Home

Limited Deployment Pilot Project: Monitoring Truck Traffic in

Contents

1. Totals sv LV Average 3 2 72 Stdev 2 1 24 Station 098 EB Evaluation 2 Hourly Errors Upper Bound 5 19 10 Totals SV LV Average 3 2 21 Stdev 1 1 14 45 6 3 2 3 THIRD ROUND OF AGGREGATED DATA COMPARISON ONE DAY RESULTS The third evaluation period reflects further adjustments to the TMS length parameter The following tables show the lower bound and upper bound errors of the aggregated data over a one day period for each evaluation site either May 26 27 or 28 2010 Once again the period of analysis was from 6 00 a m to 9 00 p m The errors for stations 037 Table 25 and 040 Table 26 did not differ significantly However Station 098 Table 27 and Table 28 The attached TMS 100 User Guide provides information on setting up the units and modifying the parameters Table 25 Station 037 May 26 2010 Station 037 NB Evaluation 3 Hourly Errors Lower Bound 5 26 10 Total sv LV Average 1 1 0 8 4 3 Stdev 0 4 1 0 10 0 Station 037 NB Evaluation 3 Hourly Errors Upper Bound 5 26 10 Total SV LV Average 0 5 0 8 2 1 Stdev 0 4 1 0 8 4 Table 26 Station 098 May 26 2010 Station 098 EB Evaluation 3 Hourly Errors Lower Bound 5 26 10 Total SV LV Average 5 3 4 9 9 1 Stdev 2 2 2 3 6 0 Station 098 EB Evaluation 3 Hourly Errors Upper Bound 5 26 10 Total sv LV Average 6 0 4 9 20 0 Stdev 1 6 2 3 12 0 46
2. After the project began InfoTek was acquired by CASE Global Technologies CGT which changed the name of the software and equipment to Traffic Monitoring System TMS This report uses the current name of TMS e Remote access e Low cost maintenance By expanding the network of truck counting sensors the data collected will e Allow increased tracking of truck traffic patterns including seasonal and holiday traffic e Inform better planning of ramp metering and congestion management e Provide detailed statistics on goods movement in the monitored area e Give information on pavement use for highway maintenance planning The efficacy of using the TMS 100 InfoTek Wizard solution was confirmed in a small scale 2006 demonstration project as part of the Innovative Corridors Initiative public private partnership program The pilot project which tested three sites each in Districts 4 and 7 showed promising results The less costly TMS 100 solution performed with at least 95 accuracy compared to the AVC system The next step prior to full commercial deployment was to test and evaluate the TMS units on a larger scale along two of the most heavily traveled truck corridors in the Los Angeles area The initial proposal for this project was to deploy the TMS equipment and software at 26 single loop sites to track where trucks are going after leaving the ports of Los Angeles and Long Beach District 7 proposed 10 locations on Interstate 710 leadi
3. www irdinc com library pdf double threshold pdf 1993 United States Department of Transportation Federal Highway Administration Field Test of Monitoring of Urban Vehicle Operations Using Non intrusive Technologies FHWA PL 97 018 May 1997 Victoria I C and Walton M Freight data needs at the metropolitan level and the sustainability of intelligent transportation systems in supplying MPOs with the needed freight data December 2004 58 Wei C et al Vehicle Classification Using Advanced Technologies Transportation research record 1551 November 1996 pp 45 50 59 1 ACCIDENT PICTURES Pictures of the accident that occurred on March 10 2010 Two Caltrans employees were seriously injured 60 61 62 2 LOCATIONS OF INSTALLED TMS 100 UNITS Station County Route Post Mile Description Cabinet Location 464 LA 60 0 281 Westbound at Boyle Ave WB on from Soto St 464 LA 60 0 281 Eastbound at Boyle Ave EB on from Soto St 98 LA 60 R2 22 Eastbound east of Indiana St EB on from Indiana St 98 LA 60 R 2 22 Westbound east of Indiana St WB on from 3rd St 99 LA 60 R3 88 East of JCT RTE 710 WB shoulder at Belvedere Park POC 499 LA 60 R8 2 East of Paramount Blvd EB shoulder 100 LA 60 R 10 6 West of Peck Rd WB shoulder 607 LA 60 R 25 464 West of JCT RTE 57 North WB shoulder 120 LA 60 R 26 57 Westbound east of JCT RTE 57 WB on from
