks electrical systems also during engine crank This implies that applicable automotive standards like 50 7637 should be met The supplier must guarantee that the DAS will work properly when powered by the electrical system in a car or truck 3 The operating temperature range of the DAS must be 20 C to 55 C or better 4 The DAS must not consume more than 10 mA at 12V in off mode 5 The DAS must shut down automatically if the battery voltage of the vehicle falls below a certain limit The threshold level must be configurable in software or in the configuration file for the DAS 6 The supplier must provide all necessary mounting equipment and cabling necessary for a full installation This must be provided for three different car models preliminary Volvo XC 70 XC60 and V50 and two different truck models preliminary Volvo FH and FM 7 The DAS must have a method for the driver to turn off and on the video recording of the vehicle s interior desirably while the logger should continue to collect all other data and video There must be an indicator telling if the video is on or off or that the driver can be assured that the video is turned off in another way 8 The main DAS unit must fit into a A cylinder that is 63
47. s most popular open source database 2011 Retrieved 08 08 2011 from http www mysql com NTFS com File System Structure Recovery Software Hard Disk Internals 2011 Retrieved 08 08 2011 from http www ntfs com Oracle Hardware and Software Engineered to Work Together 2011 Retrieved 08 08 2011 from http www oracle com index html Parallel Computing Toolbox MATLAB 2011 Retrieved 08 08 2011 from http www mathworks com products parallel computing Pea R Mills M Rosen J Dauber K Effelsberg W amp Hoffert E 2004 The Diver Project Interactive Digital Video Repurposing IEEE MultiMedia Volume 11 Issue 1 54 61 Deliverable D3 3 lt Version 1 0 gt 122 Data management in euroFOT euroFOT 01 08 2011 Saint Pierre G Kessler C amp Malta L 2011 Deliverable D2 2 Report of the results of adaptation in vehicle implementations and piloting Version 3 2 euroFOT Selpi Kanstrup L Val C Benmimoun M amp Borgen S 2009 Deliverable D3 1 Selection Procedure for Data Acquisition Sensors and Storage for euroFOT Version 1 1 euroFOT SourceForge net videolO Toolbox for Matlab 2011 Retrieved 05 30 2011 from http videoio sourceforge net Statistical Analysis Software SAS STAT 2011 Retrieved 05 30 2011 from http www sas com technologies analytics statistics stat index html tortoisesvn tigris org 2011 Retrieved 05 30 2011 from http tortoisesv
48. set up In the case of the enumerative check it is made sure that the signal values match up to the specified enumerative signal values and if the data was collected during baseline and the system is supposed to be turned off during baseline it is checked that all the signal values corresponds to the off value Also in the case of for example the BLIS signal which is a warning it is checked so that it is not constantly giving a warning Except for the above mentioned per measure quality checks there are a few exceptions as for example for the GPS data where the per measure quality is aggregated from the per sample quality vector by setting a threshold on allowed data loss from this sensor Another exception is the vehicle soeed which basically goes through all the steps of the analog check mentioned above but also if the per measure quality of the GPS speed is Deliverable D3 3 lt Version 1 0 gt 76 Data management in euroFOT euroFOT 01 08 2011 good has to fulfil some requirements on time lag and minimum correlation with the vehicle speed Further exceptions are the eye tracker signals that undergo per measure quality checks that are slightly different from above here the per measure quality is more based on thresholds of minimum time the eye tracker data is available during the trip 3 Per trip quality check The per trip quality is a tool for quick identification of trips that should be excluded from analysis and are based on thresho
49. signal browser will plot a preview of the signal whose cell has been selected by the user taking into account the plot command specified by the plot type menu After specifying appropriate values for all the elements of this module the user can plot visualise the selected signals by pressing the button Done Deliverable D3 3 lt Version 1 0 gt 103 Data management in euroFOT euroFOT 01 08 2011 lt A Signal browser FOTware aao Signal name Source Signal selection vehicleSpeed CAN 2 Mliongtudinalacceleration CAM 3 isteralAcceteration CAN Target module 4 DaccelPedalPosition CAN 5 MlbrakePedaiPosition CAN 6 C gearLevel CAN 7 7 Jsteeringngle CAN CAN 9 Mlirectionndication CAN 10 Jodometer CAN 11 JengineSpeed CAN 12 mainBeam CAN 13 lwiperFront CAM 14 JambientTemperature 15 lincarTemperature 16 Himeindex H Show previews Figure 75 The signal browser showing the preview of the first signal of the table 6 Signal viewer The signal viewer receives input from the signal browser and the event browser that specify the group of signals and events that should be plotted It also presents a draggable doted red line that works as a time slider that affects all time dependent modules Such as video viewers A single signal viewer can have several signals and events plotted at the same time Figure 76 shows a signal viewer in which three signals have been plotted It is po
50. software tool 2011 and 2 paper based The paper based questionnaire data is first scanned and converted into digital format and then imported into LimeSurvey tool by manual data entry procedure 3 2 4 2 Lessons learned The main lessons learned from the Italian FOT in terms of management of questionnaire data are 1 It is very important to develop and improve a web based survey tool for data collection in order to save time and prevent data transcription mistakes 2 In order to reach more drivers and improve response rate it is very good to duplicate or offer multiple options for filling questionnaires in i e hard copy and electronic copy Deliverable D3 3 lt Version 1 0 gt 84 Data management in euroFOT euroFOT 01 08 2011 4 Data analysis tool Data analysis in euroFOT is supported by the tool developed by WP3500 The tool is an application with graphical user interface GUI aimed for browsing replaying visualising and annotating recorded FOT data over time for one trip at a time The purpose of this chapter is to describe the results for the base software for analysis of field operational vehicle data New data collection abilities created exciting scientific possibilities and analysis challenges due to the scale and richness of data that is collected For example coding and careful analysis of video data remains extremely time consuming Other forms of data such as sensor readings while useful as separate data so
51. specific text can be added that describes each particular event not used much the EventTypelD unique database ID related to the name and the EventID unique DB ID for this particular event Note that the EventID is unique also between VCC and Volvo with events created at VCC with a 1 as last digit 2 for Volvo and 3 for Chalmers oEventinto A struct with one field per event name in oEventData The following information is stored Description the source script name that created creator and creation time the EventTypelD and information if it is a single point event or if it has duration start stop voHistory Not used for pre processing only for history information about what scripts has been applied and who applied them when a dataset is extracted though a query for analysis The Pre processing Step The pre processing see Figure 45 starts with the definition of a MEPS document for each OEM MEPS is short for Measures Events Performance indicators and Situational variables The MEPS is a Microsoft Excel document that includes all configuration items needed for the pre processing The MEPS document is then used to dynamically create vehicle specific oPreProCfg m files by MATLAB parsing the MEPS document oPreProcCfg is a configuration file unique for each individual vehicle and is the core in the pre processing everything defined in MEPS is individualized in the oPreProcCfg mat and then used to define pre processing for each
52. tested separately by the supplier and an EMC consultant The tests were successful Real driving test The DAS was installed in one truck and one car which were driven for some months software development was carried out in parallel with the in vehicle pilot tests Each new function was tested when delivered from the developers Data was extracted regularly and examined both manually and by Deliverable D3 3 lt Version 1 0 gt 37 Data management in euroFOT euroFOT 01 08 2011 Test type Comments software scripts see below Special attention was put into verifying that the DAS starts up and shut down properly as the ignition is turned on and off The shut down must work in order not to drain the battery in the vehicle The PC has an extra functionality watchdog to ensure proper shut down also in case of a severe software failure This function was tested and worked well Video quality During the pilot video quality was checked by offloading and manually reviewing data as often as possible An error was found where interlace effects occasionally could be seen in the recorded video After discussions with the manufacturer of the video grabber card a workaround was implemented in software by changing the way the card was initialized Image quality for the driver view is negatively affected by the limited dynamic range of the selected camera Extra IR illumination was added to mitigate the effect and showed a slight improvement Tim
53. that have been successfully used for FOT 1 CTAG Datalogger II CAN only logger 2 CEESAR videologging system a combined CTAG Datalogger II and video logger based on Nexcom VTC3300 with radar eye tracker and lane tracker 3 SAFER euroFOT DAS an integrated CAN and video logger based on Nexcom VTC6100 with external accelerometer and eye tracker 4 BMW DAS a CAN and video logger based on network attached storage with two radar sensors and 5 Daimler DAS a CAN video and audio logger Three implementations of web based DAS vehicle monitoring tools for quality assurance The required infrastructure and services for data storage including pre processing and uploading tools The solutions chosen vary depending to the expected total size of data The video data are stored as files in file management systems The non video data are generally stored in relational databases Oracle MySQL and MS Access are used At some partners the non video data are also stored as MATLAB files in the file management systems to allow analysis through database and or direct with MATLAB Three implementations of the base data analysis tool that allows researchers to visualise time based data from multiple sources and manually or automatically annotate the data A set of requirements for the design of data acquisition systems and sensors Fulfilling this new set of requirements in a production scale would further advance the state of the art in data acqu
54. the segment info and the measure info in the data base structure The verification has led to detection and correction of a couple of small but significant errors in the pre processing It is of course very important to resolve all problems early in order to avoid the need to rerun the pre processing on very large amounts of data later Currently the data verification results show very good similarity between raw and processed data for all measures Pre processing quality check While the aim of the pre processing data verification is to make sure that the data is not majorly altered in the pre processing the aim of the pre processing quality check is to make sure the data coming from the sensors are of good quality This is done by comparing the signal to typical signal characteristics for example check that the signal is within the sensor range or the natural range limited by the laws of physics and other external factors and creating a quality measure value that is uploaded to the database This quality measure can then be used to identify potential problems with the sensors and to help the analyst to exclude poor quality data As mentioned above there are three levels of quality checks per sample per measure and per trip 1 Per sample On this level of the quality check a quality value is created for each sample of the signal by checking that the signal value is within a reasonable range and forming a per sample quality vector Thi
55. troubleshooting lt Version 1 0 gt 22 Data management in euroFOT euroFOT 01 08 2011 2 2 2 French VMC 2 2 2 1 Initial condition Loggers CAN Only CAN Extra CAN Video Vehicles sensor s Extra sensor s Renault Cars 35 5 From the beginning of the project French partners wanted to have two levels of instrumentation for two different types of cars The first level initially called CAN only would have consisted of a single CAN datalogger used in participants personal cars The second level initially called CAN Video and Extra Sensor s required CAN logging video logging and a radar sensor to monitor headway This instrumentation level would be used in cars which would be lent to participants for some periods The reasoning behind those choices was that video was mandatory to validate classification algorithms during the analysis but that once robust enough those algorithms could be used with confidence on vehicles with simpler instrumentation where this visual validation would not be possible It also resulted from the fact that fitting a whole fleet with a complete instrumentation would be much too expensive However during the design phase of the project it appeared that studying a longitudinal regulation support system without monitoring headway on all vehicles would be quite limiting It also appeared for some hypothesis that measuring drivers glance and vehicles lateral position in lane would be
56. user starts analysing and understanding the data 4 2 2 2 Tests performed and quality assurance FOTware releases and upgrades Some details of the releases of different versions of FOTware can be summed up in the following list Version 1 0 This release included the data player module the file browser signal viewers with embedded signal browsers and video player modules Deliverable D3 3 Data management in euroFOT lt Version 1 0 gt 106 euroFOT 01 08 2011 e Version 2 0 Besides fixing some bugs and optimising some routines of FOTware v1 0 the data property monitor modules for both CAN and eye tracker signals were added in this version e Version 3 0 The updates in this release include an independent and more functional signal browser an event browser and video annotation modules e Version 4 0 This release includes an SQL module that handles the connection to an Oracle database and also manages the data retrieval procedures Management of improvement suggestions and bug fixing All FOTware users were are able to provide feedback to the tool responsible group TRG by using the short template report Template for reporting bugs and suggesting improvements for FOTware that is described in deliverable D3 2 All bugs that were found in each version were fixed for the release of a subsequent version and also as much suggestions for improvement as possible were taken into account Help and support The help a
57. was realized that because of the huge number of single files copying the data from Daimler conversion which uses ext3 ext3 Wikipedia the free encyclopedia 2011 to IZVW server which uses NTFS NTFS com File System Structure Recovery Software Hard Disk Internals 2011 file format took longer than measurement time As a consequence the format had to be changed from single frames to video format and the database and data analysis tools had to be adapted Especially the adaptation of the video tool took a lot of effort Now the tool works with both single frames and video format Future Recommendations The whole chain of data pre processing upload and analysis should be tested if possible until Pl calculation before the FOT Some problems with the signals e g strange values were realised very late in the whole process It has to be taken care that all partners who work with the raw data use the same software versions This is important since it was found that many problems such as opening mat files and playing video data were occurring because the versions of MATLAB used by the different partners differed 4 2 4 Italian VMC The Italian VMC does not use elaborate data analysis software like the other VMCs since this VMC only has questionnaires data However the Italian VMC uses standard tools such as LimeSurvey LimeSurvey the free amp open source survey software tool 2011 Microsoft Excel and SPSS IBM SPSS software for predict
58. well as detailed information on environment scenario etc A sample report is shown below Full Text Bug Listing Mozilla Firefox MAX Datei Bearbeiten Ansicht Chronik Lesezeichen Extras Hilfe A Full Text Bug Listing _ RE amp lele 2 te Meistbesuchte Seiten Y ZFMHYBR 212 Imp gt Versionsbezogener Su Transparent drop sha i Stream Based Encrypt rs Google Ergebnis F r h amp Setting Up an NLS Env How to Backup a Linu Ga Getting the NET Fram JIRA 3 0 Release Not ict AppletHowto c t Bot i Bugzilla Full Text Bug Listing Home New Browse Search PY Search Reports Preferences Administration Help Log out kuefen ika mvth aachen de Bug 44 Summary MatLab library uses lot of memory leads to out of memory exception Product euroF OT Central Management System Reporter J rg K fen lt kuefen ika rwth aachen de gt CMS Component Common Assignee J rg K fen lt kuefen ika rwth aachen de gt Status ASSIGNED Severity critical CC assaad ika rwth aachen de clement val ceesar gmail com fahrenkrog ika rwth aachen de ghaouty ika rwth aachen de hennick ika rwth aachen de kuefen ika rwth aachen de mbenmimoun ika rwth aachen de vkaff iccs gr Priority High Version unspecified Target Milesto Hardware PC Os Windows Whiteboard SVN Revision 82 Time tracking Orig Est Actual Hours Hours Wo
59. 0 gt 17 Data management in euroFOT euroFOT 01 08 2011 e Diagnosis capabilities added As a first approach the diagnostic capabilities were hard coded in firmware In particular a flow control according to ISO 15765 2 Diagnostic TP layer standard was implemented for the transport layer It allowed establishing the necessary flow control to perform DIAG ON CAN with several ECUs at the same time Currently by python scripting the user should be able to perform DIAG ON CAN functionalities according to current standards e synchronization when engine off e Firmware update through SD card or remotely via GPRS e Storage data format open non proprietary The format was documented and sample source code was provided e Two operation modes are possible ignition signal dependant power on off and CAN activity triggered power on off e Modifications on Server tool Storage and Upload Procedure As the German 1 operation centre decided to have a management system implemented on the data server a solution based on the separation of the CTAG Server tool functions was preferred This automatically initiates the process for data upload pre processing and storage on the data base As a consequence some of the functions of the server tool were separated in the form of Phyton scripts which enables the user to integrate and initiate the functions according to the definition of the process chain in the management system The changes to
60. 0 mm in diameter and that has a height of 140 mm for cars b A box that is 320 mm x 150 mm x 400 mm for trucks 9 Cameras should be as small as possible It is expected that three of the cameras will be mounted by the rear looking mirror and they should obscure the driver view as little as possible 10 All equipment must be CE marked and also pass a number of additional tests or fulfil a number of OEM s EMC requirements Deliverable D3 3 lt Version 1 0 gt 32 Data management in euroFOT euroFOT 01 08 2011 2 2 3 2 Chosen solution Based on the requirements two different options were considered 1 Develop a system based on PC technology and commercial off the shelf components COTS 2 Develop or buy a dedicated non PC based solution that is developed for FOT purpose While the option to buy a dedicated non PC based solution seems very easy and attractive our market survey and the result of our request for information RFI in 2008 suggested that there was no readily available commercial system on the market fulfilling all of the requirements specification and the budget constraints of the Swedish VMC Indeed there were very few choices of full logging system available on the market not to mention those designed for large scale field operational test Therefore it was decided to also focus on a new custom made DAS PC based solution designed to fulfil euroFOT Swedish VMC requirements specification For this a furth
61. 11 corrected referencing of resources for google maps control which was not using the resx builder in the GUI components project 541 ghaouty 18 10 49 Montag 23 Mai 2011 corrected behaviour when invoking connected callback of client socket 540 ghaouty 18 06 31 Montag 23 Mai 2011 corrected wrong port assignment in diagnostic interface service 539 kuefen 15 02 28 Montag 23 Mai 2011 added a port redirector and a svn integration for visual studio 538 kuefen 15 16 14 Sonntag 22 Mai 2011 added a possibility to define execution context for a service supported by arguments useraccount and userpass 537 kuefen 17 15 15 Freitag 20 Mai 2011 535 kuefen 10 48 18 Freitag 20 Mai 2011 new script to reset 523 kuefen 09 08 19 Donnerstag 19 Mai 2011 corrected a bug For selection the output naming schema files and the trimming of identifiers for injected arguments 519 kuefen 20 53 05 Mittwoch 18 Mai 2011 update conversion tools with naming schema and data injection mechnism 517 kuefen 16 14 35 Mittwoch 18 Mai 2011 added new dtl to mat converter to system 516 kuefen 16 09 05 Mittwoch 18 Mai 2011 extended trip general attributes to distinguish between records made by the logger and trips as segment in the logger by adding files TripNoInRe 515 mohamed 15 59 08 Mittwoch 18 Mai 2011 corrected conversion batch 514 mohamed 15 43 30 Mittwoch 18 Mai 2011 New version 512 kuefen 15 25 27 Mittwoch 18 Mai 2011 new version of conversion tool to suppo
62. 12 months CAN Only months Questionnaires CAN Video Extra sensors only Figure 5 The VMCs in euroFOT and the FOT plan in July 2009 This figure is adapted from D5 1 lt Version 1 0 gt 5 Swedish VMC 100 cars 30 trucks 12 18 months CAN Video Extra sensors euroFOT 01 08 2011 1 2 Overall Way of Work Each of the VMCs had planned to evaluate different set of safety functions and thus different set of hypotheses To answer the different set of hypotheses each of the VMCs needed different types of data and thus different solutions for collecting storing and processing the data Furthermore each of the VMCs had different constraints with regard to the kind of equipments that can be installed in the vehicles All of these differences created a complexity On the other hand these differences give us a richer experience for example in terms of using different systems and richer lessons learned Common method s was used as much as possible SP3 partners worked both at euroFOT level as well as at VMC level The tasks that could be carried out together and or could benefit more than one VMC were shared among partners and done together at euroFOT level However due to the differences described previously there were also some work i e to find specific solution s for each VMC that have had to be done at the VMC level This allows each VMC to concentrate their resources on different tasks Internally SP3 partners worked togeth
63. 3 3 lt Version 1 0 gt 60 Data management in euroFOT euroFOT 01 08 2011 Table 14 Estimated size of video data in the Swedish VMC OEM Vehicles Hours per week Video data size GB data per week Size for 12 months GB MB per h Volvo 30 30 700 630 32 760 VCC 100 10 700 684 35 547 Total 130 1314 68 307 Uncompressed data Mainly based on the size estimation the Swedish VMC has chosen the following solution for storing the FOT objective data e upload the non video data as per trip basis to an Oracle database e upload the video data as files to a fileserver To connect the non video data to video data a link to an associated video file of a trip is stored in the Oracle database as part of each trip s record The rest of this section will discuss the processing and upload of objective data The questionnaires data are planned to be stored in the Oracle database tables have been prepared see Figure 37 and the fileserver saved as files like MS Excel The overall upload task consists of several steps Generally the flow of FOT data from the vehicle s to the Swedish VMC centre can be described as in Figure 36 From the technical implementation point of view the overall upload task can be divided into two parts pre processing and uploading The two parts are defined as follows e Pre processing PP as the step between when the data being available in an office at an OEM in the form of USB discs directly from the
64. 5 GPS samples sec 17 5k signal samples sec 17 5 GPS samples sec cr 40kB sec 120kB sec oar entire car fleet a 4 overall around 12TB data eee gt j forone year of FOT i T lt A 34 5k signal samples sec lt 35 GPS samples sec 80kB sec entire truck fleet Figure 27 Data amount estimation for the German 1 OC The data estimation for the German 1 OC only refers to CAN recorded data and position information GPS It is differentiated between trucks and cars as the usage and activity levels for both vessel types are expected to be different For both vessel types around 60 to 100 signals are being recorded Depending on the volatility of the recorded signal different sampling frequencies were chosen to reduce the amount of data being recorded Trucks are estimated to run approximately 10hrs working day while cars will operate around 3hrs day Figure 27 shows the results of the overall estimation normalized to a continuous data stream of 24 7 Both VMCs have decided to reduce the amount of stored data by supporting different sampling frequencies for the signals in the field and on the storage system and to avoid explicit oversampling This approach offered the opportunity to reduce the amount of data storage needed to around 20 compared to a solution where all data is stored with the highest sampling frequency from field Geo information Data Acquisition Upload and monitoring GPRS connection Elementary
65. 512 persistence BOOLEAN tablename VARCHAR 45 m id INT name VARCHAR 45 description VARCHAR 512 unit VARCHAR 145 display BOOLEAN description VARCHAR 512 type VARCHAR 45 manual_edit BOOLEAN manual_edit_mechanism VARCHAR 45 need_refresh BOOLEAN duration DOUBLE length DOUBLE columnname_1 EventAttributeType DOUBLE columnname_2 EventAttributeType DOUBLE columnname_N EventAttributeType DOUBLE Figure 30 The left part of the database schema at German 1 OC and French VMC Deliverable D3 3 lt Version 1 0 gt 53 Data management in euroFOT euroFOT 01 08 2011 id INT 11 name VARCHAR 45 tablenameid INT scriptfile VARCHAR 128 occurencesvariable VARCHAR 128 columnname VARCHAR 45 type VARCHAR 45 variable VARCHAR 85 unit VARCHAR 150 comment VARCHAR 512 display BOOLEAN columnname_1 SituationsCalculationsAttrs DOUBLE i description VARCHAR 512 columnname_2 SituationsCalculationsAttrs DOUBLE columnname_N SituationsCalculationsAttrs DOUBLE id INT situationtypeid INT variable VARCHAR 45 modality VARCHAR 185 name VARCHAR 45 name VARCHAR 45 description VARCHAR 512 Indexes Indexes Figure 31 The right part of the database schema at German 1 OC and French VMC Deliverable D3 3 lt Version 1 0