4. 6 2 AVG Validation Resul sia id 35 6 3 TMS Evaluation Res Sa d 38 6 3 1 TMS PVR Data Comparison Results sese eee ee eee 38 6 3 2 TMS Aggregated Data Comparison Results see eee eee 40 LET ONS ee 56 8 ROTTEN COSA Je eee 58 Attachment 1 Accident PU Saver AAA AAA 61 Attachment 2 Locations of Installed TMS 100 Units sss seer eee 64 Attachment 3 Schematic of Data Collection Set Up see e eee eee eee 65 Attachment 4 TMS 100 Handbook sss sse AA 67 1 INTRODUCTION AND PROBLEM STATEMENT A large number of trucks travel to and from Los Angeles area ports contributing significantly to traffic congestion Large trucks also cause the majority of damage and stress to roadway surfaces and bridges Because truck transportation impacts every aspect of traffic and infrastructure in the Los Angeles area it must be monitored and managed effectively To address this issue it is important to have accurate real time census data on the number of trucks on the roadways peak travel times lane use and where trucks travel on the highway system after leaving the port areas Currently Caltrans relies on data collected from Automated Vehicle Classification AVC stations and manual census counts at specific locations The majority of the data collection and analysis is based on professional experience and has not been standardized or documented In addition collecting traffic data has been limited because of cost Caltrans s existing technology for auto
5. Truth LV 16 0 15 1 1 0 System Accuracy 95 LV Hit Rate 93 LV Miss Rate 6 LV False Alarm 0 An undercount was observed at station 098 lane 4 Table 12 which was caused by the class 15 errors counted at this station Comparing the video data to the PVR data revealed many instances in which the AVC system missed passing SVs after recording a class 15 error It was suspected that the system was recovering for a brief moment after reading a class 15 error and thus missed several vehicles 36 Table 12 AVC validation data Station 098 eastbound lane 4 Station 098 EB Lane 4 sv LV Errors Class 15 Cross talk Missing Ground SV 620 553 0 67 10 19 38 Truth LV 54 0 50 4 4 0 0 System Accuracy 89 LV Hit Rate 93 LV Miss Rate 1 LV False Alarm 0 6 3 TMS EVALUATION RESULTS Assessing the TMS 100 consisted of performing a one to one comparison of the PVR data to the video data and then comparing the one hour bin aggregated TMS data to the AVC aggregated data for the long term evaluation The plan was to collect PVR data collection first However because of the accident see 1 the team could only collect video and PVR data from station 098 The long term evaluation phase 2 comparing the TMS system to the AVC data continued but at a decelerated pace and with several modifications because of time constraints While comparing the TMS PVR data with the video the results