66. 6 CTAC S Datalogger Weer seek ee eo 15 Figure 7 HW Architecture for CTAG Datalogger v1 0 and v2 0 16 Figure 8 TYCO Connector CANAAN EREEREER ARORA NDERI ESAE AAEREN ANR RNANA 17 Figure 9 Connectors of CTAG s Datalogger II VS 19 Figure 10 Web based monitoring tool used by German 1 OC and French VMC 22 Figure 11 Low level instrumentation at French VMC 4 25 Figure 12 AG 20 TA0d 28 na a e a a d a o 25 Figure 13 High level instrumentation at French VMC 25 Figure 14 Video logging system at French VMC 26 Figure 15 Mobileye AWS2000 ii enseenereneneennnss 28 Figure 16 French VMC Placement of extra sensors and cameras 28 Figure 17 Swedish VMC Placement of DAS including extra sensor and cameras 35 Figure 18 The main page of Swedish VMC DAS vehicle monitoring tool 39 Figure 19 Communication of the data logger components 41 Figure 20 ZGW connections vctecicerscncecatataccsatosnnwaetascincetannsenn hod nersetnsniwne bdaueenbanetemoheiaensetescien 42 Figure 21 Placement of DAS in a BMW vehicle 42 Figure 22 The architecture of DAS used by Daimler 43 Figure 23 Placement of DAS in a Daimler vehicle VU 44 Figure 24 Web based monitoring tool at Daimler 45 Figure 25 The common process for database and data storage 46 Figure 26 Classification of French VMC and German 1 OC FOT scenario 49 Figure 27 Data amount estimation for the Ger
67. 6 The uploading steps in the Swedish VMC NAVTEQ s ADASRP a Windows based framework application for prototyping ADAS and navigation solutions is used to complement CAN and other sensor data from the vehicles with map data This is done during the pre processing of the logged data A Volvo ADASRP plug in GPSToRoute reads a file containing GPS positions latitude and longitude and then finds the most probable route given the NAVTEQ map using Dijkstra s algorithm Interesting map attributes along this route are then extracted or calculated and exported to a cvs file where every original position is followed by the corresponding map attributes The NAVTEQ map is divided into several hierarchies and the GPSToRoute plugin deals mostly with links A link can be a road several connected roads or a part of a road Each link consists of two or more shape points which define the geometry of the link The end points of the link are called start node and end node and they are sometimes also called node 0 and node 1 Thus when link direction is 1 it means the vehicle is travelling from node 0 to node 1 GPSToRoute is mosily concerned with attributes on the link level for each raw GPS Deliverable D3 3 lt Version 1 0 gt T3 Data management in euroFOT euroFOT 01 08 2011 position it maps the corresponding link and extracts those attributes For euroFOT 60 map attributes are collected including speed limit road type number of lanes and dista
68. 9 Vehicle Loos FOT217 0 VTEC ne Nemenancs 4857 126 2011 08 28 16 4444 enecte Los FOI22 C8 VTEC The Netertancs 5892 Reset 1289 Cb 290 03 Gt 100 100 2011 06 27 06 37 03 Vericie Loos FO1224 OF VIEC The Neterancs Reset 73334 FOT221 0 VTEC The Netenands 1744 Reset 1209 Gb 188 94 Gb 100 100 2011 08 28 174100 Vehicie L FOT222 E VTEC The Nethertands Reset 119210 24189 Gb 100 100 2011 06 27 644139 25 FOTZ20 CE VTEC The Nemerands 4331 Q Reset 2011 09 09 06 03 40 Labiche Lons FOI221 SIVIEC The Neteriancs 78 gt Rest 12 tit t FOT227 T VTEC The Neertands 4934 Reset 1289Cb 43508 Gb 190 100 2013 06 07 2219 11 Verecte Loos Figure 18 The main page of Swedish VMC DAS vehicle monitoring tool 2 2 3 5 Experiences and lessons learned Inbuilt diagnostics with remote status reports is a necessary tool to follow up the status of the DAS systems Connectors will fail as do cameras PCs IR illuminators HDDs etc If vehicles are operated for a long time before data HDD is collected there is a big risk of losing a lot of data before an error is discovered Note that only the status reports are sent wirelessly to the Swedish VMC centre the collected data in the HDD is collected manually by OEM personnel see Figure 36 on page 62 Failing connectors have been the main source of errors The reason has been due to human factors where the personnel assembling and handling connectors have not been skilled educated trained enoug
69. E triplD 100 AND vehicleSpeed 30 nae eae i B MEASURE TYPES ORDER BY triplD timelndex Mem MB 3 69 1434 96 i Tipes 406 HE4 PROPERTIES k F PROPERTY TYPES Segmentize data Expandsegments Prep No rows sel 30174 eA VEHICLES Executing SQL statement Segment h if SURVEYS Done SURVEY ANSWERS Populate table 4 HE9 SURVEY QUESTIONS gt Max Mo rows 100 E SURVEY _QUESTION_TYPES vehiclespeed longitudinal4cceleration lateral4cceleration accelPedalPosition brakePedalPosition gearlevel steering va 30 8600 0 6500 0 2950 17 7000 3 8000 gt l E Bees 31 1400 0 5450 0 2950 17 7000 6 6000 us m 31 2700 0 6500 0 0200 17 9000 8 8000 D vehicleSpeed 31 2700 0 6500 0 0200 17 9000 5 8000 SN iii T 31 9400 0 4400 0 2950 17 9000 5 8000 i aceon 32 0700 0 4400 0 0550 17 9000 5 8000 D accelPedalPosition 32 2800 0 5450 0 5400 17 9000 5 8000 _ brakePedalPosition 32 2800 0 5450 0 5400 17 9000 2 8000 E gearLevel w 32 7100 0 6500 0 2600 17 9000 6 8000 Figure 71 A depiction of the SQL module after successfully executed an SQL statement Bi Mi ES M M M F5 d 5 1 1 1 1 1 j ho a ES When the user opens the SQL module he is prompted for DB credentials and needs to provide his her username password the DB name and URL These two last items will be configured when installing FOTware and will be available by default However the user is able to introduce different values in c
70. European Large Scale Field Operational Tests on In Vehicle Systems 7th Framework programme INFORMATION AND COMMUNICATION TECHNOLOGIES IC T 2 6 2 ICT for Cooperative Systems Large scale integrating project Grant agreement no 223945 lt LEE ef Bringing intelligent vehicles to the road Deliverable D3 3 Data management in euroFOT Version number lt Version 1 0 gt Dissemination level lt PU gt Lead contractor lt CHAL gt Due date lt 31 08 2011 gt Date of preparation 01 08 2011 euroFOT 01 08 2011 Authors Selpi Chalmers tekniska h gskola Samuel Borgen Chalmers tekniska h gskola Jonas B rgman Chalmers tekniska h gskola Erik Svanberg Chalmers tekniska h gskola Marco Dozza Chalmers tekniska h gskola Rasmus Nisslert AB Volvo Catarina Norell AB Volvo Jordanka Kovaceva Volvo Cars Corporation David Sanchez CTAG Marga Saez CTAG Clement Val CEESAR J rg K fen IKA Mohamed Benmimoun IKA Barbara Metz IZVW Project Co ordinator Aria Etemad Telematics amp Navigation Research Ford Research amp Advanced Engineering Europe Phone 49 241 9421 246 Fax 49 241 9421 301 Email aetemad1 ford com Ford Forschungszentrum Aachen GmbH Suesterfeldstr 200 D 52072 Aachen Germany Copyright euroFOT Consortium 2008 Copyright on template euroFOT Consortium 2008 Deliverable D3 3 lt Version 1 0 gt i Data management in euroFOT euroFOT 01 08 2011 Revision and history chart Version Date Reaso
71. INT columnname V ARCH AR 45 name VARCHAR 45 unit VARCHAR 145 display BOOLEAN description VARCHAR 512 type VARCHAR 45 manual _edit BOOLEAN manual _edit_mechanism VARCHAR 45 need_refresh BOOLEAN columnname_i EventAttributeT ype DOUBLE columnname_2 EventAttributeT ype DOUBLE columnname_N EventAttributeT ype DOUBLE m Vid INT name VARCHAR 45 description VARCHAR 512 Indexes Vid INT tripID INT eventiD INT annotation VARCHAR 512 Deliverable D3 3 Data management in euroFOT Vid INT situationtypeid INT variable V ARCHAR 45 modality VARCHAR 185 tripID INT time FLOAT NOcolumnsD efinedInTDM ENUM Indexes Vid INT name VARCHAR 45 frequency FLOAT description VARCHAR 512 Vid INT signaname VARCHAR 45 2 tablename V ARCHAR 45 columnname V ARCH AR 45 unit VARCHAR 150 datatype VARCHAR 45 resolution VARCHAR 45 defaultvalue DOUBLE minvalue DOUBLE maxvalue DOUBLE Vid INT name VARCHAR 45 scriptFile VARCHAR 64 1 author VARCHAR 45 mr description VARCHAR 128 comm ent VARCHAR 512 Vid INT procID INT name VARCHAR 45 type VARCHAR 45 variable VARCHAR 45 composition BOOLEAN vdue DOUBLE Vid INT tripID INT type VARCHAR 45 starttime DOUBLE endtime DOUBLE duration DOUBLE
72. Offset Color Style Insp Cost Spaad 7 CC SL_State 100 lO A a Vehicle_Speed 7 CC SL_Speed kmh 1 Of E ee 7 CC SL_Overspeed 1 0 F Dee J Distance_Totalizer km 1 mm PF P V Vehicle _Speed kmh 1 mm v Vehicle_Longi_Accel ms2 1 0 pete a Vehicle_Trans_Accel g 1 mmm _ Left p ane Clutch_Switch_Max_Travel 1 mm F Clutch_Switch_Travel 1 mm _ Brake_Info_Status 1 mm _ wv pos bie Gee LA eS ee ee E Se i ee a i Throttle Ar 1 mm v Steering_Wheel_Angle deg 1 mmm _ uv Engine _RPM T l 1 mm Vv PU Le Fuel_Consumption mm3 1 0 ee R_Yaw_Rate de 1 0 mz v R_RelativeVelocity ms 1 EE Satellite V R_Range m 1 200 _w W penetra ee ee ptet R_Acceleration ms2 1 0 D pe R_CorrectedAzimuth deg 1 0 _ R_RelativeAzimuth deg 1 ol __ D101 R_LateralDistanceTrack m 1 0 iz _ L DistanceToLeftLaneMark m 1 mm _ v DistanceToRighttLaneMark m 1 O Vv s LeftLaneConfidence ee Be p N 50 RightLaneConfidence 1 0 a 4 GPSSpeed kmh 1 mm _ GPS de 1 0 az ed meanSpeed 88 8 kmh blis L _ oies GPSLongitude dex mmm RE Cr GPSBearing ds 1 0 a StdSpeed 24 kmh pasa Map_MatchedLongitude de i mmm _ ee 2 4 m g Map_MatchedLatitude d 1 of HSCS oon ais ele Map_MatchedHeading de 4 om PretsLUse 0 0 1 D177 s Map_NewLink 1 om e Duration 204 s Map data 2011 Google V Map_SpeedLimit kmh 1 oM v 1000 1500 2000 2500 3000 350 Event or situation panel Figure 54 Trip Viewer Visualisa
73. SpeedChange 2033 2034 isSL false E CCSLSpeedChange 2037 2037 meanSpeed kmh E CCSLStateChange 2110 2110 KeyUsed 2 StdSpeed re E CCSLSpeedChange 2158 2160 Sas i E CCSLStateChenge 2158 2158 meanHeadway i m E CCSLSpeedChange 2164 2168 Se a p or wae Sie PretCCUse 0 1 E CCSLStateChange 2169 2169 E CCSLStateChange 2214 2214 a PretSLUse 0 1 Duration Left panel in Events Situations mode Left panel in Trip Information mode Left panel recessed Each time the current time is modified all information on screen signal values videos vehicle s position on the map is instantly refreshed When the user switches to the Event amp Situations mode of the left panel he is presented with a tool to filter the events and situations he wishes to display in a list and as icons on the map view Deliverable D3 3 Data management in euroFOT lt Version 1 0 gt 90 euroFOT 01 08 2011 When he uses the third mode the user accesses general information about the trip i e standard status computed trip attributes driver and vehicle attributes When using smaller screens such as laptops it is also possible to recess the left panel in order to leave more room for the actual data The video display module Figure 56 displays the four camera views which are recorded by the French VMC forward driver steering column and feet Figure 56 Video display four camera views Forward Driver Steering Column
74. YNC and SSH server Reliability improvements after pilot testing The final connectors in CTAG Datalogger II are e MiniUSB connector for the usage with PC e FME connector for GPS antenna Deliverable D3 3 lt Version 1 0 gt 18 Data management in euroFOT euroFOT 01 08 2011 e SMA connector for GPRS antenna e SD card socket e SIM card socket The layout of all these connectors mounted in the data logger is presented in Figure 9 SPI Socket SD Card Trigger Lae Socket SIM Card Power CAN MiniUSB SerialPort J GPRS GPS Figure 9 Connectors of CTAG s Datalogger II The current operation modes of the DataLogger 2 are see D3 1 e MSC Mass Storage Device When USB is detected Both file transfer and data logger configuration is possible e Logger Steady state operation e Idle Mode When the data logger is powered by VBAT2 terminal pin 3 and detects CAN inactivity for a period of time configured by the user it goes into idle mode The data logger will come back to Normal mode again if CAN activity or trigger is detected or if a Wake up Alarm event is fired This mode is not used for the FOT i e no connection via VBAT2 e Power down data logger will switch off if ignition signal KL15 pin 2 is off it is powered by VBAT terminal pin 1 and KL15 is used The data logger will return to power ON when KL15 is ON The final specifications achieved in the CTAG Datalogger Il can be su
75. ability of database or server systems The NET interaction with MATLAB has been proven to be a very powerful mechanism to add functionalities to interact with the server systems Future Recommendations The modular approach of the software should be extended Especially the interfaces could be more standardized and could offer additional features like a command line interface As well the integration of additional external tools beyond MATLAB could be very useful 4 2 2 Swedish VMC The tool developed in the Swedish VMC is called FOTware and is based on a modular and customisable graphical user interface GUI The implementation described below refers to the FOTware version v4 0 FOTware development is based on previous experience in the Swedish VMC in field operational test data analysis from previous projects such as TSSFOT Chalmers SAFER TSS Naturalistic Field Operational Test FOT Phase 1 2011 and SeMiFOT Chalmers SAFER SeMiFOT 2011 In euroFOT FOTware is used for data quality assessment hypothesis testing outlier analysis and video annotation 4 2 2 1 Implementation FOTware is entirely developed in MATLAB using an object oriented programming approach FOTware is designed to be a completely modular and customisable tool This means that FOTware can be modified by the user and can offer endless number of different configurations combinations of modules to respond to the different analysis needs and to the different users
76. acker and lane tracker Third is the SAFER euroFOT DAS an integrated CAN and video logger based on Nexcom VTC6100 with external accelerometer and eye tracker Fourth is the BMW DAS a CAN and video logger based on network attached storage with two radar sensors Fifth is the Daimler DAS which is a CAN video and audio logger All of these systems have been successfully used for the data collection phase in this project In particular the CTAG Datalogger Il and the SAFER euroFOT DAS were equipped in more than one hundred vehicles driven in several European countries for more than one year Except the BMW DAS all DAS are accompanied with a web based monitoring tool to enable early detection of errors as part of quality management strategy How the collected FOT data are transferred from the DAS to the VMC centre and then processed in the different VMCs are also presented here German 1 and French VMC use full wireless transfer for the CAN data All VMCs that collect video employ manual data pick up and only use wireless to send status reports The pre processing and uploading steps are discussed at length in this deliverable The database and storage vary depending to the expected total size of data German 1 OC with an estimated 127TB of data uses Oracle for their database engine The French VMC stores their CAN data estimated to be 2TB total with video is estimated 8TB in a MySQL database German 2 OC stores their subjective data in MS Access
77. age of several data records from all the interacting entities for one single event such as road network log several vehicles DAS mobile phone applications logs Deliverable D3 3 lt Version 1 0 gt 59 Data management in euroFOT euroFOT 01 08 2011 3 2 2 Swedish VMC 3 2 2 1 Data storage solution Figure 35 shows an overview of the implementation visualizing the process from data pre processing to upload and interface to analysis amp be Di 7 New events Predictable computations Unpredictable computations Der aw Derived 7 a res l o i l l li Quality measures i l I Derived measures Resample Event amp data mark up l i i i l i Figure 35 Overview database and data storage The FOT objective data in the Swedish VMC consists of 1 video data and 2 non video data CAN data and eye tracker data Additionally there is also questionnaires subjective data Table 13 and Table 14 show the estimate of data size from each OEM for one year data collection Table 13 Estimated size of time history data in the Swedish VMC OEM Vehicles Hours per No of samples 10 Hz in No of Bytes per Size for 12 week per week 10 Hz 12 months measures value months GB Volvo 30 30 32 400 000 1 684 800 000 180 4 1 130 VCC 100 10 36 000 000 1 872 000 000 180 4 1 255 Total 130 68 400 000 3 556 800 000 2 385 Uncompressed data Deliverable D
78. agement in euroFOT euroFOT 01 08 2011 2 2 2 4 Tests and quality check performed Most of the tests and quality check performed are already described in Section 2 2 1 4 Some other tests performed on videologging are listed in Table 7 Table 7 The different tests performed on videologging of French VMC Test type Comments PC tests The potential and suitability of Nexcom VTC3300 as videologger were tested Video quality Many tests were done to check the quality of the video from different cameras tested in different driving with different illumination conditions Synchronisation The delay in synchronisation between video frames and CAN data test was measured Necessary changes were made until acceptable synchronisation is achieved Real driving tests Was done during piloting Web based monitoring tool Same as the one described in Section 2 2 1 4 on page 20 2 2 2 5 Experiences and lessons learned Simpler and cheaper cameras could be more suitable for FOT Standard although rather high end SuperHAD Exview sensor analogical black and white video surveillance cameras were found sufficient in the French VMC Colour cameras require larger storage and even the last generation cameras did not show the same sensitivity and dynamic range as the black and white cameras used The specific automotive cameras have higher dynamic range but they do not use standard connectivity and power supply and were more expensive than the black and
79. ail in a file called Var_db xls The variables and also the number of variables differ between BMW and Daimler In Var_db the following information is given for each variable e name and source e g raw log data map data e unit resolution and range for continuous variables e meaning of categories for categorical data Steps of data processing amp enrichment The pre processing routines are organized in a way that they always work on all files in a specific folder write the processed files into a new folder and delete the input files Through that it is ensured that each step is run only once on each file At the end of the data pre processing routines the files are stored in the structure described above Figure 51 gives an overview over the different steps of data pre processing enrichment and upload Logging in the vehicle I7 VW gt Format of OEM Readingin ofrel t parameters Merging per trip MapMatching Adding of map based gt NavTeq information Adding of derived parameters Adding of informationfrom OEM mobile device Saving in folder structure Figure 51 Overview of data pre processing enrichment and upload routines e Logging in the vehicle the data is logged in a format that can only be read by the OEM e Conversion to MATLAB The logged raw data is converted into even spaced time series data that is saved in mat format This step is done at the OEMs Deliverable D3 3 lt Version 1 0 gt 82 Data
80. alue from seven signals 8 Event browser Events are points in time or time intervals that have a type name and numerical level value that has a type specific meaning The event browser lists down all identified traffic events that are currently loaded With this module it is also possible to plot events as a Boolean or multi level signal and jump to the time in which they occur The user can specify the target signal plotter window in which he she desires to plot a certain event Figure 78 presents a sketch of the event browser D Event browser FOTware Event ID Stark 1053000 1113300 1423200 1511300 Event type 14 ceCarFollowing 15 ceCarFollawing 16 ceCarFollowing 17 ceCarFollawing 18 ceCarFollowing 1596100 1596700 19 ceCarFollawing 1737900 1746600 20 becRELongaces 0 4300 21 leCRELongacc 131027 4500 2028600 22 boeHardBraking 13 2725 a 4200 131027 4500 2028500 106700 112300 339600 345900 466300 472100 475000 482100 483600 490300 559300 565200 709200 715000 Target module Wew plotter Figure 78 The event browser module displaying some identified traffic events 1056200 A 1144900 1490800 1585900 23 loeHardEraking 24 oeRoling 0 25 oeRoling 0 26 oe amp Rolling oe amp Rolling oe amp Rolling oeRolling 9 Video and event annotation This module is intended to assist the user in video annotations In order to launch this module data must have been read either with the file brows
81. anagement of unique identities has become more complex 26 Pre processing and analysis of these large amounts of data requires 64 bit memory structures and software support 2 A sandbox database including a subset of the data should be created When developing and refining scripts it is much more time efficient to test on a smaller dataset When satisfied the analysis can be executed on the whole dataset 28 To be able to use SQL for analysis on billions of rows in a single table effective index needs to be created It is also vital to consider the storage needed for all indexes as indexes could take a lot of space 29 Use and analyse explain plans before executing queries By doing this badly written SQL commands could be avoided Deliverable D3 3 lt Version 1 0 gt 115 Data management in euroFOT euroFOT 01 08 2011 30 Try to fully use the power of state of the art servers utilizing many cores By using parallel techniques in the database server or splitting the queries into smaller pieces can boost the performance 31 Consider the cost for making backups of the database If raw or pre processed data is stored and has backup there might not be need for taking backups of the continuous data tables The time for re uploading data might take longer time but it is important to state the service level agreement for downtime 32 There is a great cost in terms of computing time for performing joins when dealing with l
82. and Feet Images displayed in the video module are perfectly synced with numeric data Keyframes caching when loading a trip allows extremely quick seek mode when changing current time by dragging the vertical red line When the real time play is activated the current time in the whole GUI is constantly refreshed current values in the inspectors are updated as well as vehicle s position on map and when necessary map and or time history view are automatically panned The vehicle position is displayed over a map Figure 57 The complete trip is lined in red the vehicle s position is shown as a black triangle and events and situations are represented as user selectable icons The user can change at will the zoom level and position of the map by selecting the same Zoom or Pan control as those used for time history data Clicking on an event or situation icon has the same result as selecting the instance in the left panel s list the corresponding event or situation form panel is displayed Figure 58 Deliverable D3 3 lt Version 1 0 gt 91 Data management in euroFOT euroFOT 01 08 2011 Parc de Boulogne Edmond ji de Rothshild i RO Figure 57 Vehicle s position on a map Events can be detected for each trip either by an automatic script based on continuous variables or by manual annotation using the event form The manual annotation is done when the driver self reports them or when an o
83. and galvanically separated CAN bus interfaces In order to make sure that the logger system cannot affect the vehicle in any way the CAN interfaces were ordered with the Deliverable D3 3 lt Version 1 0 gt 36 Data management in euroFOT euroFOT 01 08 2011 ability to send disabled Furthermore the requirement was that they must be galvanically separated so that there is no current path between the logger PC and the CAN bus Configuration Each DAS is configured with a unique identifier in a file This identifier is then used to tag the stored data as well as to identify each vehicle on the status web web based monitoring tool Installation For cars the logger unit is installed in the spare wheel well For trucks the logger unit is installed in the left luggage compartment Cameras are installed on four different locations in the vehicles Cars forward view driver feet rear view Trucks forward view driver feet right blind spot Dutch fleet left blind spot UK fleet Installation verification A software application was developed that shows live reading of all signal sources This gives the installer means of detecting e g missing camera signals and CAN bus errors GPRS subscription Cash cards from a Swedish operator are used These have to be refilled regularly During roaming conditions Dutch and UK fleet the software collects a number of status report before sending them to maximize the use of the minimum fee that appl
84. arge tables By extensive de normalization of the schema and even duplicating data analysis can be faster 33 Making data access based on roles proved to be positive 34 By converting video data from xvid to h264 the data storage need was reduced by 40 at Swedish VMC The process is very CPU intensive and a cluster of computer nodes were needed to perform this task 35 The large amounts of video data exceeded the filesystem limits of storage appliances Therefore a server providing logic to be able to create a single mount point was needed For future projects different filesystems and appliances could be evaluated for instance by using ZFS or equivalent for almost unlimited shares Review on management of questionnaires data 36 In order to reach more drivers and improve response rate it is very good to offer different options for filling in questionnaires i e paper based or web based However if participants agree it is recommended to use a web based questionnaire this will prevent data transcription mistakes Other reflections 37 In contrast to FESTA s suggestion FESTA consortium D2 2 Data requirements for FOT methodology 2008 to use relational database to store all non video data we thought it might be wise to be open to other available and emerging technologies like column based databases Hadoop database HBase HBase HBase Home 2011 and a data warehouse system for Hadoop Hive Welcome to Hive 2011 for mana
85. as then sent out to potential suppliers of the main DAS system and sensing equipment as well as companies that have been identified as possible system integrators during November December 2008 Since then to July 2009 the preliminary requirements specification and the list of potential suppliers were constantly being revised as more information became available After receiving the list of measures from SP4 SP3 partners worked together with SP4 to identify the necessary signals for the measures and particularly the two OEMs in the Swedish VMC identified where to retrieve those signals i e from which vehicle network the signals can be retrieved The information that was gathered from this activity was used to revise the preliminary requirements specification The rest of this section discusses the key elements of the technical requirements of the Swedish VMC Swedish VMC technical requirements System requirements 1 It must be possible to equip the DAS with up to four 4 CAN buses 2 The DAS should be able to store all data from four CAN buses 3 It should be possible to configure the DAS to store data in the following ways for all buses CAN LIN and J1708 All frames from a bus Specific frames from a bus i e filtering on frame identifier Specific signals from a bus i e filtering decoding 4 If the DAS is configured to decode specific signals from a bus there must be a secure method e g encryption to protect the
86. ase that is required It is also possible to keep the DB connection active and valid and to provide credentials only once during a complete MATLAB session by choosing to save the user password EJ DE credentials otg User name myusername Password seeseses Save it session DE name eurofotdb DEURL name place server damain cor Figure 72 A sketch of the form the user should fill in with his her database credentials With the SQL editor the users can modify existing SQL files and create new ones as well Once the editing has ended the user can transfer the text to the main SQL module just by pressing the button Import EJ FoT a editor ae SELECT Eee tripil vehicleSpeed u lateralicceleration accelPedalPosition brakePedalPosition gearLevel Bteeringa ngle yawRate directionIndication odometer engineSpeed mainBeam wiperFront ambientTemperature inCarTemperature FROM MAINTABLE WHERE tripIP 100 AND vehicleSpeed 30 ORDER BY tripID timeindex EENE import save Save As Independent window position Font style Figure 73 The SQL editor window 4 File browser This module enables folder browsing and presents the user with a list of files which are possible to be loaded from the folder Figure 74 The user opens a file by clicking on the file name A check box is also available for the user to decide whether eye tracking data should also be read The data currently loaded
87. ata information the user will be prompted for access credentials and should fill in the fields of the form shown in Figure 72 The options Segmentise data and Expand segments will complement the SQL command and change adapt the way in which the data is retrieved Import and export buttons that let the user transfer SQL statements to and from the simple SQL editor that is shown in Figure 73 Information fields that show relevant messages to the user such as the query execution status the total memory used to retrieve the data the time it took to retrieve the data etc Data table that shows the retrieved information organised in columns and limited by a predefined maximum number of rows Deliverable D3 3 lt Version 1 0 gt 101 Data management in euroFOT euroFOT 01 08 2011 E FOT DB SQL FOTing am A euroFOT DE SQL statement history HF DATA SOURCES SELECT timelndes tripl vehicleSpeed longitudinalAcceleration lateral4cceleratian accelPedalP DRIVERS 3 2 EES EVENTS Enter SOL statement s Enable SL completion fram DBS FEJ EVENT _T PES SELECT timelndex triplD vehiclespeed langitudinalAcceleration lateralAcceleration el M Save in history E accelPedalPosition brakePedalPosition gearLevel steeringAngle yawRate D fF MEASURES MEASURE_ENUMS directionindication odometer engineSpeed mainBeam wiperFront E ambient Temperature incar Temperature S eee phi FROM MAINTABLE WHER
88. atalogger II Test type Comments PC tests Test in laboratory simulating driving cycles were carried out in some cases even running for 48 h non stop Environmental The reference standard for the product validation was ISO 16750 tests Road vehicles Environmental conditions and testing for electrical and electronic equipment in its different parts e Electrical tests Supply voltage range verification 9 18 V for 12 V nominal 19 30 V for 24 V nominal overvoltage immunity 20 V for 12 V nominal 34 V for 24 V nominal and reverse voltage immunity 14 V for 12 V nominal 28 V for 24 V nominal variations and discontinuities in supply voltage and superimposed alternating voltage e Environmental storage at 85 C operating temperature range 30 C 70 C thermal shock resistance 30 70 C Humid heat up to 97 HR Tests have been performed with no remarkable physical effect on the device whose behaviour was considered as acceptable e Mechanical Vibration tests following automotive profiles mechanical shock resistance 500 m s2 pulses e Chemical Chemical tests according to 16750 part 5 Deliverable D3 3 lt Version 1 0 gt 20 Data management in euroFOT euroFOT 01 08 2011 Test type Comments e Dust and fluid tests IP50 carried out successfully Achieved IP50 rating limited dust protection no protection from liquid for dust and fluid tests EMC tests EMC requirements were met e marking
89. ay time to collision and time to line crossing Based on the list of signals the following extra sensors were considered eye and head tracker GPS accelerometer and yaw rate sensor radar neadway position sensing lateral position sensing 2 2 DAS Solutions Used in euroFOT An overview of the data acquisition systems used in the different VMCs is shown in Table 3 Detailed descriptions of the chosen DAS solutions used in each VMC are provided in Sections 2 2 1 to 2 2 5 German 1 operation centre is presented first then French VMC and then Swedish VMC this sequence was chosen as there is a gradual increase in complexity of the data acquisition systems to be used in these VMCs German 2 operation centre is presented after the previous three this OC is the only OC in euroFOT that studied SafeHMI The Italian VMC is presented last as this VMC does not actually use sophisticated DAS as the other VMCs Deliverable D3 3 lt Version 1 0 gt 9 Data management in euroFOT Category Main component Number of CAN channels collected Operating System Storage device Cameras Radar GPS Deliverable D3 3 Data management in euroFOT euroFOT 01 08 2011 Table 3 The main features of the data acquisition systems used in the different VMCs German 1 OC CTAG Datalogger Il Embedded Linux SDHC up to 8 GB Existing vehicle radar via CAN Built in the data logger French VMC Light Heavy instrumentati
90. cameras Trucks only Volvo standard SmartEye embedded Embedded AntiSleep 2011 Ethernet and CAN Cable harnesses were manufactured by VCC One model for cars and one model for trucks The same cable harness is used for both the left hand controlled trucks Dutch fleet and for the right hand controlled trucks UK fleet Mounting brackets for cameras and the eye tracker were manufactured at VCC lt Version 1 0 gt 34 euroFOT 01 08 2011 w FY T p i s a T EEA TTAR we a DU LE ee net Figure 17 Swedish VMC Placement of DAS including extra sensor and cameras Deliverable D3 3 lt Version 1 0 gt 35 Data management in euroFOT euroFOT 01 08 2011 A number of functions special features that improves the performance of this DAS are worth mentioning here HDD diagnostics HDD status is constantly monitored by using a commercial tool called HDSentinel Hard Disk Sentinel HDD health and temperature monitoring 2011 This application uses a technique called SMART to get information from the HDD The current status of the HDD is written to the status report that is uploaded once for each trip Since consumer grade HDDs i e non automotive standard are used it was considered extra important to keep track of the HDD status to get an early warning if the HDD performance starts to decrease read write errors etc HDD shutdown The selected HDD has a function that will make it spin down after a fe
91. ch is a transparently maintained replicate of some tables of the complete database In the later case the most complex search options are disabled Deliverable D3 3 lt Version 1 0 gt 87 Data management in euroFOT euroFOT 01 08 2011 Be He Hs Trip Begin Date on Available 23281 Fitered 603 Selected 0 23100 LagHigh 28 Jul 2011 11 49 39 23100 LagHigh 29Jul2011 18 53 71 23103 LagHigh 01 Jul 2011 08 38 22 23103 LagHigh 01 Juk2011 17 54 18 23103 LagHigh 01 Jul 2011 18 56 35 23103 LagHigh MAMARE EEE 23103 LagHigh 06 Ju 2011 17 38 22 23103 LagHigh 07 Jul 2011 14 04 32 23103 LagHigh 07 Jul 2011 18 00 52 23103 LagHigh 09 Juk2011 09 06 22 23103 LagHigh OSJuL2011 11 02 48 23103 LagHigh 10 JuL2011 11 43 59 23103 LagHigh 11 Jub2011 12 41 56 23103 LagHigh 12 Jul 2011 11 00 29 23103 LagHigh 14JuL2011 15 12 24 23103 LagHigh 15JuL2011 10 53 04 23103 LagHigh 16JulL2011 09 42 76 Between 1 juil 2011 hi Errors Select Be 00 13 11 km 00 36 27 km 00 21 13 km 00 16 11 km And 30 ao t 2011 hi Logger ID 23115 Contig ID 36 km ClioHigh Conversion Date t T 5 1 janw 1970 ess SS os a Pi sF n 1 janv 1970 Custom Attribute Lon x CC Y 23103 23103 23103 aut LagHigh LagHigh LagHigh 16 Jub24011 10 55 15 16 Juk2011 12 38 23 16 Juk2011 13 35 20 4 5 7 23103 LagHigh 23103 LagHigh 18 Juk2011 06 33 29 on 16 Juk2011 17 50 56 43 23106 o
92. ckaging the main DAS unit with all of its components per vehicle based was found helpful during installation This is especially important when there are many loose things for example when DAS includes cameras and extra sensors 19 Related to lesson learned 11 the PC SAFER euroFOT DAS has to be opened to replace the CF card this was a tedious job Therefore it is recommended to find a PC if using with the CF card or any storage medium used accessible from the outside Deliverable D3 3 lt Wersion1 0 gt 114 Data management in euroFOT euroFOT 01 08 2011 5 2 Databases amp Data Storage To summarize the aspects related to database and storage we can see a number of lessons learned Since the VMCs have different implementations and solutions the lessons learned varies Handling large amount of CAN and video data has been a challenge Both copying and also converting video files is time consuming This was solved by an adaptation in the database and analysis tools at German 2 OC In the Swedish VMC a cluster environment has been used to increase the speed which in turn brought problems handling unique row id s Another issue has been the upload into an SQL database There have been difficulties with the design and performance problems There are also different storage solutions and from these we can learn about the benefits and disadvantages of choosing a SQL based or a file based data storage The following works well 20
93. compact flash and SD cards with industrial grade from the beginning 12 The use of office non automotive hard disks as storage medium in the DAS collecting video data was considered OK especially when the DAS is equipped with hard disk health check functionality as in the SAFER euroFOT DAS case 13 Inbuilt diagnostics in the DAS and web based monitoring tool are necessary to be able to follow up the status of the DAS systems to minimise risk of losing a lot of data due to undetected error s and for general quality check 14 Failing connectors have been the main source of DAS errors in the Swedish VMC In most cases these are due to human errors when assembling and or handling connectors It is recommended that all personnel involved in installation must be well trained to do the job and that proper quality check must be done after every installation 15 It is recommended to give extra attention to cameras during installation and daily monitoring Loose cameras were often found in the Swedish VMC Be ready with some unexpected work 16 If cable harnesses and mounting brackets are to be manufactured internally beware that one cable harness tool about 8 hours to manufacture 17 Many parts or components used in DAS might not be available in automotive grade It is recommended particularly if building in house solution to give plenty of time for making DAS components acceptable to OEM s standard Other reflections 18 Pa
94. configuration file from being read by any unauthorized person Deliverable D3 3 lt Version 1 0 gt 30 Data management in euroFOT euroFOT 01 08 2011 System requirements 5 The DAS must have functionality to send out CAN frames by routing frames from one bus to another 6 It must be possible to equip the DAS with and collect video data from at least four 4 video cameras Desirably the system should be able to handle six 6 cameras 7 The DAS must be capable of combining images from all cameras into a single image 640x480 or similar 8 The DAS must be capable of storing video data at a rate of 10 Hz or more 9 The DAS must compress all data from all sensors all CAN buses all cameras etc to a size below 1 GB per hour 10 It must be possible to equip the DAS with and collect data from two 2 LIN buses or two 2 digital I O s 11 It must be possible to equip the DAS with and collect data from one 1 J1708 bus 12 The DAS must have a GPRS or 3G modem 13 The modem must comply with European cell phone standards 14 The DAS must have a GPS receiver GPS data coordinates UTC time speed etc must be possible to be stored just as any other data GPS data must be stored at a rate of at least 1 Hz 15 The DAS must have an option to be equipped with and collect data from either a forward or a rearward looking RADAR delivered by the DAS supplier 17 The supplier must provide information about the average
95. d i e the delay if any is much smaller than a frame duration Lane tracking system The lane tracking system used is a Mobileye AWS2000 an aftermarket Lane Departure Warning and Forward Collision Warning see Figure 15 MOBILEYE AWS 2000 2008 It has a modified firmware which allows retrieving by CAN the values measured by computer analysis of the front scene Currently only the position on Lane is saved position of other cars can be retrieved as well but radar provides better measurement in at higher distances Deliverable D3 3 lt Version 1 0 gt 27 Data management in euroFOT euroFOT 01 08 2011 Figure 15 Mobileye AWS2000 Eyetracking system The eyetreacker used is an Antisleep EHT from Smarteye It is a single camera eye tracker camera computer and IR illumination placed in the centre of the dashboard Integration The integration of equipments is described in D2 2 Report of the results of adaptation in vehicle implementations and piloting and the installation guide which is an annex of D2 2 Saint Pierre Kessler amp Malta 2011 Figure 16 shows the visible elements of the instrumentation In the low level instrumentation only the radar can be seen inside the front bumper A I Feet Camera Mobileye Front Camera Cockpit Camera in roof light Face Camera Eyetracker Figure 16 French VMC Placement of extra sensors and cameras Deliverable D3 3 lt Version 1 0 gt 28 Data man
96. d by both French and Swedish VMCs The two VMCs worked together with Bosch to develop the requirements specification on these sensors and in identifying possible solutions Collaborations with other SPs Regarding interaction with other SPs SP3 has contributed to the task of feasibility ranking of SP2 hypotheses input from SP3 was made from the technical point of view This task was coordinated by Chalmers but each VMC gave its input for this task In addition to collaboration with SP2 SP3 also worked together with SP2 for in vehicle adaptation for instance it is important for both SP3 and SP2 to make sure that the DAS can fit in the specified compartment in the vehicle where the DAS was planned to be put The measure list from SP4 was first assessed in each VMC within SP3 then in a meeting involving SP3 SP4 and SP6 to clarify any open issues like unavailability of signal s for certain measure s and to discuss alternative solution s that might work for all VMCs In the latter meeting representatives of each VMC within SP3 were present Overall this collaboration was to make sure that SP3 as a whole can provide all the necessary measures that SP4 identified as necessary to answer hypotheses defined in SP2 within available budget Safety installation insurance and maintenance issues related to the installation of the DAS in the vehicles were some of the topics of interaction between SP3 SP5 and SP2 Concrete results from the interacti
97. d to oDBdata Calculate Kalman filter sensor fusion for position and heading enhancement Calculate all derived measures Calculate in tiers Add to oDBdata Add driver ID Separate EyeTracker data from Sort measures and events alphabetically simplify usage other data different frequency Calculate all events tier based Add to oDBdata Save oDBdata on server and transfer disk Display statistics Handle significant digits and too large infinite values Read MAP data information data enrichment and add to oDBdata Convert video streams to X 264 and store on server and transfer disk with checksums Remove some orignial measures from oDBdata as stated in the oPreProcCfg Check that all files When tranfer disk is almost full move to SAFER Chalmers for uploading remove data from original disk Figure 45 The pre processing steps in the Swedish VMC lt Version 1 0 gt Data management in euroFOT 72 The OEM specific information about per vehicle configuration is added in the MS Excel document called MEPS This document is then parsed by Matlab and a Matlab structure called oPreProcCfg containing all relevant configuration information for each indivudual vehicle is used in all steps of the pre processing euroFOT 01 08 2011 3 2 2 2 Upload and enrichment tools The Uploading step Ones the oDBdata files are available at t
98. data originate from a physical signal it has to obey the laws of physics and hence we can predict the behaviour or shape of the signal For example if the velocity oscillates between 10 and 100 km h it does not behave in a physically meaningful way Note however that a sudden deceleration Deliverable D3 3 lt Version 1 0 gt 58 Data management in euroFOT euroFOT 01 08 2011 from 100 to 0 km h indicates a hard braking event which should not be removed by some automatic attempt to remove low quality data e Incoherent relationships among data Some measures in a data set are related to each other For example acceleration and velocity are tightly connected to each other and when the value of the acceleration is positive the velocity should increase Otherwise there is an incoherent relationship between those signals In some cases the same information is acquired with independent sensors For example velocity may be acquired from the CAN bus as well as from GPS In such cases these two signals can be compared for quality assurance purposes With respect to speed Tests were only performed on a limited amount of pilot data sets and on dummy data sets generated with the aim of testing the frameworks functionality and stability Therefore no information related to performance speed can be reported here Future recommendations Current database schema support data records which are trips In the future we need to take into account the stor
99. ded green fulfil requirements yellow not enough information could be found and red did not fulfil requirements Provider Name CTAG Datalogger Il IMC imc busdaq2 IMC imc busdaqx IMC ime busLOG S IMC imc C1 IPETRONIK M LOG IPETRONIK S LOG Micro movement MAHTechS M5X PRO DAS yes GATE FMS 500 The best cost benefit price was found for the CTAG Datalogger II which was selected as the DAS of the German 1 OC after first discussions with CTAG This DAS integrates all needed hardware in a very compact design that can be easily integrated in the vehicles The data logger is presented in Figure 6 Deliverable D3 3 lt Version 1 0 gt 14 Data management in euroFOT euroFOT 01 08 2011 Figure 6 CTAG s Datalogger II CTAG s DataLogger Il is a storing device basically designed to receive and transmit CAN data by 4 channels and also to log the information onto an SD card Apart from this main feature the data logger can be located by a GPS receiver and is able to keep diagnostic communications with several ECU s at a time The data logger can be controlled with a server application Internet based solution for remote fleet management tracking logging data logger configuration on real time The main features of CTAG Datalogger I can be summarised as follows 4 CAN 2 0B Channels Logging GPRS Quad Band Modem GPS receiver CAN baud rate configurable by user Built in Real Time Clock with internal battery
100. e W X Y Z W X y Z Figure 62 Branch Usage of Version Management Release versions of the software are accessible through the version management system at a predefined location Deliverables for each software component For each release an individual branch of the Deliverables folder is created Software updates can be gathered through this folder All binary files have a defined version number which is created by a group of four numbers AA BB CC DD where AA is the corresponding major release BB is the subversion revision Deliverable D3 3 lt Version 1 0 gt 95 Data management in euroFOT euroFOT 01 08 2011 number CC is a continuously incremented number and DD defines year and day within the year of the build time This versioning ID shall guarantee traceability of origin creation time and corresponding sources of the binaries Help and support on usage extensions and corrections of the software is provided by the developers of the software packages throughout the entire FOT operation time Processes have been defined to ensure proper documentation and traceability To request additional features and to correct probable software errors a Bugzilla based bug tracking system is used Home Bugzilla bugzilla org 2011 All users of the software have access to this bug tracking portal and can report their issues and or recommendations Reporting includes the definition of the software component which is referred to as
101. e complete fleet with radars e Equipping all five high level instrumentation cars with video logging lane tracking and eyetracking devices Table 6 The part list used for the two levels of instrumentations used in the French VMC Item Low Level High Level CTAG Datalogger Il e Max 4 CAN Channels 2 channels used 4 channels used GPS n e GPRS data transfer TRW AC20 radar e not part of standard vehicle equipment VideoLogger e custom made for CEESAR H 264 Cameras A B amp W SuperHAD Exview Mobileye AWS e added with special firmware Smarteye Eye tracker e The components of those two levels of implementation are described in the rest of this section Deliverable D3 3 lt Version 1 0 gt 24 Data management in euroFOT euroFOT 01 08 2011 Low level instrumentation Figure 11 Low level instrumentation at French VMC In addition to the CTAG Datalogger Il the low instrumentation level consists of an ACC radar This is a standard AC20 radar from TRW Figure 12 including a gyrometer allowing function on a vehicle without ESP and with a software which instead of controlling ACC and or FCW just provides the characteristics distance speed azimuth etc of the vehicles in front of the car Figure 12 AC 20 radar Integrating the radar in the DAS required the development of a callback mechanism to forward some information mainly vehicle speed from the vehicle to the radar High leve
102. e data should be created When developing and refining scripts it is much more time efficient to test on a smaller dataset When satisfied the analysis can be executed on the whole dataset e To be able to use SQL for analysis on billions of rows in a single table effectively indexes needs to be created It is also vital to consider the storage needed for all indexes e Use and analyse explain plans before executing queries By doing this many bad written SQL can be avoided e Try to fully use the power of state of the art servers utilizing many cores By using parallel techniques in the database server or splitting the queries into smaller pieces can boost the performance e Consider the cost for making backups of the database If raw or pre processed data Is stored and has backup there might not be need for taking backups of the continuous Deliverable D3 3 lt Version 1 0 gt 78 Data management in euroFOT euroFOT 01 08 2011 data tables The time for re uploading data might take longer time but it is important to state the service level agreement for downtime e There is a great cost in terms of computing time for performing joins when dealing with large tables By extensive de normalization of the schema and even duplicating data analysis can be faster e Storing large amounts of data in a traditional relational database should be questioned for future projects of this scale or bigger The trend is that more specialized software
103. e stamping The DAS time stamps data by a software mechanism This means that the performance time stamping performance is affected by the current workload of the CPU In order to get a better understanding of the time stamping performance a software tool was developed that outputs text files containing just the time stamps for further analysis in e g Excel The time jitter for CAN and video and the average latency between the signal sources were measured CAN quality The DAS collects and stores all data from four CAN buses Data is compressed before storage to file A test environment was created to verify that the logger does not miss any CAN data A function was implemented in the DAS software to detect and report CAN bus hardware buffer overflow Eye tracker A software application was developed to extract eye tracker signals from tests collected data and to calculate derived values The eyetracker software has had to be updated several times to address issues found during the tests Accelerometer The accelerometer is connected to one of the four CAN buses and is tests powered from the PC Web based monitoring tool To accommodate DAS quality check during FOT a web based monitoring tool was also built This tool has functions for showing the fleet status in four different views 1 Main page where all vehicles are shown Figure 18 2 Vehicle page where all trips for a specific vehicle is shown 3 Trip page where all diagnostic paramet
104. e tool should be easy to use with the interface facilitating the annotation process 2 It should be flexible and not limited to specific research questions allowing users to define annotation categories 3 It should be expandable so that the users could implement new modules and functions 4 It should provide users with an interactive timeline based visualisation of observational data 5 It should be able to efficiently retrieve information from both file management systems and relational databases The development of this software is completed in spring 2011 4 1 Method Procedure Used The development of the data analysis tool involved defining the use cases and requirements development of the technical design implementation and testing of the data analysis tool Common templates were used as a base for gathering requirements and use cases by all VMCs The templates described in Deliverable D3 2 Annex 3 Kovaceva et al 2011 gave an efficient structure for unifying use cases collected from each VMC which were then used to derive requirements for the VMCs partners The requirements were categorised into high priority i e common for all VMCs partners and low priority i e specific for different Deliverable D3 3 lt Version 1 0 gt 85 Data management in euroFOT euroFOT 01 08 2011 partners For example each VMC had specific low priority requirements due to differences in sets of safety functions sets of hypothese
105. easons the map matching tool and map data enrichment tool were available quite late in the project and thus were not integrated in the pre processing at the beginning Therefore we have had to add a way to add map data on already uploaded data Not possible to know all possible analysis derived measures and events needed a priori The piloting phase of the pre processing and uploading included the full chain until test analysis and methodologically answering a few research questions but since the amount of data needed for proper statistics and full analysis was obviously not available it was difficult to plan for and implement all possible derived measures events and quality checks before analysis As a result of this the analysts would have to review each hypothesis and revisit the implementations of derived measures and events and probably implement several new or refined ones along the way during analysis Deliverable D3 3 lt Version 1 0 gt 74 Data management in euroFOT euroFOT 01 08 2011 Successes A user friendly interface using spreadsheets for defining pre processing significantly simplified usage and implementation The pre processing and uploading was implemented so that adding new measures derived measures and events etc in the development phase was easy That is a framework was implemented to allow for only needing to add information about the measures in a XLS spreadsheet and when needed implement specific filt
106. ed in euroFOT Sections 2 2 e shares the lessons learned Section 2 2 Here DAS is defined as the in vehicle hardware and software that constitutes a complete logging system including external sensors used for the purpose of logging necessary signals for FOT In the deployment of field operational tests data logging is a key issue and euroFOT is no exception Success of the project relies somehow on the DAS performance it should log all the information planned to be logged with no data loss or corruption be proven to work in different conditions and be reliable for use in a long period at least one year by a large number of vehicles 2 1 Method Procedure Used The steps that were followed throughout WP3300 work to provide best suited data acquisition systems for euroFOT can be summarised as follows 1 Gather initial requirements and or constraints from the different VMCs or specifically OEMs and identify necessary signals and sensors 2 Gather technical requirements specification considering safety ethical and legal issues as well as cost 3 Identify possible DAS solutions conduct market survey and evaluation 4 Decide which DAS solution s to use and make the DAS ready for euroFOT Step 1 is summarised in Section 2 1 1 The highlights from steps 2 4 are summarised in Section 2 2 2 1 1 From functions and hypotheses to signals and sensors The initial requirements and or constraints that had to be taken into account
107. ed to work with dedicated servers and MySQL MySQL The world s most popular open source database 2011 for the database whereas the German 1 OC uses virtual machines servers and an Oracle database system Oracle Hardware and Software Engineered to Work Together 2011 Deliverable D3 3 Data management in euroFOT lt Version 1 0 gt 51 Eurofot Database Schema Version 2 2 21 3 2011 euroFOT 01 08 2011 Vid INT columnname VARCHAR 45 name VARCHAR 45 unit VARCHAR 45 display TINYINT description VARCHAR 512 type VARCHAR 45 Vid INT Vid INT columnname_1 ENUM columnname_2 ENUM Vid INT columnname V ARCHAR 45 columnname_1 ENUM columnname_2 ENUM abbreviation VARCHAR 45 description VARCHAR 512 reference VARCHAR 128 dassification INT version VARCHAR 11 columnname V ARCH AR 45 name VARCHAR 45 unit VARCHAR 45 display TINYINT description VARCHAR 512 type VARCHAR 45 tripID INT annotation VARCHAR 512 Indexes Vid INT driverID INT vehidelD INT Ostattime DATE endtime DATE duration DOUBLE length DOUBLE Qisvalid TINYINT periodtype INT experimentalphase TINY configID INT name VARCHAR 45 childrenEventtypeid INT columnname V ARCH AR 45 description VARCHAR 512 Vid INT eventtypelD
108. eindex which is a uniformly increasing time index in 100 millisecond steps from O except for EyeTracker where timeindex is the sampled time in uneven milliseconds This is what is used for all in trip analysis Then one field per measure All fields have the same length oMeasurelnfo Include all metadata information about the individual measures A structure with fields of all names that are in voMeasureData There is always a Measurelnfo for a MeasureData field For each measure information i e number of significant digits measure description unit enumeration description data source creator and creation time for traceability as well as per measure quality are stored The per measure quality has three levels 1 the overall MeasureValid 0 or 1 2 qualityDetails specific information that may differ between measures and Deliverable D3 3 lt Version 1 0 gt 70 Data management in euroFOT euroFOT 01 08 2011 3 PreProcQuality a general procedure applied to all measures in the same way but with thresholds Also if there is a per sample vector available for this measures a reference to this vectors name is found here oEventData A struct with one field per event name e g e RELongAcc06 Each field is a Struct array corresponding to each of the events found of this type for this trip The data included is Trip ID StartTime StopTime EventLevel value that triggered the event specific definition for each event Text
109. election and integration It should be considered that the system will be running during 12 months in vehicles owned by non professional drivers These drivers are real customers from euroFOT partners OEMs so no further maintenance calls are desirable e GPRS connectivity should be taken into account As the vehicles can be driven around different European countries and data servers are located in Germany network compatibility should be considered from the very beginning In this sense specific negotiations with service providers GPRS connection are to be arranged e g flat rate etc Roaming connection fade outs and other data missing scenarios should be considered during the development and pilot testing The different configuration access parameters within the different European GPRS network providers can cause some configuration problems Deliverable D3 3 lt Version 1 0 gt 21 Data management in euroFOT euroFOT HeetManager Interface euroFOT 01 08 2011 1917 19483 2 0 GPS Sste es 9 PDoP 254 YDoP 224 GPS Longitude 121001791655557 GPSlettude 4933150475 GPS Bexing 0 GPS Akiude 0 GPS Speed 0001852 Available Detaloggers a o OEM IP Reached within Y Sminutes Fv Last Connection O0 eff Last Synchronization State 1 23114 MAN 80 226 173 48 42226 40 05 April 2011 05 55 21 05 April 2011 0614 15 idle d 2 o 23119 Ford 80 226 173 18953864 4 2 20April 2011 18
110. en possible configuration of FOTware Deliverable D3 3 Data management in euroFOT lt Version 1 0 gt 99 euroFOT 01 08 2011 Figure 68 Annotation workstation with FOTware 1 Data player The data player is the core module in FOTware and controls the playback of the data Modules can be created manually from a list in the player or automatically via scripts A slider bar displays the current position in time and can be dragged to jump to other positions The sliding is instantaneous and all plots and video windows are continuously updated as the slider is moved By using the mouse scroll wheel the user can change the playback speed if the data player is in play state If the player is in paused state the scrolls will shift the current position in time The data player displays information of the current trip date segment and time Closing the data player closes all other currently open windows and stores the setup When FOTware is started again the user can choose to load the configuration with the same windows layout and settings as the previous one Visualization CORE OI x Add new module Signal plot module Set rate Current trip 2009 04 27 09 19 35 Set time Current segment 1 Current time 726 67 Play Playback speed 0 00 Figure 69 Data player 2 Video viewer The video viewer displays video sequences by using ffdshow a media decoder and encoder ffdshow tryouts 2011 and videolO a lib
111. ent For example the Compact Flash problem would most likely have been detected in an earlier stage Deliverable D3 3 lt Version 1 0 gt 113 Data management in euroFOT euroFOT 01 08 2011 if the system had been left continuously running for one or two weeks in a test rig The pilot car was run for several months without failure Review on hardware software technical methods and tools used 8 Data must always be saved in the DAS either in the internal or external storage like hard disk or SD cards because wireless connection is not always reliable GPRS connection issue was experienced quite often in German 1 OC 9 If the vehicles are driven across many different countries and GPRS is used for transferring data to the home country it is important to understand the services from the different network providers As this could affect the set up of the transfer mode from the DAS to the server in this case the server at the VMC centre 10 Proper test and care should be taken when IR illumination is used as part of the DAS for example as part of eye tracker Although most drivers did not get any problem several drivers experienced dry eye but got used to this after a while 11 Compact flash cards and SD cards are among the common DAS problems experienced in euroFOT The consumer grade compact flash and SD cards have had to be replaced with those of industrial grade not long after FOT started Therefore it is recommended to use
112. ent m Instance Edit Generator Script File E A CCSLSpeedChange fi Situations Tables 10 Situations Ho Situations Calculations H Speed_Headway_PrctCCUse PretSLUse H SaveSituation 2 H Speed_Headway_PrctCCSLUse_Duration _ E Speed_Duration BEG Trip Attributes E i Processes Edit Refresher Script File l 4 i CleanUpRadarRange m Delete this item Figure 59 Experiment builder Deliverable D3 3 lt Version 1 0 gt 93 Data management in euroFOT euroFOT 01 08 2011 no AO i 7 I 1 i Figure 60 Dependency graph 4 2 1 2 Tests performed and quality assurance Release Process and Version Management Releases are defined beforehand Release and features are defined in the development roadmap The releases are signed with a version number The version number is a group of three numbers AA BB CC where AA is the major release BB is the Minor release and CC is a revision ID The roadmap for the releases foresees the usage of the numbers following to Pre Release FOT preparation FOT startup software evaluation FOT operation corrections and feature extension Data Evaluation respecting results gathered from SP6 Final release version PEDE Q To ensure software quality all software components and features are tested by the developers before releasing To keep track of software changes a version management system is used The version management system is
113. er extensive market survey on the DAS components was conducted in early 2009 Part of this market survey was done via internet survey For this activity previous experiences from earlier FOT projects were very advantageous Request for tender RFT documents for full DAS Vehicle PC eye tracker and vehicle communication interface were prepared and sent out in July 2009 No answer was received for the RFT for full DAS Thus there is only one solution that is to build a new custom made DAS based on PC technology and commercial off the shelf components Relation between the chosen solutions to previous SAFER owned data acquisition systems The euroFOT partners in the Swedish VMC are partners of SAFER a Vehicle and Traffic Safety Centre at Chalmers The SAFER partners including the euroFOT partners in the Swedish VMC have previously involved in other FOT studies in Sweden In those studies the SAFER partners have used two different data acquisition systems The designs of these two systems can be used by the Swedish VMC partners for this study However due to more complicated and restricted requirements for euroFOT both systems would need further development to fulfil the euroFOT Swedish VMC requirements specification For example EMC requirements are not yet fulfilled by the previous data acquisition systems EMC requirements have never been a strict requirement that had to be fulfilled by the DAS in the previous FOT projects This was beca
114. er on different parts of the data management chain The interconnection and relation between the work packages inside SP3 and with other subprojects were coordinated at euroFOT level by SP3 leader and WP3x00 leaders As a whole SP3 provides common ground as well as specific solution s for each VMC 1 3 Organisation of D3 3 This deliverable is organised into six chapters e Chapter 1 introduces the purpose of this deliverable an overview of SP3 work the chosen way of work as well as the term vehicle management centre VMC which will be used throughout this deliverable e Chapter 2 describes the data acquisition systems used in the different VMCs e Chapter 3 describes how we process and store the FOT data e Chapter 4 describes the tools used in the different VMCs to extract view and annotate the collected data e Chapter 5 summarises our experiences and lessons learned in terms of data management in the euroFOT project and makes recommendations for future FOTs e Chapter 6 sums up the contributions to the state of the art made by SP3 partners This chapter also concludes this deliverable Deliverable D3 3 lt Version 1 0 gt 6 Data management in euroFOT euroFOT 01 08 2011 2 Data Acquisition System DAS This chapter e summarises the process that was used by euroFOT partners to select and finally made available the data acquisition systems in the different VMCs Section 2 1 e describes the data acquisition systems us
115. er or the SQL module and an event must have been selected with the event browser Differently from any of the other modules which have one and only one window for each instance the number of windows in this module may vary In fact the number of windows depends on the number of annotations to be done This module is dynamically generated from an excel file Depending on the content in the excel file the module can be used for different types of annotations and can have different appearances The excel sheet must be generated by following specific rules in order to be interpreted by this annotation module More specifically there are 6 different types of annotations which can be described in the excel sheet e g single choice multiple choice etc Furthermore annotations can be organised into different sheets inside the same excel document Each sheet generates a new window A browser window is also generated to enable the user to visualise hide or select the current annotation window by pressing the Deliverable D3 3 lt Version 1 0 gt 105 Data management in euroFOT euroFOT 01 08 2011 buttons shown on the left window in Figure 79 Each annotation will be presented to the user as buttons list box or pop down menu depending on the nature of the annotation and on the number of possible choices In the browser window a table with the current status of each annotation is presented By clicking on this table the user can plot the
116. erred to an external server via UMTS 8x FlexRay 2x Ethernet 3x High Speed CAN 1x Low Speed CAN 1x MOST Figure 20 ZGW connections This figure is from D2 2 Figure 21 Placement of DAS ina BMW vehicle This figure is from D5 2 Deliverable D3 3 lt Version 1 0 gt 42 Data management in euroFOT euroFOT 01 08 2011 Daimler The DAS logs data from several CAN buses and cameras The architecture of the data acquisition system is shown in Figure 22 and the placement of the DAS in a Daimler vehicle is shown in Figure 23 Hardware Power Supply 12 V KI 30 Storage 5V DC DC Additional 12 V CAN activity KI 30g GPS receiver USB P R O VE te C h VA Video2PROVEtech VA Video 4 1 Standard 4 1 Video Configuration CAN Integration of Navigation data camera2Quad Daimler extern BMW Definition Trigger event Data logger Linux CAN2PROVEtech VA Trigger2PROVEtech VA CAN Data DBC amp Signals Daimler Body CAN Chassis CAN camera 1 2 3 4 Headunit CAN Trigger button amp micro Dynamics CAN Definition p e trigger audio comment Figure 22 The architecture of DAS used by Daimler This figure is from D5 2 The DAS used by Daimler also includes a trigger box and a microphone see Figure 23 that the driver can use to make auditory comments regarding incidents or other important issues e g concerning the navigation system This is a unique feature of the DAS co
117. ers for a specific trip is shown 4 Image snapshot page where snapshots from all cameras for a specific trip are shown From the main page red colour is used to highlight which DAS or vehicle needs attention or has some issues Information about the CF card and HDD space left and health status remaining money on SIM card and last upload from vehicle are also presented in this page Information about trips and their details like various measurements to check the different sensors are working fine and trip duration are shown in vehicle page Details of detected errors during a trip version of logger configuration file etc are shown in the trip page Deliverable D3 3 lt Version 1 0 gt 38 Data management in euroFOT euroFOT 01 08 2011 Staten Web 2 me Dace Logoul Chance Password came New Veet Delete Vee A TheMetherlends L Hung space 4 n 06 Comment f a thy Erroe Count Oi FOI22 CU VTEC The Nemerands 4728 Reset 1191C gt 485 68 Go 100 100 2011 08 28 16 2605 Yebiche Loos EQI22 S VTEC The Netertancs 859 Reset 13 30 G 334716 100 100 2011 06 28 17 1 FOt218 CN VTEC The Nemertands 4931 129360 41742 Gb 100 100 2011 06 28 17 45 17 Veh cie Loos FOTZ20 OF VTEC The Netrertancs 049 Reset 12 14 tr 4 01213 VTEC The Nefertands 4101 1101 0b 344866 100 to 11 06 28 16 4 de Lans FOOT 14 7 TEC The Mefertands 473 Reset 10 FOT2Z25 UE VTEC The Nemertancs 5067 12910 34057 Gb 100 100 2011 06 28 16545
118. ers or handling criteria in MATLAB with dynamic references in the spreadsheet This document also made sure that all documentation was gathered in one place only Several layers of quality checks were implemented The solution we used includes a number of layers of data quality checks that is used and added to the database making it possible for analysts to readily access and filter data with respect to quality on a per sample on some measures per measure and per trip basis The quality checks were also implemented so that qualities was calculated in several tiers allowing for more complex but still easily traceable quality checks One data format was used throughout the pre processing uploading and analysis based on MATLAB structures once the raw data had been read This simplified implementation aspects and application of functions and filters throughout the project This format is complete enough to easily be able to transform the data into most other formats if a parser writer is written Hypothesis breakdown In order to be able to implement the piloting and prepare for main analysis significant work has been done by other work package to break down the VMC specific hypothesis into specific Hypothesis Breakdown document one per hypothesis not all finalized yet This has been a great help in the implementation of the pre processing and piloting and assuring a solid link between analysis and data management Performance W
119. esses we only have process we have BB processes but manually started Future recommendations e Overall the cluster environment using MATLAB parallel toolbox Parallel Computing Toolbox MATLAB 2011 has worked very well The drawback has been that management of unique identities has become more complex e Separating the handling of incoming data from pre processing has been a success Especially since all steps in the pre processing were not finished before data collection started e Evaluate at an early stage all the software needed At the Swedish VMC some software packages decoding CAN NAVTEQ map data needed Windows as operating system causing a more complex automated flow on the Linux based high performance computing HPC e By converting video data from xvid Home of the Xvid Codec 2011 to h264 H 264 MPEG 4 AVC Wikipedia the free encyclopedia 2011 the data storage need was reduced by 40 at Swedish VMC The process is very CPU intensive and a cluster of computer nodes were needed to perform this task e The large amounts of video data exceeded the filesystem limits of storage appliances Therefore a server providing logic to be able to create a single mount point was needed For future projects different filesystems and appliances should be evaluated for instance by using ZFS ZFS Wikipedia the free encyclopedia 2011 or equivalent for almost unlimited shares e A sandbox database including a subset of th
120. files for each trip the handling of the videos e g copying was no problem with the pilot data about 50 hours 3000km of data logging After converting the first months of FOT data it was realized that because of the huge number of single files copying the data from Daimler conversion in ext3 ext3 Wikipedia the free encyclopedia 2011 to IZVW server in NTFS NTFS com File System Structure Recovery Software Hard Disk Internals 2011 file format took longer than measurement time AS a consequence the format had to be changed from single frames to video format and the database and data analysis tools had to be adapted Especially the adaptation of the video tool took a lot of effort Now the tool works with both single frames and video format Deliverable D3 3 lt Version 1 0 gt 83 Data management in euroFOT euroFOT 01 08 2011 3 2 4 Italian VMC 3 2 4 1 Data storage and upload solution Since the Italian VMC only collects questionnaires data see also Section 2 2 5 the storage solution and upload procedure used are much simpler compared to those used in other VMCs Drivers background data timing of handing out and receiving back of filled in questionnaires incentive scheduling and timing and raw file of filled in questionnaires are stored in MySQL PHP database The questionnaire data is collected in two ways 1 electronically based by using an open source tool LimeSurvey LimeSurvey the free amp Open source survey
121. g Jerk per data source Add to measures for this data source in oDataSet instantiate oDBdata for the first time For measures not in oPreProcCfg get nearest neighbour interpolation per source corresponding to resample Add to oDBdata primarily for debugging Add metadata about filter and resampled data to oDBdata Perform processing of some OEM sensitive measures to less sensitive derived measures pre pre processing on CAN Create a common time vector at 10Hz based on the CAN velocity timestamps Add metadata information to the oDBdata Add common time information from resample to oDBdata EyeTracker specific pre processing including data extraction filtering derived measure and quality calculations but not resampling stored in source frequency Save the file in the intermediate format oDataSet mat temporarily Remove the first Xs and the last Ys of data different at VTEC and VCC due to startup shutdown effects on data sources Run a first data verification script on oDBdata Add errors and or warnings information to oDBdata Extract baseline and treatment information for this trip base on CAN Calculate per sample quality on oDBdata for a few measures data sources Add to oDBdata Calculate per measure quality on oDBdata for measures using per sample quality Add to oDBdata Calculate per trip quality Based on per measure quality Ad
122. gement of the data transmission is also important when to send the collected information via GPRS should be defined when the vehicle is running and collecting data At the end of a trip Data Size As the data to collect in the framework of euroFOT is limited there is no use in recording and storing all data circulating in the CAN network of the vehicles An appropriate data filtering through SP2 and SP4 information is compulsory This is a key factor to reduce storage and transmission requirements SD storage As a back up in case of transmission failure and in case GPRS is not able to cope with the necessary bandwidth for the real time data transmission an SD Card storage is necessary As there is the risk to write and overwrite it continuously a reliable and high grade card is requested Flexibility EuroFOT comes up with different requirements in terms of data acquisition systems specifically for the OEMs DAS should withstand their different reliability tests and specifications and in terms of collected data It is important then to integrate open protocols so that the DAS can be customized per OEM and per application This is done through python scripting Finally due to all the above mentioned characteristics it has been revealed that a success factor for the project was the inclusion of the DAS provider among the project partners so that the system can be updated and upgraded to euroFOT final requirements This includes future
123. ging large amounts of data for future projects of this scale or bigger 38 In euroFOT filesystem was also used in addition to the relational databases For example the Swedish VMC and German 2 OC stored the FOT data as Mat files This is to allow analysts to work through the database and or directly with MATLAB The effectiveness of using file management system FMS for complex analysis of large FOT data is yet to be tested Among the drawbacks of using FMS compared to the database management system DBMS are 1 Data access on FMS is not as flexible and granular as in DBMS 2 consistency check on data is not at all enforced in filesystem 39 Due to budget constraints the use of special features like compression and partitioning of relational databases for example Oracle and proper data warehousing has not been thoroughly explored in euroFOT It would be interesting to see how these features could enhance the solutions we adopt 5 3 Data Analysis Tools The following works well 40 A crucial step in the development phase was testing the analysis tools by as many users as possible The feedback from the users was valuable input for finding the Deliverable D3 3 lt Version 1 0 gt 116 Data management in euroFOT 41 euroFOT 01 08 2011 errors fixing the bugs and making the improvements according to the user s suggestions SQL software tools may require a fairly high level of knowledge to use so it was advantageous to develop
124. gure 25 when choosing database engine fileserver etc specific VMC needs and constraints have to be taken into account and thus suitable solution s were chosen at VMC level as these decisions to certain degree also depend to the IT infrastructure of the different partner organisations Deliverable D3 3 lt Version 1 0 gt 46 Data management in euroFOT euroFOT 01 08 2011 Tasks 4 6 of Figure 25 are summarised in the next section Guidelines for upload strategies and procedures as well as test strategies and procedures are discussed and made together However considering the collected data types data format as well as resources the implementation was done separately Within SP3 it was decided that e IKA CEESAR and ICCS work together to implement the data management framework to be used in German 1 OC and French VMC Note German 1 OC and French VMC use the same CTAG datalogger Il e Chalmers Volvo and VCC work together to implement the solution for Swedish VMC Outside SP3 the partners in German 2 OC work together to implement the solution for German 2 OC and the same applies in the Italian VMC Deliverable D3 3 lt Version 1 0 gt 47 Data management in euroFOT euroFOT 01 08 2011 3 2 Database and Storage Solutions Used in euroFOT The table below is summarizing the different implementations in the VMCs Table 11 The database and data storage solutions used in the different VMCs Category German 1 OC French VMC Ger
125. h The Compact Flash CF cards recommended by the PC supplier Transcend 32 GB consumer grade MLC type worked well during the pilot but almost immediately broke down when a larger number of vehicles started to run All cards were replaced with Transcend 16 GB industrial grade SLC type cards The PC has to be opened to replace the CF card and the recommendation is that one should try to find a PC with the CF card or whatever storage medium that is used accessible from the outside The PC supplier Rugged Mobile Systems UK has been of great help They provided a lot of services so that what was delivered to the Swedish VMC was a complete and tested PC with an installed video grabber In practice the Compact Flash problems lead to a significant amount of extra work for the Swedish VMC opening up all PCs for CF replacement but in general it is of course recommended to find a supplier that can do as much work as possible It is easy to underestimate the work needed for manufacturing cable harnesses and mounting brackets One complete cable harness took 8 hours to manufacture Deliverable D3 3 lt Version 1 0 gt 39 Data management in euroFOT euroFOT 01 08 2011 Since the task to provide suitable DAS for the Swedish VMC turned partly into a development project there was no equipment DAS available on the market that fulfilled the requirements it would have been convenient to build a complete test rig for emulating the environment I
126. he OEM they are moved to Chalmers The process of how the upload is performed is described in the flowchart below This include the steps ofrom when the software is arriving at Chalmers from the OEM until it is uploaded into the database and the files are moved to appropriate folders Uploading in the Swedish VMC in euroFOT This flowchart shows the Swedish VMC s upload software process The transfer disk An instance of matlab Upload OEM rae _ is attached to a is identifying a new specific pn pal es poy 64 bit Windows 7 drive and initiates configuration file is m ee oe workstation processing read I Physical handling and configuration oDBdata mat files All video files for Copy eye tracker are copied with this trip is copies ripae data with checksum checks with checksums iini checksums 1 a le Le oo CL her dd oa Upload Event information to Upload trip DB SQL inserts information to 1 Check if events are database SQL uploaded inserts 2 Add event metadata 3 Add event data Per trip process Trip and event handling Create text files to use with SQL loader 1 Create trip_ID vector aes 2 Rearrange column order PAR e T lonp tale es 3 Create SQLloader control file ee 4 Prepare a full matrix with all data eee i 5 Create Matlab output format string a 6 Handle too large values and infinite T Open output and write text file Per trip process Handle SQL loader Figure 4
127. he device The status report should include but not be limited to DAS status Remaining hard disc drive space or low disc space warning configurable Detected errors Desirably quality measures for some signals in order to assess the quality of sensors e g eye tracker Deliverable D3 3 lt Version 1 0 gt 31 Data management in euroFOT euroFOT 01 08 2011 System requirements Desirable a measure of the hard disk drive health 27 The DAS must have functionality to minimize or turn off data transfers over GPRS 3G when roaming 28 The DAS must have a storage capacity for collected data of at least 80 GB 29 It must be possible to extend the storage space either by changing storage media e g using a larger hard disc or by adding an external storage media on USB It is understandable that the storage media larger than 80 GB could not operate in the full operating temperature range 30 The process of exchanging storage media must be easy and fast lt 5 minutes This means that there must be a hard disc drive bay or that the hard disc drive is external to the main DAS unit 31 The DAS must be able to encrypt data on the storage media 32 The data format of the collected data must be available to euroFOT partners in Sweden for development of tools for reading the data files 33 There must be a mechanism e g watchdog that guarantees that the DAS always is shut down when the ignition is turned off Hardware
128. he limit at the time of writing this report not all of the total collected data at each VMC has been uploaded to the respective VMC s database and storage However with so many other technologies available for example column oriented databases Hadoop database HBase and the data warehouse for Hadoop Hive it is wise to be open to the potential benefits of those technologies or similar for the management and analysis of FOT data of the scale of this project or bigger Regarding the data analysis tools the modular approach adopted here has allowed reuse of the same modules in different use cases and or different machines It is felt that the modular approach could still be extended the interfaces of the tool could be made more standardised additional features like a command line interface could be added and external tools beyond MATLAB could be integrated In summary the best suited data acquisition systems the required infrastructure and services for data storage the base software for analysis general quality assurance as well as support in installation and maintenance strategies to both SP2 and SP5 have been provided to the project SP3 has achieved all of its objectives Deliverable D3 3 lt Version 1 0 gt 119 Data management in euroFOT euroFOT 01 08 2011 Acknowledgements We acknowledge and thank the following for their contribution to the work described in this deliverable Helena Gellerman Chalmers Riku Kotiranta Chal
129. ices were modified in such a way that the use and the handling of the system could be logged on a SD card in the mobile devices The main requirement for the DAS is that the system must be able to collect videos of driver behaviour in particular videos of driver s face movements of head and upper body of the driver movements of arms of the drivers and the view to the navigation systems forward view and rearward view 2 2 4 2 Chosen solution The two OEMs in this OC have chosen to use their own in house solutions from the beginning of the project BMW developed its own in house DAS solution Daimler contracted MBTech to develop their in house DAS solution MBTech was chosen because of their knowledge of Daimler vehicles and that they have the appropriate installation facilities and experiences Deliverable D3 3 lt Version 1 0 gt 40 Data management in euroFOT euroFOT 01 08 2011 2 2 4 3 Final implementation to fulfil euroFOT requirements BMW The components that make up the BMW DAS are listed in Table 10 and the placement of DAS is shown in Figure 21 Table 10 Part list for the DAS used by BMW Item ReadyNAS 2100 Power management ZGW Zentraler Gateway Switch Video server ACD2000Q UMTS Modem Viprinet 300 including UMTS antenna 2dBi 8 ports Hub Short range radar sensors 2 ahe CMC2002HP 1 3 7 PAL ahe CM 108 7 CCD Mini 3 KISSte Configurable Intelligent System Control DC DC Converter
130. ies 2 2 3 4 Tests and quality check performed Table 9 The different tests performed on the SAFER euroFOT DAS Test type Comments PC test Of special interest was to test the power management part of the PC since this is a known area of problems in automotive applications The PC has three power inputs KL15 KL30 and Gnd It also has settings for delaying Start up and shut down times with respect to KL15 changing state The following deviations were found e Higher current consumption in off mode than specified 17 mA instead of 10 mA This is only a problem for car installations and was solved by adding an external relay e Unstable operation after resuming from hibernation and sleep Since these modes were not planned to be used this was considered a minor issue Climate A complete system was tested in a climate chamber at different chamber tests temperatures ranges from 20 to 60 degrees Celsius The tests were successful The chipset in the PC has a function to shut down if the internal temperature reaches a certain limit This function has not been tested since it was considered to be a potentially destructive test and that the available number of PCs was low at the time EMC tests A complete system except for the eye tracker was tested at an authorized EMC testing facility SP Technical Research Institute of Sweden to verify that it complied with the requirements for e marking The tests were successful The eye tracker was
131. is displayed to the user The user can also write directly or copy and paste a folder location to have access to it Deliverable D3 3 lt Version 1 0 gt 102 Data management in euroFOT euroFOT 01 08 2011 B File Chooser v 1 2 Pex Do Weersidozza NET WDesktop MARCO Projects4 EuroFoTivMCspaceSPaysul tools Scripts estDatai Ho data file in currenth open Read also eve tracking data TestData_ VCC GUI mat TestData VOC GUL ET mat testdata volvo mat Figure 74 The file browser module 5 Signal browser The signal browser displays the name and source of all the signals that are currently loaded and allows the user to select one or more signals that can be plotted in a signal plotter module or visualised in a signal viewer module Figure 75 shows a representation of this module when some data has been loaded The elements of the signal browser are Signal table It lists down the signal names and sources and allows the individual and multiple selections of signals Signal selection buttons These can select or unselect all signals present in the signal table Target module menu It lets the user select the target signal plot or signal visualisation module in which the selected signal should be plotted or listed down Plot type menu With this menu the user can specify the desired plot command among linear plot plot histogram hist and cumulative distribution cdf Signal preview checkbox When selected the
132. is used for corner case purposes New techniques as columnar databases Hadoop Welcome to Apache Hadoop 2011 including HBase HBase HBase Home 2011 and Hive Welcome to Hive 2011 are making progress and should be evaluated Also new products for hardware and software acceleration could be considered for better performance It is also an upcoming challenge to see how different storage and analysis software can be used to complement each other By mastering this researchers can be much more efficient in analysis projects and by using the right tools at the right time new methods for analysis projects can be developed e For pre processing and analysis these large amounts of data requires 64 bit memory structures and software support e Data access shall be role based e Data access on filesystem is not as flexible and granular as in databases e Network capability is important to consider when handling large amounts of data At the Swedish VMC copying data in parallel from three USB disks to storage exceeded 100mbit connection and 1gbit was needed e All data sent between OEM and the Swedish VMC was encrypted using Truecrypt TrueCrypt 2011 ensuring the sensitive data 3 2 3 German 2 OC The data logging of Daimler and BWM contains CAN data which are converted to even spaced time series data with 10Hz and video data with 5Hz The data are logged in the vehicles in data formats that are specific for the OEM The OEMs are responsible f
133. isition systems and sensor technologies and their applicability for FOTs 47 lessons learned Working methods that proved positive include using common templates for gathering requirements having specialised focused groups to enable joint efforts in performing specific tasks separating the handling of incoming disks data from pre processing stage using modular approach for developing the base data analysis tool Common DAS hardware issues include broken compact flash and SD cards failing connectors and loose cameras Recommendation made include use industrial grade compact flash cards and SD cards give extra attention when installing connectors and cameras and finally be open to the potential benefits of using database features like partitioning and compression to enhance the standard relational database using proper data warehousing or even to other technologies e g column oriented databases Hadoop database HBase or data warehouse for Hadoop Hive for the management and analysis of FOT data of the scale of this project or larger projects Deliverable D3 3 lt Version 1 0 gt 1 Data management in euroFOT euroFOT 01 08 2011 1 Introduction The goal of subproject 3 SP3 Data Management of euroFOT is to provide robust and flexible data management solutions for euroFOT Specifically the objectives of SP3 euroFOT 2008 are to 1 Provide the best suited data acquisition systems DAS for euroFOT 2 Provide the
134. ith respect to quality The quality checks are performed on per sample per measure and per trip basis Pre processing data verification To ensure that the pre processing in itself does not cause any significant alteration of the measure data a data verification step has been added at the end of the process This consists of a number of checks and comparisons between the processed data and the corresponding raw data briefly described below Time vectors are checked to ensure that all values are consistent i e monotonously increasing and evenly spaced Signal values are checked to determine if they are constant or consist of invalid NaN values only and that the raw and processed data are identical in such cases Mean value and standard deviation of raw and processed data are compared This test is passed if the mean value and standard deviation differs by less than 5 of the root mean square RMS value of the signal By means of cross correlation signal similarity and potential time shift is determined The test is passed if the cross correlation coefficient is more than 0 9 and the time shift is less than one second For enumerative type signals an additional check is performed to ensure that the number and values of the signal levels in raw and processed data are the same Deliverable D3 3 lt Version 1 0 gt 75 Data management in euroFOT euroFOT 01 08 2011 The results from the pre processing data verification are stored in
135. ithms Analysis and validation Figure 34 Pre processing 3 2 1 3 Lessons learned Designing the data storage and database structure faced several challenges Initial strategy and specifications related to database configuration were updated and amended many times during project evolvement New challenges emerged which we did not think about at first place and needed to overcome So the experience obtained by the whole activity on the project provided us invaluable knowledge in huge dataset manipulation techniques Limitations and issues Main difficulty was the design of the database schema and the methodology on storing data in a relational database system The time it takes to upload large datasets from field operation tests is critical That is if the sum up of all the steps that needs to be performed for data uploading is for example at the same speed as it was recorded then 20 hours of driving data would take up to 20 hours to upload This is unacceptable and not reasonable since data collected from multiple drivers may need to be uploaded in one day Also the way that analysts need the data influence the design of the database schema It is necessary to avoid performing joints between tables with millions of rows of data during the analysis phase to have them presented in useful way Another issue was the foreign keys which needed for database integrity So creating an optimal database schema is optimal for database s desirable operation
136. ive analytics 2011 for handling the data from the questionnaires and the information about the drivers After all the data from the questionnaires both paper and electronically collected are in LimeSurvey tool see also Section 3 2 4 they are then exported into Microsoft Excel In Microsoft Excel the data are pre processed and quality procedures are applied to prepare the data suitable for analysis according to the Data Analysis Plan of euroFOT SP6 The advanced statistical analysis will be done with statistical packages such as SPSS 4 3 Remarks This chapter described data analysis tools implemented in each VMC The data analysis software has accomplished the given goals and has met the user needs Namely the software is e User friendly interface facilitating the annotation process e Flexible and not limited to specific research questions allowing users to define annotation categories e Expandable so that the users could implement new modules and functions e Interactive with timeline based visualisation of data Data management in euroFOT euroFOT 01 08 2011 e Efficient in retrieving information from both file management systems and relational databases Further improvement is possible by augmenting the tool with a command line interface and by integrating functions for statistical analysis Deliverable D3 3 lt Version 1 0 gt 112 Data management in euroFOT euroFOT 01 08 2011 5 Summary of Lessons Learned a
137. l instrumentation a DATALOGGER set ee p lt 4 amp 4j y A f N 4 1 N Figure 13 High level instrumentation at French VMC Deliverable D3 3 lt Version 1 0 gt 25 Data management in euroFOT euroFOT 01 08 2011 The high level instrumentation starts from the low level instrumentation and adds the following elements e Videologging system videologger and cameras e Lane tracking system Mobileye e Eye tracker Smarteye 2 2 2 3 Final implementation to fulfil euroFOT requirements CAN only system Same implementation as in German 1 OC see Section 2 2 1 3 The only difference in CTAG Datalogger Il that had to be implemented specifically for French VMC was the synchronization mechanism which consists on some frames received from the videologger by the Datalogger serial port The Datalogger acknowledges the reception and correction of the frames and saves the information of the frames within the log so that the video frames can have a time stamp with the same reference as the CAN data one Videologging system The videologging system Figure 14 consists of a videologger and videocameras Figure 14 Video logging system at French VMC It was initially thought that off the shelf video surveillance systems could fulfil all those needs However after an extensive screening process it appeared that no such system existed Some systems were autonomous embeddable and cheap but
138. latency and jitter for each signal source 18 The synchronization inaccuracy between different signal sources must not be larger than 100 ms When post processing the data based on the latency and jitter figures 19 The DAS must autonomously log data from all signal sources CAN video LIN or digital I O J1708 GPS etc There must not be any driver interaction needed at all 20 When the ignition is turned on the DAS must automatically boot up and start logging 21 When the ignition is turned off the DAS must automatically stop logging and turn off The shut down time should be as short as possible 22 The operation of the DAS must be controlled by a configuration file 23 In the configuration file it must be possible to specify parameters for each signal source where applicable Such parameters might include but not be limited to CAN baud rate Which CAN frames or signals to collect Video frame rate Video quality 24 It must be possible for our personnel to edit the configuration file and to update the DAS with the new file 25 The DAS must have functionality to automatically connect to a server via GPRS 3G and check if there is any updated software or configuration file If there is the new software and or configuration file should be downloaded 26 The DAS must have functionality to automatically connect to a server via GPRS 8G e g once per trip and upload a status report containing the current status of t
139. lds of a minimum trip duration of 120 sec minimum trip length of 100 m and that not more than 95 of the trip has zero speed Furthermore it is checked that the CAN source is available for the trip and that the per measure quality of the vehicle speed is good The quality measures from all of the above mentioned steps are presented in the segment info and the measure info in the data base structure In the processing and quality of data from pilot data the following indicative data and processing quality have been found GPS data is generally available for all trips and there are reliable data around 95 99 of the time Lack of data is mainly due to obstructed signals during tunnel passages etc CAN data is available more or less 100 of the time Eye tracker data is available for most trips and there are reliable head position and gave angle data generally in the interval 30 70 of the time This fairly large variation is probably due to differences e g in head position in relation to the camera for different drivers With respect to speed One aspect of pre processing is that it must take significantly shorter time to process than to record to be able to keep up with data delivery This has been a guiding factor in implementation with focus on identifying bottlenecks Many parts of the process are so fast in relation to the bottlenecks that no time is needed to be spent to optimize these parts Also due to the hardware availability
140. lla org http www ipcmax com datasheet Nexcom_VTC_3300_Transport PC pdf 2007 08 03 Retrieved 08 08 2011 from http www ipcmax com datasheet Nexcom_VTC_3300_Transport PC pdf IBM SPSS software for predictive analytics 2011 Retrieved 05 30 2011 from http www spss com Jamson S Chorlton K Gelau C Schindhelm R Johansson E Karlsson A S et al 2009 Deliverable D4 2 Experimental Procedures Version 0 7 euroFOT Kircher K Heinig I amp Brouwer R 2009 Deliverable D4 1 Performance Indicators Events and Situational Variables within euroF OT Version 1 3 earoFOT Kovaceva J Leon Cano J A Sanchez D Dozza M Selpi Val C et al 2011 Deliverable D3 2 Base Functionality Data Analysis Software Updated Version Version 2 0 euroFOT Kvaser USBcan Professional 2011 Retrieved 08 30 2011 from http www kvaser com index php option com_php amp Itemid 258 amp eaninput 7330130003576 amp lang en amp product K vaser 20USBcan 20Professional LimeSurvey the free amp open source survey software tool 2011 Retrieved 05 30 2011 from http www limesurvey org Matroska Media Container Homepage 2011 Retrieved 08 08 2011 from http matroska org MOBILEYE AWS 2000 2008 Retrieved 08 30 2011 from http www clifford nl pdf mobileye UK U AWS 2000 pdf MPEG4000XLP 2011 Retrieved 08 30 2011 from http www ampltd com dload mpeg4000xIp pdf MySQL The world
141. man 1 OC 50 FOUG ZOrIrFaMCWONKs na as nn ea dd aa A nd AR one AR 50 Figure 29 Database schema at German 1 OC and French VMC 52 Figure 30 The left part of the database schema at German 1 OC and French VMC 53 Figure 31 The right part of the database schema at German 1 OC and French VMC 54 Figure 32 General layout of data processing at German 1 OC and French VMC 55 FigUre S3 ANalySiS 010 PR ease cuetoiescusbensasouteiescusbuisasonteersuidensssnntonsscyaeensediut 56 FOUG 34 Pre DOSSIER a ee a a ere ree 57 Figure 35 Overview database and data storage VU 60 Figure 36 Overview of data flow from the vehicles to the VMC centre in the Swedish VMC 62 Deliverable D3 3 lt Version 1 0 gt vi Data management in euroFOT Figure 37 Figure 38 Figure 39 Figure 40 Figure 41 Figure 42 Figure 43 Figure 44 Figure 45 Figure 46 Figure 47 Figure 48 Figure 49 Figure 50 Figure 51 Figure 52 Figure 53 Figure 54 Figure 55 Figure 56 Figure 57 Figure 58 Figure 59 Figure 60 Figure 61 Figure 62 Figure 63 Figure 64 Figure 65 Figure 66 Figure 67 Figure 68 Figure 69 Figure 70 Figure 71 Figure 72 Figure 73 Figure 74 Deliverable D3 3 euroFOT 01 08 2011 The database structure used in the Swedish VMC 63 Part 1 of the database structure used in the Swedish VMC 64 Part 2 of the database structure used in the Swedish VMC 65 Part 3 of the database str
142. man 2 OC Swedish VMC Italian VMC Estimated size of data storage Up to 12 TB Up to 8 TB rough 25 TB Up to 100 TB 450MB MySQL for pilot Database engine Oracle for MySQL MS Access Oracle MySQL PHP production Mat files and Other storage sh a ae VUEG Mat files in filesystem video in filesystem Video Y N N Y Y Y N Logger Y N Y Y Y Y N Questionnaires Y N Y Y Y Y Y Enrichment tools Y Y Y Y Y Fleet number of vehicles 200 40 30 130 5 0 Estimation of number of recorded driving hours 120000 20000 16600 80000 NA Deliverable D3 3 lt Version 1 0 gt 48 Data management in euroFOT euroFOT 01 08 2011 3 2 1 German 1 OC amp French VMC 3 2 1 1 Data storage solution The solutions chosen for the German 1 OC amp French VMC are similar in terms of software architecture processes and database schema but is different in certain details The Similarity is mainly driven by the use of the same data logger CTAG logger for CAN only logger At both places it was investigated which methodologies could be applied for the data retrieval processes as well as for monitoring and data analysis processes of the FOT Based upon duration number of vehicles and vehicle distribution German 1 OC and French VMC classified their FOT scenarios Figure 26 and selected a mostly automated processing chain for all three project phases data retrieval pre and post processing including enrichment result evaluation The automated approach in all steps of the
143. management in euroFOT euroFOT 01 08 2011 Reading in of relevant parameters Data logging includes more signals than needed for the euroFOT analysis only signals needed for euroFOT are kept for the db If necessary missing values are interpolated at this step Merging per trip Due to peculiarities of the data logging it is possible that one trip is logged in more than one data file Files belonging to one trip are merged Map matching The NAVTEQ tool is used to derive information from the digital map based on GPS position Adding of map based information The parameters derived from the map are merged with the vehicle data per trip Adding of derived parameters Algorithms have been implemented that derive new parameters from the logged raw signals The algorithms include adding of continuous signals e g standard deviation of lane position of categorical signals e g weather based on wiper status and temperature and of event detection e g incidents The derived parameters are added as new variables to each trip Adding of information from mobile device In the condition mobile device the information logged on the mobile device is added as new variables to the vehicle data whenever the mobile device has been active Saving in folder structure The data consists now of one file per trip that contains all the parameters needed for the euroFOT analysis The files are stored in the final folder structure that allocates driver ID a
144. med directly after data conversion The checks on trip basis are performed within the pre processing step Based on an error handling integrated in the framework errors due to missing or wrong information are automatically presented in order to identify the problem and define a suitable solution The way to check if data quality is acceptable or not and the way to proceed when the quality is not enough is different depending on where the data comes from On the highest level there are basically two kinds of data objective data which comes from sensors integrated in the vehicle and subjective data which comes from questionnaires filled by the drivers e Objective data on vehicle sensors o CAN data o GPS and map data e Subjective data questionnaires The following checks are applied in order to assess the quality of the incoming data by means of automated quality checks integrated in the framework e Missing data Because of malfunctioning sensors unfastened plugs faulty acquisition systems or human error there are many possible factors which cause data loss e Values out of range The acquired data contains information about the driver or the environment This information has a physical meaning and therefore also physical bounds When the registered values are out of bounds we may conclude that something has gone wrong For example if the velocity is 511 km h it is clearly erroneous e Wrong dynamic behaviour of data If the
145. ment any additional power supply except for the eye tracker Furthermore the VTC6100 has an elaborate power supply with input for both KL15 and KL30 directly from the vehicle PCI internal to the PC Internal to the PC Ampltd MPEG4000XLP MPEG4000XLP 2011 MPEG4 HW compression PCI The video grabber is fitted internally to the PC It has four CVBS video inputs The SW API generates a callback for each coded frame MPEGA4 Industrial specification Transcend 16 GB industrial grade The Compact Flash CF memory is used both for the Windows XP Embedded installation and for primary data storage during operation Seagate FreeAgent Go 500 GB USB external to the PC This HDD has two features that makes it suitable for this application e It inhibits spin up if temperature is below 0 C Spinning up a HDD at low temperatures can damage the HDD e t spins down after several minutes of inactivity Data from the previous trip is copied from the Compact Flash memory to the HDD at the beginning of each trip thereafter the HDD spins down It is believed that HDDs will last longer if not operated during driving Kvaser USB CAN Professional Kvaser Professional 2011 USB external to the PC Cars only Custom made by VCC USBcan Trucks only It is a two axis accelerometer with an additional yaw rate sensor To minimize power consumption in off mode Cars only For eye tracker power supply For driver to turn off
146. mers Dag Borjesson Previously at Chalmers Lena Kanstrup Previously at Volvo Jorge Leon Cano Previously at VCC Vassilis Kaffes ICCS Katia Pagle ICCS Leandro Guidotti POLI Roberto Tadei POLI Friedemann Kuhn DAG Mariana Just BMW We also would like to thank all other individuals that are not named here but have in some ways contributed to the work reported in this deliverable Deliverable D3 3 lt Version 1 0 gt 120 Data management in euroFOT euroFOT 01 08 2011 References Chalmers SAFER 2011 Retrieved 08 08 2011 from https www chalmers se safer EN Chalmers SAFER SeMiF OT 2011 Retrieved 08 08 2011 from https www chalmers se safer EN projects traffic safety analysis semifot Chalmers SAFER TSS Naturalistic Field Operational Test FOT Phase I 2011 Retrieved 08 08 2011 from http www chalmers se safer EN projects traffic safety analysis tss naturalistic field Csepinszky A amp Hagleitner W 2011 Deliverable D5 2 List of drivers vehicles equipment legal aspects and report on installation of the logging devices Version 2 1 euroFOT Dozza M Kessler C Petersson M Rakic M Tattegrain Veste H amp Heinig K 2010 Deliverable D2 1 Updated Specifications amp Requirements for Testing In vehicle Systems for Driving Support Version 1 7 euroFOT Embedded AntiSleep 2011 Retrieved 08 30 2011 from http www smarteye se productsoem embedded antisleep euroFOT conso
147. mmarized as follows e The datalogger is 110 W x 36 D x108 H mm and weights only 274 g excluding cabling Thus it is possible to install it in several places in the vehicle glove box compartment trunk space under driver or passenger seat etc allowing a naturalistic experiment approach e No fan no cooling device Noiseless device e Storage temperature ranges from 40 C to 85 C Operating temperature is located between 30 C and 70 C fulfilling euroFOT requirements regarding thermal stress e Power consumption limited to 23 mA in stand by mode CAN activity wake up mode Powering on by ignition signal means no consumption In this case shut down time is configurable If the battery voltage is below or drops under a certain threshold the system shuts down e Filtering options have been added to the configuration tool Using standard dbc files the device can be configured for extracting CAN signals from CAN messages Deliverable D3 3 lt Version 1 0 gt 19 Data management in euroFOT euroFOT 01 08 2011 e Triggering conditions for data filtering can be programmed by callback functions Python Scripting By data compression also implemented the transmitted amount of data was reduced to 1 3 of the logged data In case of both German 1 OC and French VMC the data stored in the device are not readable for security reasons before applying specific euroFOT dedicated conversion tools e Data in euroFOT is transmit
148. model itself these evolutions are automatically detected and if validated result in automated database structure evolution and refreshing of data where only impacted steps are re evaluated Beneath the data analysis tool used in both VMCs all enrichments are executed by embedding algorithms to the TDM Algorithm Execution Plan Distributed Execution computation cluster Definition and Encapsulation using versioning and dependency analysis using meta information performed by Petri Nets Start Ww encapsulation adding meta information gt required inputs i gt generated outputs gt versioning on distribution of parallel algorithms re calculation due to affect analysis by algorithm evolution Vx y t t date yyyy mm dd W w algorithm definition amp implementation committing to server system y V1 0 automated analysis of i G JE V1 0 le gt dependency and versioning i algorithm evolution End w 4 V1 0 ear A als forit m poi ol Pl algorithm classification quality check affect analysis Figure 33 Analysis tool Deliverable D3 3 lt Version 1 0 gt 56 Data management in euroFOT euroFOT 01 08 2011 Server Data Mining Data Mining Data Data Mining Data _ Phase 1 Phase 2 Clean Up Phase3 RAWdata pre p rocess aggregated dat aggegateddata probalby adaptatioin and evolving of algor
149. mpared to other DAS used in euroFOT The trigger box has the following functions e Record audio comment After audio comment button is pushed an audio comment is recorded The default time setting is 15 seconds and can be restarted every time the measurement system is in state running The start and end of an audio comment is announced by a double beep e Record driver familiarity with the route Button Route status Initially the route status is set to 0 After start up of the measurement system the system beeps and blinks to advice the driver to select whether the route is 1 Known or 2 unknown Deliverable D3 3 lt Version 1 0 gt 43 Data management in euroFOT euroFOT 01 08 2011 Main components in spare wheel compartment gt y Driver view with attached microphone _ Camera with view to navigation system Rear view camera Figure 23 Placement of DAS in a Daimler vehicle This figure is adapted from euroFOT Daimler installation guide 2 2 4 4 Test and quality check performed BMW and Daimler follow their standard internal their company tests The details of the tests are not described here but can be found in Deliverable D2 2 Saint Pierre Kessler and Malta 2011 To monitor that all FOT vehicles and data acquisition systems are working properly at any time a web based monitoring page was also built by MBTech for Daimler Figure 24 No such tool is used by BMW Deliverable D3 3 l
150. n 0 23 2011 08 10 Version for internal review 1 0 2011 08 31 Final version for submission Deliverable D3 3 lt Version 1 0 gt ii Data management in euroFOT euroFOT 01 08 2011 Table of contents FRY IST OTANI SUT CORRE a e an as denses ce oo tee 11 PARIS CE COME CIS eee ee EE 111 SCORE tenet aene V TOLTI U e E E E E E E EE vi X CUUIV SUMMAT serseri EEEE E EEE EEEE ne E E i OO E E eee 2 1 1 Vehicle Management Centre VMC siens 4 12 OM etal WV A ON OR a ee etes ce Aou et 6 1 3 DiS AI SALOME 01 DS eses asnnctocide aes ease etsudecae R 6 2 Mata AC GUIS System DAS cercei e ee aE Eea E AEE 7 2 1 MenO ETOC dro DS a nat do 7 2 1 1 From functions and hypotheses to signals and sensors cccccccceeeeeseeeeeeeeeees 7 22 DAS Solutions Used in euroFOT ss 9 OR E E E E E E E 13 PA A CAS EE OM E E E A E A eessactsactosetedesesnet 13 2202 O E WA OM A a ot 14 2 2 1 3 Final implementation to fulfil euroFOT requirements seeeeeeeeeeeeesssseeeeeeee 16 2 2 1 4 Tests and quality check performed 20 22 155 ExXp n enees and lessons learned 2 ni 21 DDD FCIAC A A CR RP E 23 222 Ina CONCIAONS ess ses E nes etes EEEE R eee 23 Delle CO SOON ht cars A 24 2 2 2 3 Final implementation to fulfil euroFOT requirements cc eeeeeeeeeeeeeeee 26 2 2 2 4 Tests and quality check performed 29 222 5 Experiences and lessons CARR ressens se rS 29 LS DN ENM E 30 o A E E EES TE a E A AE T E A 30 22 MOTOS E
151. n of the platforms that we consider video to be a very important source of information when dealing with naturalistic driving data Our view is that in most cases when analysing driver behaviour in naturalistic data it is necessary to include analysis of the drivers actions and the context in vehicle as well as surrounding traffic environment based on video The video annotation module presented here has been significantly inspired by the 100 car naturalistic driving study FESTA methodology and to optimize the procedures to fit different studies several parallel projects have been giving input to its design The different tools and interfaces needed in and between the different platforms will be further developed as new projects and data emerge and the platforms will grow and develop to fit specific needs brought forth by researches and engineers Deliverable D3 3 lt Version 1 0 gt 117 Data management in euroFOT euroFOT 01 08 2011 6 Conclusions This deliverable D3 3 describes the achievements made by euroFOT partners particularly in SP3 in terms of data management Five different data acquisition systems used in this project have been presented in this deliverable First is the CTAG Datalogger II CAN only logger that was used in German 1 OC and French VMC for the low level instrumentation Second is the CEESAR videologging system which is a combined CTAG Datalogger Il and video logger based on Nexcom VTC3300 with radar eye tr
152. n s operate in silent mode that is they collect data but do not give any signals to the driver From the viewpoint of the driver the function s is are off The part of the data collection during which the function s are switched on by the experimental leader such that they are either active all the time or can be switched on or off by the driver These are those factors that are kept constant within one analysis The data are filtered such that only occurrences in which the controlled factors assume the intended values are selected These are covariates they are not kept constant within one analysis but their values are logged and their influence on the results is considered Quantitative or qualitative indicator derived from one or several measures agreed on beforehand expressed as a percentage index rate or other value which is monitored at regular or irregular intervals and can be compared to one or more criteria lt Version 1 0 gt B Data management in euroFOT DAS related terms ADR Avi CAN DAS Driving cycle EMC Eye and head tracker HDD GPS GPRS J1708 LIN Radar RTC Signal source VCI SSL Deliverable D3 3 euroFOT 01 08 2011 The European Agreement concerning the International Carriage of Dangerous Goods by Road Multimedia container format and filetype Avi can be used as a placeholder for different media codecs as xvid or h 264 Controller Area Network
153. n tigris org TrueCrypt 2011 Retrieved 08 30 2011 from http www truecrypt org VTC 3300 2007 Retrieved 08 30 2011 from http www ipcmax com datasheet Nexcom_VTC_3300_Transport PC pdf VTC6100 2010 04 05 Retrieved 08 08 2011 from http www rm systems co uk wp content uploads 2010 07 VTC6100 pdf Welcome to Apache Hadoop 2011 Retrieved 08 08 2011 from http hadoop apache org Welcome to Hive 2011 Retrieved 08 08 2011 from http hive apache org Wittenburg P Brugman H Russel A Klassmann A amp Sloetjes H 2006 Elan a professional framework for multimodality research Proceedings of Language Resources and Evaluation Conference LREC 2006 Xvid org Home of the Xvid Codec 2011 Retrieved 08 08 2011 from http www xvid org ZFS Wikipedia the free encyclopedia 2011 Retrieved 08 08 2011 from http en wikipedia org wiki ZFS Zimmerman P Bolhuis J Willemsen A Meyer E amp Noldus L 2009 The Observer XT A tool for the integration of and synchronization of multimodal signals Behavioural Research Methods Volume 41 Issue 3 131 735 Deliverable D3 3 lt Version 1 0 gt 123 Data management in euroFOT euroFOT 01 08 2011 Annex 1 Keywords Analyse Annotation euroFOT CAN DAS Data acquisition system Database Data logger Data storage Data Analysis Tool GUI Logger Market survey Measures Requirements specification Sensors Software Architectu
154. nagement in euroFOT euroFOT 01 08 2011 3 2 1 2 Upload and enrichment tools To reach the aim of a mainly automated processing chain the implemented approach of the German 1 OC and French VMC distinguishes strictly between the framework and embedded algorithms in an algorithm pool encapsulated by the priory mentioned framework The framework was designed to be of universal use for the data enrichment procedures and algorithms execution in all steps of the project pre post processing and data enrichment follow the same structure and are described by the Trip Data Model TDM The TDM describes all artefacts handled by the implemented framework Artefacts of the meta description are signals data structures attributes events situations and algorithms including the enrichment algorithms on meta level This meta level information is used by the framework to bind incoming data and external data sources like geo map data together with the algorithms from the algorithm pool lt determines a correct execution order of the algorithms and keeps track of evolving algorithms For this purpose versioning information is included in all generated output artefacts of the framework which guarantees consistent data at every time of the project The versioning information is as well used to determine if already processed data sets require reprocessing due to evolved algorithms Hence when evolutions are made either inside an algorithm or in the data
155. nation conditions colour cameras were dismissed not only because colour isn t necessary for analysis and requires larger storage but also because even last generation cameras didn t show the same sensitivity and dynamic range as the black and white cameras used Specific automotive cameras were also considered and tried for the front scene but their interfacing was not standard connectivity and power supply they were more expensive and the benefits they brought in high dynamic range compensation were not sufficient to justify the effort Feedback from pilot testing showed that with typical driving scenarios video compression on 4 channels was sufficiently efficient to allow recording nearly 200 hours of video on the single embedded 80 Gb automotive hard drive This proved much more than enough for euroFOT requirements Video records are transferred to the storage server by network connection when vehicles are at CEESAR using a simple windows batch script As video files are directly recorded with the very efficient H 264 codec no further re encoding is necessary and they are directly used as is by the data analysis tool Synchronisation information being transmitted regularly during recording from the videologger to the datalogger they are embedded inside the CAN amp GPS data acquisition files Compensating for latencies in video encoding and network transfer between framegrabber and embedded PC very precise synchronization was obtaine
156. nce to next crossing rs j CAN video GPS Sensor data CAN video GPS Complete FOT data Matlab merge FOT safety script analysis Sensor data GPS coords Matlab extract Road attributes csv format ry ASP PE ah ee ales L jane ow GPSToRoute Volvo ADASRP plugin Navteq license required Figure 47 Overview of the process of complementing logged data with map attributes 3 2 2 3 Lessons learned Limitations and issues Data quality is difficult Even if a lot of quality checks are made in the data management chain it is difficult to utilize this easily e g what is poor quality for different types of analysis for different measures and events How can it be made transparent to the users what quality measures to use and what they mean How we proceed The use of data quality will have to be in focus for the initial analysis and simple guidelines have to be written Even if this has been done to some extent in the piloting the main analysis is still ahead and with that the understanding of use of quality Implementation validation must be done iteratively Lessons learned from previous studies have shown that manual review and evaluation of ALL measures and events before production upload should be done This is something that has to be done several times and takes significant time but it is crucial for the quality assurance and good analysis Availability of map data For several r
157. nd Recommendations In the previous three chapters we have described the data management solutions used in euroFOT The variety of solutions used allows us to learn a lot Here we summarise the lessons we learned from this project from the data management point of view It is our hope that the lessons we list here and the recommendations we make here could benefit future projects The following way of work proved positive 1 Using common templates and common terms to gather the requirements were very helpful in particular these help during harmonisation and prioritisation of the requirements 2 Working together for common interests has proven very constructive For example the French and Swedish VMCs_ worked together in defining requirements specification for several sensors conducting market survey as well as in evaluating some of the possible solutions This increases the throughput and reduces the burden on both VMCs 3 Collaboration with SP2 the OEMs that know the inside and outside of the vehicles SP4 responsible for study design performance indicators and list of measures SP5 that manage the FOT and SP6 responsible for analysis from early time was absolutely useful and necessary Their inputs are the base requirements for the data management Review on quality assurance 4 Setting quality threshold is still a challenging issue in FOT This issue has been raised by previous projects and FESTA but it is wor
158. nd condition to each trip Before storing the data in the final structure an overview table is created that gives basic information for each trip 3 2 3 3 Lessons learned The whole chain of data pre processing upload and analysis should be tested if possible until Pl calculation before the FOT Some problems with signals e g strange values were realized very late in the whole process The data pre processing and upload procedures should be flexible and allow changes after the FOT has started This is necessary because new problems might occur that did not exist in the pilot For instance in the Daimler fleet vehicles data logging continued sometimes for hours after the end of a drive As a consequence more than one drive were often logged in one data log This has not occurred in the pilots because due to an error in the logger installation the shut down procedure did not work properly in the fleet vehicles but it worked well in the pilot car It took some months to realize and find the error and then to change the cabling in all fleet vehicles Because of the error a new step had to be included in the data pre processing procedures that now separates each data log into single drives Data handling especially of video data should be tested with a large amount of data already in the pilots In the beginning also the Daimler video data was stored as single frames instead of video format Although there were a large number of separate
159. nd support H amp S service is provided by the TRG people and will be available during the extent of the euroFOT project FOTware users who know the TRG members can write emails and or make phone calls to specific persons in order to get assistance It is also possible to leave messages in a log book where users can write errors additional suggestions for improvement etc that can be found in the FOT analysis room However this log book is recommended for non urgent issues All users that do not know who to contact should in first instance contact the analysis tool responsible at the Swedish VMC centre All trouble issues found and their constantly updated statuses shall be reported in one document It is important to keep track of all relevant follow up actions and also the name s of the assigned TRG member s who is are dealing with each support case All TRG members should have access to this document in order to update it accordingly A FOTware user manual is also available In this document users can find directions for making a proper use of this tool Since FOTware is modular and customisable it is used by different users for different data analysis purposes Further since FOTware was developed in collaboration with SeMiFOT also the SeMiFOT project will make use of it At the current time the tool is used mainly for annotation of video and signal quality inspection In addition since the tool started to be used there was a contin
160. ne is Oracle 11g Oracle Hardware and Software Engineered to Work Together 2011 The database schema used in the Swedish VMC is shown in Figure 37 Overall the FOT data can be divided into several groups based on whether it is a raw data enriched data and also based on who can access which data to ensure consistency and security Therefore access rights are enforced in both the Oracle database and the fileserver Deliverable D3 3 lt Version 1 0 gt 61 Data management in euroFOT euroFOT 01 08 2011 Q SY P nd gt 2 Questionnaires Status report sent wirelessly Hard disk Empty Hard disk CAN data eye nn tracker data Oracle subjective data At each OEM e Do pre processing steps e Upload data to OEM de PR e Copy parts of the pre Empty transfer processed data to transfer disk At VMC centre Transfer disk e Do daily monitoring of status reports Upload data from transfer disk to Oracle database and fileserver for video Empty transfer disk Video data Matlab files subjective data Figure 36 Overview of data flow from the vehicles to the VMC centre in the Swedish VMC Deliverable D3 3 lt Version 1 0 gt Data management in euroFOT 62 euroFOT 01 08 2011 Figure 37 The database structure used in the Swedish VMC This structure is split into six parts to make it readable see the next six figures Deliverable D3 3 lt Version 1 0 gt 63 Data
161. nstead most full scale tests were done in real vehicles For example the Compact Flash problem would most likely have been detected in an earlier stage if the system had been left continuously running for one or two weeks in a test rig As it was now the pilot car was run for several months without failure Configuration of each logger with a unique identifier takes a lot of time One has to connect a monitor and keyboard and log into the system to alter a file Instead another method would have been beiter e g by letting the SW application read out the serial number of the PC or something similar 2 2 4 German 2 OC The partners involved in German 2 OC are not members of SP3 Their works to select and finally made available DAS systems for FOT data collection in this OC were made outside of SP3 Here only a summary of DAS systems used in German 2 OC is presented Further detailed can be read from Deliverable D2 2 Saint Pierre Kessler and Malta 2011 and Deliverable D5 2 Csepinszky and Hagleitner 2011 2 2 4 1 Initial condition Loggers CAN Only CAN Extra CAN Video Vehicles sensor s Extra sensor s Daimler Cars 15 BMW Cars 15 Within euroFOT only German 2 OC studied safe human machine interaction safe HMI systems Specifically they studied the BMW and the Daimler solutions for built in navigation systems and mobile devices Z205 manufactured by Harman Becker For the purpose of this project the used mobile dev
162. nsufficient Spec Dev Incomect Tester Request User Accept FM Validate Tester Validated Clarify PM Close PM Not accepted PM Figure 65 New feature change request Implement Dev Implementation Completed Complete Dev Incorrect Tester Request User Validate Tester Clanfy PM Validated Close PM Not Accepted PM Figure 66 Improvement 4 2 1 3 Lessons learned Difficulties and Solutions The definition of use cases in the beginning tended to be useful to get a basic understanding of the tool chain However designing a modular platform lead to the need of many changes for these use cases The use cases and the implementation needed to be reworked during the development and evaluation Sharing the development work among three different project partners made synchronization of the work difficult in the beginning Especially the usage of three version management system three bug trackers and the exchange of files lead to conflicts This was finally solved by defining clear workflow schemes and by unifying the development to on central version management and bug tracking system Deliverable D3 3 lt Version 1 0 gt 97 Data management in euroFOT euroFOT 01 08 2011 Different scenarios for usage of the tools made a modular approach challenging as interfaces of the software components needed to be matched in all types of operation e g no avail
163. nt Event Low Medium H Back View AIS 1 on T an a eree Precipitating event RHT alking singing danang applic 2 pen Initial Conflict Type This Vehicle Lost Control Blow out or fl i j Side View sve Hands on wheel internal distraction not vehicle or passenger Conflict with a lead vehicle RUE basse Good Bad Not applic i Reading a i Precipitating event LHT o hands One hand Two hands Unknown Right or left hand traffic Secondary Conflict Type P g Wireless device Right hand Left handt Unknown Conflict with a lead vehicle This Vehicle Lost Control Blow out or fl Talkingfistening Horizontal alignment RHT Near Crash type Seat Belt A rs Straight OR Gee Start of precipitating event CRE start tiated a sae Vehicle related secondary task 9 lo nown Horizontal alignment LHT Edit TIME h Adjusting climate control i ere Straight Conflict partner type Dini ning Vertical alignment None Evasi RHT Visual obstructions P VOSIVO manoeuvre TETEE Eating with Eating with Drinking w Drinking o Activity Conflict partner location RHT No driver present DE mr Number of lanes Rain snow fog smoke sand dust Smoking In front of subject Reflected glare 1 Headlights Reaching f Lighting ci Smoking c _ Extinguishi Locality Conflict partner location LHT Evasive manoeuvre LHT Curve or hill 2 No driver present Building billboard or other roadway infrastructur Daydreaming Urban In front of subject Trees crops vegeta
164. on are 1 the installation strategies and verification which was used by SP2 for pilot installation and SP5 for FOT installation 2 the maintenance strategies like the use of web based monitoring tool procedure to deal with DAS related problems during data collection and procedure for data handling Deliverable D3 3 lt Version 1 0 gt 12 Data management in euroFOT 2 2 1 euroFOT 01 08 2011 German 1 OC 2 2 1 1 Initial condition Loggers CAN Only CAN Extra CAN Video Vehicles sensor s Extra sensor s Ford Cars 100 MAN Trucks 100 VW AUDI Cars 40 The numbers above shows the plan in July 2009 Only 60 MAN trucks finally joined the FOT The main requirements constraints affecting DAS choice in German 1 Operation Centre and French VMC for the CAN only logger There must not be any integration or use of additional sensors beside the DAS itself From the vehicle itself only data available on CAN busses can be collected GPS information location of vehicles is not available on CAN bus but must also be collected DAS should be a small device that can be integrated in the vehicles without significant noticeable modifications to the vehicle Drivers should not be bothered with change of hard discs configuration of DAS or upload of data DAS should be remotely configurable DAS must be able to store collected data on a temporary storage on the DAS and when possible send the logged data autonomously o
165. on instrumentation CTAG Datalogger II CTAG Data logger Il and Nexcom PC VTC3300 2 4 Embedded Linux SDHC up to 8 GB Long range ACC radar 200m Built in the data logger lt Version 1 0 gt Embedded Linux Windows XP SDHC up to 8 GB and 80GB automotive hard disks for video 4 dashboard view forward rear feet Long range ACC radar 200m Built in the data logger 10 Swedish VMC Nexcom PC VTC6100 Windows XP Embedded compact flash card 16 GB 500GB office hard disks For cars 4 forward view rear view driver feet For trucks 4 forward view blind spot driver cabin feet Existing vehicle radar via CAN External GPS German 2 OC Daimler BMW PROVEtech V ReadyNAS A 2100 4 4 Linux Embedded Linux 3 USB disks 4x 1TB hard disks 4 driver face forward view rear view navigation system Existing vehicle radar via CAN External GPS and existing vehicle GPS 4 driver face forward view rear view and centre stack to see driver hands and legs 2 Short range radar sensors Existing vehicle GPS euroFOT 01 08 2011 French VMC German 2 OC Categor German 1 OC i Swedish VMC ik Light Heavy Daimler BMW instrumentation instrumentation Lane tracker Existing vehicle Mobileye Existing vehicle Existing Existing information via information via vehicle vehicle CAN CAN information information vi CAN Accelerometer Existing vehicle Exi
166. ongs to a selected trip The program works with video data stored as single frames as well as with standard video format After starting the program a drop down menu can be opened that shows all trips per driver see Figure 80 trips for driveriD 5 can be selected Here the experimenter can select the trip for which he wants to view the video iix SEF406 20100427 __194224_1 SEF406 20100427 __194224_2 Trip SEF406 20100428_132622_2 SEF 406 20100428_201201_1 SEF 40620100429 _135934_1 SEF 406 20100429_135934_2 SEF 406 20100429_135934_3 SEF406 20100429_135934_4 Ende SEF406 20100429_194638_1 SEF406 20100429 215510 _1 SEF 406 20100430_052123_1 Anfang 28 4 2010 5 45 27 28 4 2010 6 14 40 SEF 406 20100430_134558_1 QUE SEF406 20100430_134558_2 SEF406 20100430 134558 3 19 4 SEF 406 20100430_153150_1 Dee SEF 40620100430 200345 1 SEF 406 20100501_184516_1 SEF406 20100501_ 200650 _1 SEF406 20100501_212455 1 SEF406 20100502 140705 _1 SEF406 20100502 140705_2 SEF 406 20100502_140705_3 SEF 406 20100502_172202_1 FAR PRE SEF406 20100502 205652 1 Fahrt Laden _Abbruch SFF4NA 20100502 233549 1 xi Figure 80 Selection of trips for driver ID 5 Deliverable D3 3 lt Version 1 0 gt 108 Data management in euroFOT euroFOT 01 08 2011 After selecting a trip the vehicle data and the video data are loaded From all parameters in the mat file raw map derived variables variables can be selec
167. or converting the logged raw data into even spaced time series data in mat format MATLAB format The converted CAN data and the video data are copied onto 2TB external hard drives Whenever a hard drive is full the OEM sends it via mail to the University of Wurzburg Here data pre processing and upload to the database take place The server for the German 2 OC is installed at the location of WIVW GmbH in Veitshochheim and it contains data from BMW and Daimler 3 2 3 1 Data storage solution Data backup routines Raw data Backups for logged raw data as well as for video data are run at BMW and Daimler At IZVW two copies of the raw data are stored on external hard discs Deliverable D3 3 lt Version 1 0 gt 79 Data management in euroFOT euroFOT 01 08 2011 Pre processed time series data The pre processed CAN data is stored as mat files on the server at IZVW There are 2 hard drives with 2TB each One of them is used for BMW data and the other one for Daimler data For both internal hard drives of the server a backup is run every weekend Other data This section relates to the following types of data e description of the data structure e mat structures containing the used variable names e MATLAB routines for uploading pre processing and analysing the data e MATLAB routines for analysing the data e other routines e g for data conversion extraction of map based data e overview tables generate in the process of data analysis
168. perator discovers them from the video Based on the video the signals and operators own experience the operator can discard an event uncheck the check box Valid Event in case the automatic detection was a false positive He can also annotate it with comments in the text field and modify some editable attributes for instance by manually selecting values on a graph display After each modification the events attributes which are calculated and dependent on some of the edited attributes are refreshed by the framework and the results are displayed in the form This manual observation and validation of the detected events also allows troubleshooting and refining of the detection and characterisation algorithms IV Valid Event Braking Event Parent Event None Duration 57 Speed at beginning Speed at the end 26 Speed reduction during braking 3839 Maximum deceleration 23 Maximum braking pressure 135 ABS activation Gear engaged at begoning Gear engaged at the end Figure 58 Event module Trip data model front end tool This tool allows performing all modifications in the data model e g adding a new variable defining a new script It supports definition or modification of a new object in the trip data model signal situation event and trip attribute by using the Experiment Builder shown on Figure 59 The experiment builder calls the enrichment manager which runs a successive set of processes A process Is defined by its inputs
169. perties and instantaneous value from seven signals 105 The event browser module displaying some identified traffic events 105 A possible configuration of the annotation module inspired by the 100 car study a oso asa oreo E E E E 106 Selection of trips for driver ID 5 108 selection of variables to be displayed together with video 109 Display of video together with selected time series data 109 Jump to defined sections in the video based on values of a selected variable 110 Deliverable D3 3 lt Version 1 0 gt viii Data management in euroFOT euroFOT 01 08 2011 Executive summary Data management is essential for a successful field operational test or naturalistic driving study The complexity of data management tasks and thus the resources needed are often underestimated This deliverable D3 3 Data management in euroFOT is a public deliverable of subproject 3 SP3 Data Management of euroFOT Its purposes are 1 to describe the data management solutions the data acquisition system database and storage and the base of analysis tool used in euroFOT 2 to summarise how we came to choose those solutions 3 to summarise the quality assurance checks performed throughout the whole data management chain 4 to share our experiences and the lessons learned The main contributions from D3 3 to euroFOT and a wider community include Five state of the art data acquisition systems
170. project made specialized control structures and well traceable process interactions essential Duration gt 1 year contionous regular A monitoring French VMC German1 VMC 6 month 3 month lt 1 month manual inspection lt 10 100 200 gt 300 Fleet Manual exchange of automated datatransfer storage media data retrieval Figure 26 Classification of French VMC and German 1 OC FOT scenario Whilst in the German 1 OC CAN only data is recorded and later enriched by geo map data the French VMC also is managing in addition to those data types video data Both VMCs have started an estimation of the amount of probable incoming data based upon the data recording requirements of SP6 as far as these were available In addition to that signals being of probable interest for SP6 were added to the amount of recorded data as well The initial data amount estimations lead to the need of investigating probable optimization potentials Therefore optimization on process for the incoming data amount from the field in the very first beginning of the project as well as optimization for processing pre and post processing and storage structuring had to be identified The results of the data amount estimations have been reviewed and were verified to decrease uncertainties during experiments prior to the start of the pilot of the FOT Deliverable D3 3 lt Version 1 0 gt 49 Data management in euroFOT euroFOT 01 08 2011 52k signal samples sec 52
171. r extra sensors were also investigated This was also done in the hope to get a better price per unit German 1 OC and French VMC conducted the market survey on data logger together and finally chose the CTAG data logger Since then the evaluation of CTAG data logger was led by German 1 This was to allow French VMC to focus their efforts on finding and evaluating video loggers and sensors To fulfil euroFOT requirements specification for both German 1 OC and French VMC further development on CTAG data logger was needed In this case CTAG worked together with German 1 OC and French VMC in finding the best way to fulfil the requirements specification Regarding external sensors eye and head tracker radar neadway sensing lateral positioning as well as cameras are of interest to both French and Swedish VMCs Therefore these two VMCs worked together in defining requirements specification for those sensors conducting market survey as well as in evaluating some of the possible solutions When compared to other VMCs the Swedish VMC has more experience on video acquisition Furthermore the Swedish VMC has done extensive market survey on both dedicated video logger as well as on vehicle PC which can also be used as video logger The French VMC has found that collaboration with the Swedish VMC in discussing requirements and candidate sub components of the video logging system has proved useful Accelerometer and yaw rate sensors were considere
172. r upon request to the server via any kind of WIFI or GPRS communication DAS should have ability to compress and encrypt the collected data DAS must be reliable and applicable for automotive application high durability with regard to vibration temperature etc Based on these requirements the following hardware components need to be integrated in the DAS CAN Interfaces at least for 2 CAN busses CPU unit GPRS modem Integration of SIM card upload via GPRS Integration of hard disc GPS receiver GPRS antenna GPS antenna Power supply unit Deliverable D3 3 lt Version 1 0 gt 13 Data management in euroFOT euroFOT 01 08 2011 With all of these requirements added to the common requirements a market survey was carried out to find a suitable DAS solution for German 1 OC and French VMC for the CAN only logger During the market survey several data acquisition systems were investigated by gathering all available information about these systems Thereby several aspects were examined Apart from the requirements on the DAS the costs were also considered Finally all gathered information was documented onto a database Based on this the most suitable DAS requirements costs etc was selected 2 2 1 2 Chosen solution From the requirements defined in Section 2 2 1 1 a market survey was performed and several DAS fulfilling the requirements were found Table 4 Result of market survey on CAN loggers Items are colour co
173. rable D3 3 lt Version 1 0 gt 109 Data management in euroFOT euroFOT 01 08 2011 Sprungprogramm 15 x Laden Speichern Einstellungen Extras Sher 100 lee 80 60 40 20 0 6 CHAHHELS2 CHAHHELS4 til 0 50 55 60 65 70 75 60 v a 1 a 0 6 0 4 Time 08 07 2010 13 12 16 20 dogo sesa du a chste Fahrt Marker Fahrer hr RENE 30 o Fahrt SEF402 a _nachster Wien 50 55 60 65 70 75 80 Suche entspricht von 1 CarFollow CarFollow Ze tBereich unten En Status Variablen bis 1 CarFollow CarFollow Buchert Figure 83 Jump to defined sections in the video based on values of a selected variable The calculation of parameters needed for the data analysis is not done based on a GUI instead code is written in MATLAB to extract the needed information Because the database consists of a large number of trips that are stored independently of each other all indicators are calculated first per trip or per relevant event e g incident The resulting tables are the basis for the final data analysis In the final analysis the tables giving information per trip or event can be used to plot indicators and to calculate aggregated indicators per driver per condition or overall The implemented routines create one table per driver This allows a repeated extraction of the needed indicators already before
174. rary providing access to video files using a wide variety of codecs in MATLAB SourceForge net videolO Toolbox for Matlab 2011 The video playback is controlled by the core module The video source in the video viewer can be changed by choosing the desired camera view from a popup menu It is also possible to choose among different camera views by pressing any numeric key from 1 to 5 where each of these numbers makes reference to the following available views Figure 70 1 Forward view Deliverable D3 3 lt Version 1 0 gt 100 Data management in euroFOT ee IN euroFOT 01 08 2011 Driver view Feet view Rear view Eye tracker view 7 Video viewer FOTware ole 1 Forward view Figure 70 A sketch of a video display module when the forward camera view is chosen 3 SQL module This module handles the access to the euroFOT Oracle database the execution of SQL statements and the retrieval and display of data Figure 71 shows a sketch of this module in which the following main items can be identified Tree menu that shows the hierarchy and names of the euroFOT tables and signals SQL statement history menu that keeps the statements that have been executed during a certain analysis session SQL statement text field that contains the statement that is executed when the button labelled Execute is pressed Before executing any SQL statement or populating the DB tree menu which requires reading the database metad
175. re SQL Upload strategies Video Video logger Visualisation Deliverable D3 3 lt Version 1 0 gt A Data management in euroFOT euroFOT 01 08 2011 Annex 2 Glossary FOT related terms and definitions are taken from the official euroFOT glossary Glossary FOT Net WIKI 2009 FOT Field Operational Test Function Use case Situation Situational variable Hypothesis Baseline period phase Treatment period phase Controlled factors Variable factors Performance indicator Deliverable D3 3 A study undertaken to evaluate a function or functions under normal operating conditions in environments typically encountered by the host vehicle s using quasi experimental methods Implementation of a set of rules to achieve a specified goal A specific event in which a system is expected to behave according to a specified function One specific level or a combination of more specific levels of situational variables An aspect of the surroundings made up of distinguishable levels At any point in time at least one of these levels must be valid A specific statement linking a cause to an effect and based ona mechanism linking the two It is applied to one or more functions and can be tested with statistical means by analyzing specific performance indicators in specific scenarios A hypothesis is expected to predict the direction of the expected change The part of the data collection during which the functio
176. res are the focus of SP4 Thus inputs from SP2 and SP4 were essential to make sure that the chain from hypotheses to requirements specification for DAS is preserved that is to say that at any point in time the reasoning of why certain signals or certain external sensors were necessary is clear Furthermore SP3 also collaborated with SP5 and SP6 as these two subprojects are direct users of SP3 SP5 is the subproject that is responsible for managing the field operational study FOT within euroFOT including handling ethical and legal issues and SP6 is responsible for conducting the analysis Therefore SP5 s and SP6 s needs were always valued and considered by SP3 to make sure that SP3 provides the right solution for the project This deliverable D3 3 summarises the work that has been undertaken mainly by SP3 partners towards achieving the objectives of SP3 Specifically the objectives of this deliverable are to e summarise the process we followed to select and finally made available data management solutions for euroFOT e describe the DAS solutions used in euroFOT e describe how we process and store the FOT data including what kind of quality checks done e describe the tools used in euroFOT to extract view and annotate the FOT data e share the experience and the lessons we learned through this project from an FOT data management perspective i i i i Lummemum j j
177. rked Hours Left Complete Gain 8 0 0 0 0 0 0 0 0 0 Deadline 2010 10 18 J rg K fen 2010 10 13 21 01 39 CEST Description rrent implementation of the NET MatLab library uses a lot of memory when lab fi is avoids currently pr i s F tlab 1 an es Th consumption shall be identified J rg K fen 2010 10 13 21 04 15 CEST Comment 1 Jorq K fen 2010 10 13 21 05 38 CEST Commen t2 ally communicated the deadline 2010 10 15 is Mond Home New Browse Search J 7 Reports Preferences Administration Help Log out kuefen ika nvth aachen de Figure 63 Sample Report for a reported Bug All reported features and requests are thereby documented and can be either directly be handled or be scheduled for future release versions For software errors bugs and software extensions changes different workflows and state schemas are used The process state graphs for the processes Software Correction Bug Feature Functional Change and Improvement are shown below Not accepted Tester Clarify PM Assigned In hs Resolved Validate Tester Validated Accept Dev Resolve Dev Assign Tester Close PM Reopen Tester Figure 64 Software error bug Deliverable D3 3 lt Version 1 0 gt 96 Data management in euroFOT euroFOT 01 08 2011 Implement Dev Specify Req Specification Implementation Completed Spec complete Reg Complete Dev I
178. rt output file naming scheme 510 kuefen 15 08 36 Mittwoch 18 Mai 2011 added suppoirt for new naming scheme to nmea to kml and dtl to mat converters 507 kuefen 14 43 54 Mittwoch 18 Mai 2011 integrated conversion tool with new global trip attributes 506 kuefen 14 42 42 Mittwoch 18 Mai 2011 504 kuefen 17 22 00 Dienstag 17 Mai 2011 added three example sets mhanli f eha evin ehall he Sf amavan Action Path Copy from path Revision Modified trunk euroFOT 1_Software MatLab Interaction Upload Processor Analysis ConversionBatch m Showing 100 revisions From revision 420 to revision 560 1 revision s selected E Hide unrelated changed paths Statistics Stop on copy rename Include merged revisions s a Show Al Next 100 ok Figure 61 Subversion Changelog The main development is executed in the so called trunk For releases and corrections of the software within a release the branches are used For each release of a software component a corresponding development branch is created which is marked by a tag before opening when merging back to the trunk and after closing a release The nomenclature of the tags is defined by PREP_ releaselD MERGE_ Cycle releaselD and POST _ releaselD The following graph shows the preferred usage of trunk branches and tags Post W X Y Prep final W X Y Release tag Release branch W X Y Release Releas
179. rtium 2008 Annex I Description of Work of European Large Scale Field Operational Tests on In Vehicle Systems euroFOT ext3 Wikipedia the free encyclopedia 2011 Retrieved 08 08 2011 from http en wikipedia org wiki Ext3 FESTA consortium 2008 Deliverable D2 2 Data requirements for FOT methodology FESTA FESTA consortium 2008 FESTA Handbook Version 2 FESTA ffdshow tryouts 2011 Retrieved 05 30 2011 from http ffdshow tryout sourceforge net Flament M Cavallo F Csepinszky A amp Hagleitner W 2009 Deliverable D5 1 Description of the VMCs and Common Guidelines Version 1 1 euroFOT Glossary FOT Net WIKI 2009 August Retrieved August 20 2011 from http wiki fot net eu index php title FOT_Glossary H 264 MPEG 4 AVC Wikipedia the free encyclopedia 2011 Retrieved 08 08 2011 from http en wikipedia org wiki H 264 MPEG 4_AVC Hagedorn J Hailpern J amp Karahalios K G 2008 VCode and VData Illustrating a new Framework for Supporting the Video Annotation Workflow Extended Abstract of AVI 2008 Hard Disk Sentinel HDD health and temperature monitoring 2011 Retrieved 08 30 2011 from http www hdsentinel com Deliverable D3 3 lt Version 1 0 gt 121 Data management in euroFOT euroFOT 01 08 2011 HBase HBase Home 2011 Retrieved 08 08 2011 from http hbase apache org Home Bugzilla bugzilla org 2011 Retrieved 08 30 2011 from http www bugzi
180. s and constraints methods of data collection and data type database implementations and upload tools and data analysis plans After definition of the requirements the common software architecture was designed The architectural design refers to the broad functions high level abstraction of the system General principles were adhered in the design and coding of the data analysis tool testing help and support software releases and software reuse At the end the base data analysis tools were implemented in each of the VMCs The tools have some common functionalities as well as some VMC specific functionalities Detailed descriptions of the data analysis tools developed in each VMC are provided in Section 4 2 4 2 Base of Data Analysis Tools Used in euroFOT Detailed descriptions of the data analysis tools developed in each VMC are provided in Sections 4 2 1 to 4 2 4 4 2 1 German 1 OC and French VMC French VMC and German 1 OC share a similar framework for data analysis software although implementation varies a little between the two VMCs The variations appear in 1 the implementation of the visualisation of the data since German 1 OC is not using the video display option and 2 the implementation of the enrichment scripts since the both VMCs have different Pls and set of events that need to be detected and characterised The aim of the data analysis software is to allow the definition and calculation of the intermediary results as
181. s quality check is used for intermittent sensors sensor that give no or unreliable data under certain conditions in our case GPS radar lane tracker and eye tracker For GPS it is checked that the receiver has satellite coverage at the time point the sample was taken and for radar and lane tracker compound quality measures are formed from the CAN signals containing information about the availability of these systems For the eye tracker signals the per sample check is performed in a slightly different way and no per sample quality vector is formed Instead a filtered version of the signal is created that contains all the values that can be considered to be of good quality This is done by processing the eye tracker signals trough a weighted median filter that takes into account quality signals from the eye trackers as well as constraints on eye movements 2 Per measure In the per measure quality check a quality value for the whole signal is created The per measure quality checks that are performed on the majority of the signals can be separated in to an enumerative signal check and an analog signal check depending on the type of the signal For all signals that go through the analog check it is checked that the signal is not constant and that the max and min values are within the specified signal range Additionally for some signals thresholds on sample to sample variations and percentage of time the signal can have a zero derivative are
182. saeeseeseeseesaeesaeeseeseeesueeseeeseeeseeeaaes 92 Experiment DUNOSN isa cedectecdteadendeheutendsdecthanceinctandtdead a iee n EAE iea A EEEE ineine 93 Dependency ADS en en NEE EERE AAE 94 Subversion Changelog 2 tetes sets 95 Branch Usage of Version Management VV 95 Sample Report for a reported Bug 96 Sowar Gnor DUO Re ae pen nes Reno ben eee co cu 96 New feature change request 97 MOOV RATS PPT PRE PR a PAR PE 97 Screenshot from a two screen possible configuration of FOTware 99 Annotation workstation with FOTware c ccccseccsecceeeceeeseeeseeeceeeeeeeseeeeeeeeaeees 100 ala DANS a a un did 100 A sketch of a video display module when the forward camera view is chosen 101 A depiction of the SQL module after successfully executed an SQL statement 102 A sketch of the form the user should fill in with his her database credentials 102 The SQL editor WINdOW cccccccceccccccececcccceccccecececcccecuaesececscseauavauaesecstaeavaeaess 102 THG Tile browser MOUSE a anses doa des de nee oct ac a den 103 lt Version 1 0 gt Vil Data management in euroFOT Figure 75 Figure 76 Figure 77 Figure 78 Figure 79 Figure 80 Figure 81 Figure 82 Figure 83 euroFOT 01 08 2011 The signal browser showing the preview of the first signal of the table 104 Three signals plotted in the signal plotter module 104 Statistical pro
183. ssible to modify some properties of the signal graphs by changing certain values in the table located below the axes M Plotter 1 Signal viewer FOTware ota g File Edit wiew Insert Tools Desktop Window Help OS ASIR Measures see table x oli IE iF iy 0 20 0 5 Signal Source Color Style Width Marker Face Size Y scale Delete Type l vehicleSpeed CAN ca Ea 2 Ej none none 6 ca 1 Fi plot 2 longitudinalAcceleration CAN ml slt me onele none 6 me 3 pot 3 brakePedalPosition CAN wl ea el nonel none 6 me 1 plot Figure 76 Three signals plotted in the signal plotter module 7 Data properties monitor This module shows using a table the instantaneous values and statistical properties of currently selected signals from the signal browser Figure 77 Deliverable D3 3 lt Version 1 0 gt 104 Data management in euroFOT np timestamps om ltitude GPS omAmbientTemp omDaylight amFix Time SPS omaFSTime _Quality omGPS_ Quality Value F 3425e 05 63 6500 17 2500 1 Mah Mah 1 Mean T 34256 05 21 8452 12 9822 1 Mahl Mah euroFOT Std 3 6057e6 07 23 1677 1 29272 0 Mah Mah Min r 34256 05 23 1000 11 4700 1 3 1 709e 70 0 D Max T 34256 05 F3 7 00 18 0300 1 3 1711e 10 0 1 Median F 3425e 05 16 3100 12 5000 1 Mah Mah 1 01 08 2011 Range 0 96 8700 6 5600 0 2026000 0 1 Figure 77 Statistical properties and instantaneous v
184. sting vehicle External two axis information via information via accelerometer with CAN CAN additional yaw sensor Eye tracker One camera One camera system system Audio functionality Driver route identification interaction box for audio comments and route identification and microphone External Navigation Added for Added for device certain period certain period Data transfer Wireless CAN Wireless CAN data Local Network Wireless status Wireless Wireless data transfer information disk status status pick up information information CAN video data disk pick up disk pick up CAN video CAN video data data Encryption Yes Yes Yes Yes Compression Yes Yes Yes Yes Yes Deliverable D3 3 lt Version 1 0 gt 11 Data management in euroFOT euroFOT 01 08 2011 The task to decide which DAS solution s to use and make the DAS ready in each VMC was done at VMC level However as much as possible the work to gather technical requirements specification to identify possible DAS solutions and to conduct market survey and evaluation were done together Here we give some examples of those joint tasks Collaborations between VMCs Generally several specialised focused groups were formed to enable joint efforts in performing certain tasks for example for investigating some sensors like eye trackers radar cameras and map providers The possibilities to use the same eye tracker cameras and othe
185. t Version 1 0 gt 44 Data management in euroFOT q ET M 42 Anwendungen Orte System 2 Euro FOT Status Seite Mozilla Firefox Datei Bearbeiten Ansicht Chronik Lesezeichen Extras Hilfe euroFOT 01 08 2011 d Mo 12 Jul 15 43 pva 0 amp A Ie http localhost amp v rl gt 6 Euro FOT Status Seite 6 Euro FOT Status Seite da l6 monit service man phpMyAdmin e Wireless Gateway L Mercedes Benz Lechoology Euro FOT Status bersicht der Versuchsfahrzeuge S EF 404 IP 10 1 1 4 D1 offline zuletzt online 11 07 2010 00 20 09 Kilometerstand 18681 8 Km interne Festplatte 21 42 GB frei externe Festplatte 423 17 GB frei Position MBtech PSS ae S EF 407 IP 10 1 1 9 D6 offline zuletzt online 12 07 2010 12 13 58 Kilometerstand 24788 4 Km interne Festplatte 21 48 GB frei externe Festplatte 387 52 GB frei Position S EF 414 IP 10 1 1 14 D11 offline zuletzt online 12 07 2010 12 40 03 Kilometerstand 8575 Km interne Festplatte 21 45 GB frei externe Festplatte 5 68 GB frei Position S EF 402 IP 10 1 1 6 D2 offline zuletzt online Kilometerstand Km interne Festplatte externe Festplatte Position S EF 403 IP 10 1 1 5 D3 offline zuletzt online 10 07 2010 13 47 13 Kilometerstand 26579 1 Km interne Festplatte 4 49 GB frei externe Festplatte 429 49 GB frei Position
186. ta 1 10Hz 2 SmartEye oMeasurelnfo oEventData N number of segments could be trips or any other segment of data oEventinfo voHistory Figure 44 Overview of MATLAB struct oDBdata mat oDBdata The main level of the struct with two indices One 1 for all 10Hz data and 2 for all Eye Tracker data This is the struct that the pre processing produces as output and what the uploader takes as input OSQLInfo Since this is also the format returned when querying the database for analysis the information and history about how the data was retrieved and who retrieved are stored here This is not relevant for the pre processing and upload oSegmentinfo This is called Segmentinfo since it is also used for analysis but for pre processing and upload it is the trip information Here all the information about e g trip_ID driver_ID trip start stop OEM name as well as trip quality information and references to video files are stored The segment info can when returned as a query be a struct array but for pre processing and upload this is always only one instance corresponding to the trip voMeasureData A struct array with one instance per segment That is when data is retrieved for analysis data is extracted as segments For pre processing the number of segments will always be one 1 and corresponds to the whole trip This is where all measure data end up including derived measures There will always be a field called tim
187. te View Bad LIT or ef band Horizontai niyen Write Annotations Horizontal alignment RHT Straight Horizontal alignment LHT Straight Vertical alignment Level Number of lanes Locality Urban Relation to junction Non junction Weather Fog Light conditions Daylight Road type Walking pace street Traffic Control No traffic control Road surface condition Dry Traffic Density Freeflow ts Event severity Normal driving Crash deformation extent Tica sno resid Ronda Union MAIS AIS 1 Initial Conflict Type Conflict with a lead vehicle Secondary Conflict Type Confict with a lead vehicle Near Crash type Rear end striking Conflict partner type None Conflict partner location RHT In front of subject Conflict partner location LHT In front of subject Driving Lane RHT Rightmost lane Driving Lane LHT Leftmost lane Heading Within Lane Vehicle aligned with lane Narrative Edit here Pre event manouvre RHT Going straight constant speed Pre event manouvre LHT Going straight constant speed Precipitating event RHT This Vehicle Lost Control Blow out or fl Precipitating event LHT This Vehicle Lost Control Blow out or fi Start of precipitating event CRE start Edit TIME here Evasive manoeuvre RHT No driver present Evasive manoeuvre LHT No driver present Post manoeuvre control RHT Control maintained Post manoeuvre control LHT Control maintained
188. tecture v1 0 RTC amp RAM supply Trigger 32 bit RISC Microcontroller Serial port 12V 24V HW architecture v2 0 system Power supply CAN processor Application processor Serial RTC amp port Battery Figure 7 HW Architecture for CTAG Datalogger v1 0 and v2 0 HW upgrade for Datalogger v2 0 CTAG Datalogger Il to fulfil euroFOT requirements The following modifications had to be implemented in the CTAG Datalogger in order to meet the new HW requirements for comparison see Figure 7 e A new architecture was implemented with 2 microcontrollers CAN microcontroller extending up to 4 CAN channels and Application microcontroller e GPRS and GPS connexions XT65 Siemens The Datalogger v1 0 did not have any wireless external communication module neither GPS receiver Deliverable D3 3 lt Version 1 0 gt 16 Data management in euroFOT euroFOT 01 08 2011 e SD and SDHC compatible In version 1 0 there was a limit of 2 GB in the internal storage capability With the new 2 0 version SDHC cards can be used boosting the off line performance of the device e USB 2 0 High Speed 480 Mbps e New CAN and supply connectors but still not automotive e Configurable LED indicators e New automotive compliant connector for CAN and power supply A new automo
189. ted that shall be displayed in one of two diagrams Figure 82 displayed on the right hand side of the video einstellungen_Diagramm BE Diagramm 1 Diagramm 2 w 4 a fa ka Hinzuf gen gt Status Variablen lt Entfernen CarFollow CarFolow O E lt Entfernen OK Abbrechen Figure 81 Selection of variables to be displayed together with video The GUI allows viewing the video while the selected variables are displayed over time The length of the time window used for depicting the selected variables as well as the speed with which the video is displayed can be chosen Sprungprogramm Ms E x Laden Speichern Einstellungen Extras 00508 98 100 88 25 47 80 60 DR N N 40 eo 20 CHAH HEIMI CHAHHEL2 1 i 1 aoe ELDELE CHARA EES PE 0 45 50 55 60 65 70 75 3 v 1 1 0 8 0 6 0 4 Time 0807 2010 13 1212 30 Gi DIE spa ms 0 2 Esher hr a n chste Fahrt Marker Mo 0 ca a ae 45 50 56 60 65 70 75 Suche CarFollow CarF ollow entspricht von 1 z it Yariabl a is ZeitBereich eit Yariablen bis ES EN F lt Variable gt gt Suchen 30 Figure 82 Display of video together with selected time series data Furthermore it is possible to jump to defined events based on a selected variable Figure 83 This enables the experimenter to systematically control for instance event detection algorithms based on the video Delive
190. ted trips and testing data model or algorithms evolution Deliverable D3 3 Data management in euroFOT lt Version 1 0 gt 88 euroFOT 01 08 2011 Visualisation and annotation tool This tool allows simultaneous visualisation of all trip linked data Data can be browsed and visualised using the GUI called Trip Viewer shown in Figure 54 The trip viewer is the main module that controls all functionalities The user can alter between the tabs in the panel on the left of the window Each tab represents a separate functionality Signals Events amp Situations and Trip information In the first mode the user can browse through the trip signals which are listed in a table see the left part of Figure 54 With the check boxes in front of each signal the user selects the signals that should be displayed as a time history on time graph see the right side of Figure 54 The user can choose the style graphical properties of the line for each of the signals on the plot The values of the signals in the current time shown as red line on the time history graph on Figure 54 can also be displayed in the module nspector Figure 55 Panel tab selector Time history view BB Trip Viewer 23103 06 Jul 2011 13 44 45 Driver 2 Phase 2 Vehicle 3 os mee File Help S VE OF Signals Events amp Situations Trip Information 400 Ill Il CC SL_State Plot Name U Gain
191. ted via GPRS each certain time configurable It could also be transmitted after a complete trip cycle thus preventing overwriting If the memory card is full and it is not possible to transmit data the Datalogger stops logging and saves the last data It does not overwrite data e Real Time Clock synchronized with GPS time The Datalogger timestamps and synchronizes frames from video logger if required The timestamp resolution is around 1 ms in CAN data The maximum latency of a single CAN message is around 40 microseconds Sample rates for CAN data are configurable from 0 to 2 ms e Storage data format is open non proprietary for euroFOT purposes e Output data format is asc compatible with most commonly used CAN analysis network tools The Datalogger provides code examples allowing data access The Datalogger has several standard servers http rsync ssh allowing data uploading real time monitoring and remote configuration e The Datalogger is capable to log up to 4 CAN channels including extended CAN The Datalogger is able to send out CAN frames by phyton scripting and to perform Diagnosis functionality also by programming specific scripts in Phyton e Other functionalities such as configurable LED indicators SDHC capability USB interface or remote configuration firmware upgrade have also been implemented and successfully tested 2 2 1 4 Tests and quality check performed Table 5 The different tests performed on CTAG D
192. tems used in the different VMCs 10 Table 4 Result of market survey on CAN loggers cccceseeeeceeeeeeeeeeseeeeeeeeeeseeeeeeneenseeeess 14 Table 5 The different tests performed on CTAG Datalogger Il 20 Table 6 The part list used for the two levels of instrumentations used in the French VMC 24 Table 7 The different tests performed on videologging of French VMC cccsseeeeseeeees 29 Table 8 The components used in the SAFER euroFOT DAS 33 Table 9 The different tests performed on the SAFER euroFOT DAS 37 Table 10 Part list for the DAS used by BMW 41 Table 11 The database and data storage solutions used in the different VMCs 48 Table 12 Data storage at German 1 OC and French VMC 51 Table 13 Estimated size of time history data in the Swedish VMC 60 Table 14 Estimated size of video data in the Swedish VMO ccccseeeeeeeeeeeeeeeeeeeeseeeeess 61 Deliverable D3 3 lt Version10 gt yo Data management in euroFOT euroFOT 01 08 2011 List of Figures Figure 1 SP3 objectives and work packages VV 2 Figure 2 Data management chain considered in euroFOT SP3 ccccceeeeeeeeeeeeeeeeneeeesees 2 Figure 3 FOT chain introduced in FESTA SP3 role is marked with red Curve 0 c0008 3 Figure 4 Collaboration between SP3 and other subprojects SPs in euroFOT 0008 4 Figure 5 The VMCs in euroFOT and the FOT plan in July 2009 5 Figure
193. th mentioning again Different quality threshold might be needed for different types of analysis this applies to signals measures and video and or other sensors requirements One cannot really exhaustively list in advance all the possible analysis that could be done Other reflection 5 Non project members often ask why not use common solutions for all VMCs While using common solution s seems ideal it was inevitable to finally choose different solutions as the requirements from the different VMCs in fact from different partner organisation are different Often the solution has to fit partners own standard existing IT infrastructure and expertise In the spirit of research having different solutions is in fact very good as it gives us an opportunity to assess the suitability of the different solutions for FOT 5 1 Data Acquisition Systems The following works well 6 Having specialised focused groups to enable joint efforts in performing certain tasks for example investigating some specific sensors like eye tracker radar cameras and map providers were found useful The following could have been done better 7 Unavailability of suitable DAS fulfilling euroFOT requirements in the market particularly for a large scale FOT has left us with no other solution than to partly develop our own solution in particular for the Swedish VMC It would have been convenient to build a complete test rig for emulating the environm
194. tion Relation to junction Driving Lane RHT Moving vehicle with or without load Lost in thought Looked but did not see Parked vehicle Non junction Righ most lane Post manoeuvre control RHT External distraction Weather E Control maintained Driver attention allocation Looking at traffic other than critical event Fog Driving Lane LHT Attention towards the critical event j nS Personal iene Nome Current Light conditions a Post manoeuvre control LHT hyg o f Daylight Combing brushing fixing hair Heading Within Lane Control maintained Road type Vehicle aligned with lane Gaze direction RHT Navigation Walking pace street Forward Side View Start of evasive maneuver Looking at map Viewing naviga Interacting with Traffic Control Narrative Horizontal abn ne Motel conio Edit here Edit TIME here Driver activity RHT Road surface condition Gaze direction LHT Checking centre rear view mirror Dry End of evasive maneuver F d Write Annotations ry Driver activity LHT Traffic Density Edit TIME here Checking centre rear view mirror Free flow FOL OTT AE TA a CR 1 T Figure 79 A possible configuration of the annotation module inspired by the 100 car study Unique and novel features of FOTware include modularity and customisability as well as the programming approach and the definition of events to communicate across modules FOTware implements the required functionalities listed in Chapter 3 Further new analysis needs may raise when the
195. tion and annotation tool with signals selection panel Deliverable D3 3 lt Version 1 0 gt 89 Data management in euroFOT GPS and map data euroFOT E Inspector Timestamp 1572 7 Z 06 Jul 2011 14 10 58 CC SL_State C SL_ Speed CC SL_Overspeed Distance Totalizer Vehicle Speed Clutch Switch _ Travel Brake _Info_Status Throttle Steering wheel Angle Engin _RPM R_Range Distance ToLettLaneMark DistanceToRighttLaneMark Map MeLink hiap SpeedLimit hiap_islrban Map_MumofLanes Map RoadType Map_Roadtlumber Map _FRoadtame Low Bean High _ Bear External_Temp CC_ active 92 kmh no_over_ speed 21049 km 69 3 km Pedal_not_pressed Brake pedal _not_pressed 0 Arbitrary 2 3 deg 1714 Trimin new link 90 kmh Mot Urban 1 lane FCO or FCA NAO null Light_oft Light_oft 20 deg _c 01 08 2011 Figure 55 Inspector showing the current values of the selected signals The user changes the current time by dragging the red cursor entering a value in the inspector or selecting an event or situation in which case current time is set to the beginning of the instance LA Trip Viewer 23103 06 Jul 2011 13 44 45 Driver 2 Phase 2 Vehicle 3 ROC X r File Help Ay Signals Events amp Situations f Trip Information y Signals Events amp Situations Trip Information Sr A Sonate LO Tip 0 6 88 VE OF Display Name Icon Color
196. tive connector was integrated in order to fulfil the high level OEM requirements In order to guarantee suitable reliability a standard automotive connector was required For this reason the new version of the device is equipped with a micro quadlock TYCO connector shown in Figure 8 Micro Quadlock TYCO connector for o 4CAN channels o Power supply and ignition signal KL15 o Trigger e Two power in versions are available 12V and 24V e Consumption reduction KL15 mode The high energy consumption was solved by using the ignition signal KL15 for the power supply i e the data logger will be supplied with energy only when the engine is on In the current version of the Datalogger consumption in sleeping mode or idle mode to be woken by CAN activity is under 30mA There is another possible supply mode connected to KL15 not directly to Bat via VBAT2 where consumption in ignition signal mode when ignition is off is OMA approximately 0 1 HA Figure 8 Tyco Connector SW Upgrade for Datalogger v2 0 CTAG Datalogger Il to fulfil euroFOT requirements e CAN processor Real time operating system with enhanced CAN driver e New Application processor based on Linux operating system with applications for CAN amp GPS logging e GPRS connection to internet that allows real time data synchronization and real time monitoring tracking with CTAG Server tool Deliverable D3 3 lt Version 1
197. tures in acquired files as Raw data amp all derived files converted in Harmonized and CAN amp Framework s data in manually GPS Trip Frameworks coded Structure Trip Structure artefacts MATLAB MATLAB derived Format Format measures events situations Database Meta data in N A N A N A N A Processed Storage relational model data in relational model suitable for querying and analysis Signals storage in DB is optional Backup As files As files As files As DB In file storage Trip related data signals and derived artefacts is stored as MATLAB files following a structure which is relevant for both Raw data and Pre processed or Processed data The Framework provides all sort of tools to manipulate or display this data The framework also ensures permanent consistency between file storage database storage and compliance with Trip Data Model To do so it integrates a complete database management tool which creates and maintains automatically the database using the Trip Data Model This flexible approach even allows the framework to work with or without database server A feature allows the creation of a subset of pre processed data based on any filtering criteria with which it is possible to efficiently prototype and test further evolution of the Trip Data Model and calculations before pushing the evolutions in the production environment Storage engines and server solutions chosen vary the French VMC has decid
198. ucture used in the Swedish VMC 66 Part 4 of the database structure used in the Swedish VMC 67 Part 5 of the database structure used in the Swedish VMC 68 Part 6 of the database structure used in the Swedish VMC 69 Overview of MATLAB struct oDBdata mat ccccccceeceeeeeeeeeeseeeeseeeseeeeseeesaeeens 70 The pre processing steps in the Swedish VMC sansonnsnnnsnnnnnnnnnonnnnnnnnrrnnrnnrnennne 72 The uploading steps in the Swedish VMC cccccecccseeeeceeeeseeeeeseeeeseeeseaeeesees 73 Overview of the process of complementing logged data with map attributes 74 General layout of German 2 OC database 80 Adapted folder structure required for CAN data stored as mat files 81 Folder structure used for storing the driving data 81 Overview of data pre processing enrichment and upload routines 66 82 TADS EXPO ar 88 Log visualization integrated in Trips Explorer 88 Trip Viewer Visualisation and annotation tool with signals selection panel 89 Inspector showing the current values of the selected signals 90 Video display four camera views Forward Driver Steering Column and Feet 91 Vehicle s position on a Map an eue aa ao nue an asso Dean na tan 92 Event module cc cece cecccecececece cece cece cece eesaee
199. uous and direct communication in between the users and the programmers to eliminate bugs and increase overall usability of the tool At the present time the tool can only analyse one trip event at a time and can provide parameters to be used for Statistical analysis however FOTware enables only descriptive statistical analysis of individual signals from the CAN bus or eye tracking In the future this tool will be also used for outlier analysis and hypothesis testing The extent to which new functionalities will need to be added for these new types of analysis is still uncertain at the present moment In any case Statistical analysis is envisioned to be performed on commercial programs already developed for this purpose such as SPSS 4 2 2 3 Lessons learned The general experiences from FOTware are very positive and previous Swedish FOTs have shown the benefit of using a modular tool It is also a large advantage to use MATLAB where the majority of analyses are carried out anyway The scripting language provided through MATLAB allows several users to develop modules and features independently Changes can automatically be incorporated through the use of version control system and no compilation or restart of the software is required Deliverable D3 3 lt Version 1 0 gt 107 Data management in euroFOT euroFOT 01 08 2011 Difficulties and Solutions Initially there were concerns of the performance that could be reached in MATLAB particularly
200. urces may provide more valuable insights about activity when viewed and analysed together Visualisation tools are needed to address the challenge of analysing and synthesising information from multiple data sources and at multiple time scales Sharing the content of the user findings and communicate it is another fundamental issue in the analysis tools Although several other systems have been developed to support various aspects of this visualisation challenge data analysis software represents a unique focus on the visualisation of multiple diverse data sources For example numerous systems exist for coding and annotation of video such as VCode Hagedorn Hailpern amp Karahalios 2008 Diver Pea Mills Rosen Dauber Effelsberg amp Hoffert 2004 ELAN Wittenburg Brugman Russel Klassmann amp Sloetjes 2006 and The Observer XT Zimmerman Bolhuis Willemsen Meyer amp Noldus 2009 While these systems have powerful coding capabilities they are not designed for analysis or visualisation of heterogeneous data nor do they support easily extensible automatic analysis facilities They are also commercial products with limited open extensibility to visualise new types of data The main contribution from WP3500 is data analysis software that allows researchers to visualise time based data from multiple sources and manually or automatically annotate the data The main design goals that motivated the work in this package are 1 Th
201. use the vehicles used in the previous projects were all classified as test vehicles Therefore non EMC classified equipment could be installed When compared to the previous SAFER owned data acquisition systems this new system has a different hardware platform but was built using some of the designs ideas and software pieces from the previous SAFER owned data acquisition systems 2 2 3 3 Final implementation to fulfil euroFOT requirements The parts that are used to make up the DAS used in the Swedish VMC SAFER euroFOT DAS are listed in Table 8 Figure 17 shows the placement of the DAS in the spare wheel well of a Volvo XC70 car and the placement of the cameras Table 8 The components used in the SAFER euroFOT DAS Item Comment PC Nexcom VTC6100 Intel Atom 1 6 GHz The Nexcom VTC6100 VTC6100 2010 is a rugged PC developed for automotive applications It has regulated 5V and 12V outputs 1A each This made it possible to power all equipment without Deliverable D3 3 lt Version 1 0 gt ao Data management in euroFOT Item GPRS 3G modem GPS receiver Video grabber Compact Flash Hard disk drive CAN interface 2 BLIS interface 2 J1587 to CAN interface Accelerometer Bosch Relay Cameras 4 IR illuminators 2 EMI filter fuse Switches 2 Combined GPRS 3G and GPS antenna Eye tracker Cable harnesses Mounting brackets Deliverable D3 3 Data management in euroFOT euroFOT 01 08 2011 Com
202. w minutes of inactivity By not having the HDD up and running all the time it is believed that the HDDs will last longer Inbuilt diagnostics It was early decided to add as many diagnostic functions as possible to the DAS software This turned out to be helpful both during early tests and during execution of the FOT both for detecting problems with cable harnesses and for detecting failures within the different components of the DAS For each driving cycle ignition on to ignition off a status report XML is generated in the DAS which is then uploaded wirelessly to a server at the beginning of the next driving cycle see description of the web based monitoring tool in Section 2 2 3 4 The following errors can be detected e Missing video signal e g loose video cable or defect hardware e Obscured camera e Low CAN busload settable threshold e CAN error frames e CAN bus hardware buffer overflow e Decreased HDD performance see HDD diagnostics above e Missing HDD e g loose USB cable or defect hardware e Missing CAN interface e g loose USB cable or defect hardware e SW errors Furthermore the status report contains information about e HDD temperature e GPS signal strength e GPRS signal strength e Histograms over signal values configurable e Average signal values configurable e Remaining storage space on Compact Flash and HDD e Remaining money on SIM card e Version number of all software in the DAS Silent
203. w O3 Jub2011 14 43 41 00 29 22 km Figure 52 Trips Explorer a Ra s BN 5 5 Fi Fi F9 RI Fi Bo bo Bo Boo Bo mo Bo Boo Boo Po Po Fi Fi Fi Fi Fi Fi ha O0 0 0 O1 0 O O1 O1 00000 oa The Trips Explorer also integrates access to trip related data either actual trip data in Trip Data Viewer or processing logs in a dedicated visualisation tool pictured below gt w Signal LeftLaneConfidence has not been assigned for this configuration gt w Signal RightLaneConfidence has not been assigned for this configuration Calculating delay between GPS and CAN Data gt w GPS synchronization seems to fail Needs checking Processing signals Synchronization finished done done 0 Processing Experiment Eurof OT gt 0 Execution of node N_1 Process CleanUpRadarRange done gt 0 Execution of node N_ 2 Process AssignNetworkType done 0 Execution of node N 3 Process AssignExperimentalPhase gt done gt 0 Execution of node N_ 6 Process ClassifySpeedLimit gt done gt 0 Execution of node N_ 50 Event CCSLSpeedChange i Found 0 events to be refreshed done Figure 53 Log visualization integrated in Trips Explorer Finally when the Framework is used in prototyping mode the Trips Explorer allows running the processing on a set of selec
204. we create the tables columns in such a way that data which are supposed to be huge for instance events and their attributes are stored in single tables Thus we prepare the data in compatible form and use database specific techniques like sql loader to perform fast and reliable uploading Also the above separation gives us the ability to create different types of users which will have different rights to database manipulation The structural characteristics of the database are defined by SQL statements that comprise the Data Definition Language DDL create alter drop If the above statements fail during the execution we cannot rollback the database to a previous consistent state So we create another protection layer by giving the ability to the user to make an informative inquiry of the upcoming changes to the database during meta data process The overhead and the speed bottleneck created by foreign keys are overcome by transferring the control from the database to our library for huge tables The users also have the ability to remotely send upload the data to a specified database simplifying a lot the data upload process Finally the structure of the database gives the analysts the ability to define filters and directly access the data corresponding to the entire dataset i e data from all the trips which are relevant and necessary to test each hypothesis Performance With respect to quality Data quality on sample and signal basis is perfor
205. well as their storage in the database in the most interactive way The software allows both visualisation of data trip visualisation and annotation tool and modification of the data model definition of new Pls or events linking of new scripts with trip data model front end This software is mainly programmed in MATLAB Four processes are taking place in the workflow of the data e Pre processing and enrichment Raw trips are converted to a standard readable format MATLAB file initially only containing unsynchronised time history signals The signals are then synchronised and harmonised according to a uniform data model The trips are then enriched with aggregated data by applying user defined scripts e Upload to database Trips data from the MATLAB files is then stored in relational database e Analysis of data The analysis is performed by directly querying aggregated data and its description from the database using standard statistical tools such as SAS Statistical Analysis Software SAS STAT 2011 Users manipulate trip data only by e Defining new objects in the data model e Defining algorithms in scripts e Directly manipulating trip related data in trip data viewer GUI explained in the following section Deliverable D3 3 lt Version 1 0 gt 86 Data management in euroFOT euroFOT 01 08 2011 4 2 1 1 Implementation The GUI consists of three main parts 1 The Trips explorer a search and listing tool 2
206. white one Deliverable D3 3 lt Version 1 0 gt 29 Data management in euroFOT euroFOT 01 08 2011 2 2 3 Swedish VMC 2 2 3 1 Initial condition Loggers CAN Only CAN Extra CAN Video Vehicles sensor s Extra sensor s Volvo Cars 100 Volvo Trucks 30 The partners in the euroFOT Swedish VMC are Chalmers Volvo and VCC Within euroFOT data management tasks in the Swedish VMC were especially challenging due to the complexity of the data acquisition systems and the number of safety functions tested Furthermore fleet handling was also comparatively arduous because 15 of the participating Volvo trucks are based in the UK and 15 in The Netherlands many of these trucks often be driven outside their home country Thus special management and maintenance strategies were needed All of the participating VCC cars are based in Sweden While waiting for the list of measures from SP4 SP3 partners inside Swedish VMC started gathering requirements based on partners experiences from previous smaller FOT studies run in Sweden The result from this step was a preliminary requirements specification that was used as a basis for request for information RFI This RFI was used to identify companies that are interested in participating in the project as suppliers of equipment and or services for the data acquisition system At the same time when preparing the RFI a survey was conducted to get an initial list of potential suppliers The RFI w
207. x v x x v Driver Driver 2 Driver Driver 1 Driver 2 Driver Driver 1 Driver 2 Driver Driver 1 Driver 2 Driver 1 Driver Driver 2 Driver Driver Driver Driver 1 2 va 1 RE Drive X CAN Video data For both OEMs data logging does not only include CAN but also video data The formats used for video differ between the OEMs Figure 50 Folder structure used for storing the driving data BMW saves the video data as single frames in jpg format The naming of the frames allows relating each frame to a certain data point in the driving data For Daimler the logged data can be converted either to single frames or to videos It has been decided that using video format is handier and therefore the data is stored as videos Daimler converts the video data into videos in mjpeg format For longer drives the whole video is stored in several files with a length of about 50 minutes each At IZVW the different Deliverable D3 3 lt Version 1 0 gt 81 Data management in euroFOT euroFOT 01 08 2011 video sections for one trip are merged into one mjpeg video which is stored in a Matroska container Matroska Media Container Homepage 2011 3 2 3 2 Upload and enrichment tools Description of data The variables of the logged CAN data used for euro FOT as well as all derived and added parameters are described in det

Download Pdf Manuals

Related Search

Related Contents

Open the Instruction Manual  Panasonic BT-LH900E Flat Panel Television User Manual  Adendo No.1 - Hospital Universitario Erasmo Meoz  User Manual  3M Quest Edge Intrinsically Safe Cable-Free Noise  配布資料 - 日本弁護士連合会  Kerosene Portable Heater - Mi-TM Equipment Sales & Service  - BA  Kramer Electronics C-HDMI/DVI-10  EUROLITE NH-30 Hazer User Manual (#3046)  

Copyright © All rights reserved. Failed to retrieve file