6. 0 0 0 Time C1L1 C2L1 C1L2 C2L2 C1L3 C2L3 C1L4 C2L4 C1L5 C2L5 C1L6 C2L6 6 00 226 0 286 7 118 30 327 1 366 6 284 47 6 15 315 2 315 9 142 48 276 2 301 14 234 52 6 30 366 0 320 16 162 58 325 2 369 10 287 64 6 45 478 3 432 12 172 66 412 1 405 14 309 58 7 00 486 1 430 11 214 53 458 2 421 14 341 80 7 15 483 1 409 14 204 55 435 2 418 16 312 90 7 30 520 1 448 13 291 61 436 2 382 16 325 79 7 45 417 20 401 26 319 33 426 4 394 21 292 96 31 Table 7 Sample processed hourly aggregated AVC data Each column shows lane by lane vehicle classification where C1 is lt 40 feet and C2 is gt 40 feet Time C1L1 C2L1 C1L2 cC2L2 C1L3 C2L3 C1L4 C2L4 CB C2L5 C1L6 C2L6 6 00 673 1 899 32 382 118 789 2 1014 23 721 143 7 00 1645 6 1497 48 690 225 1471 7 1496 52 1171 254 8 00 1868 39 1680 92 1109 217 1849 10 1671 66 1238 368 9 00 1861 32 1575 132 895 357 2066 10 1545 139 1058 513 10 00 1804 4 1330 288 621 615 1655 8 1216 194 962 571 32 6 RESULTS The results from the AVC validation were positive and within the expected parameters described by Caltrans The results for evaluating the TMS 100 were compiled with limited data and a modified evaluation plan to account for the reduced staffing caused by the collision that occurred on the first day of the TMS evaluation 6 1 PERFORMANCE MEASURE PARAMETERS FOR VIDEO COMPARISON Various parameters were measured and analyzed while comparing the ground truth video data for those 30 minute video recordin
7. TMS 100 User Guide Single Loop Configuration Cable Mapping Cable TMS 100 Wire Detector Loop Output Chan1 8 Ground Grounded to detector Chan1 8 1 Lane 1 Chan1 8 2 Lane 2 Chan1 8 3 Lane 3 Chan1 8 4 Lane 4 Chan1 8 5 Lane 5 Chan1 8 6 Lane 6 Chan1 8 7 Lane 7 Chan1 8 8 Lane 8 Chan9 16 Ground Grounded to detector Chan9 16 1 Lane 9 Chan9 16 2 Lane 10 Chan9 16 3 Lane 11 Chan9 16 4 Lane 12 Chan9 16 5 Lane 13 Chan9 16 6 Lane 14 Figure 15 SINGLE LOOP CONFIGURATION CASE Global Technologies 21 TMS 100 User Guide Detector Output Diagram The following diagram shows the back of the detector rack labeling its wiring connections The diagram shows an input file with 2 channel detector card in the dual loop configuration the lead loop is the first detector output of the detector card and the second detector output is the trail loop of the detector In a single loop configuration the first output is lane 1 and the second out put is lane 2 lane 3 and lane 4 lane 5 and 6 etc INPUT FILE v LOTS SLOT SLOTS SLOTS SLOTS SLOTS e SE F w C o F ILDA Cable N 2 El 16 Pin connector K GECCOEDEL 1 ROEM CGECCOERE S TSTS TS gt aG GERE La HGL ERE CGECCOERE TiS e OD ob UGH be 16 Pin connector ILDA Cable Figure 16 Detector output diagram CASE Global Technologies 22 TMS
8. Table 27 Station 098 May 27 2010 Station 098 EB Evaluation 3 Hourly Errors Lower Bound 5 27 10 Totals sv LV Average 3 4 2 4 51 3 Stdev 1 6 1 4 29 3 Average 2 6 2 4 9 4 Stdev 1 5 1 4 9 6 Table 28 Station 098 May 28 2010 Average 2 8 1 6 61 6 1 9 1 8 41 0 Average 1 7 1 6 5 6 Stdev 1 8 1 8 10 6 6 3 2 4 FOURTH ROUND OF AGGREGATED DATA COMPARISON ONE MONTH RESULTS The fourth evaluation period reflects longer evaluation periods The following data shows the lower bound and upper bound errors of the aggregated data over a one month period for each evaluation site excluding weekends in March 2011 because of low LV counts Once again the period of analysis included data from 6 00 a m to 9 00 p m only The outliers were removed from the data analysis and were not presented below The data was collected in one hour aggregated bins Therefore the hourly errors in the following tables were calculated using every bin in the aggregated data as an error value The daily errors were calculated from a day s worth of aggregated data from the one hour data bins there is only one error value per day The daily hourly average errors were computed by taking all of the hourly errors in that day and averaging them together to produce one error value This error value is different than the daily error because the daily error was comprised of a single error value while t
9. 100 Figure 7 The last step is to connect the TMS 100 cable to the back of the unit Figure 8 Figure 7 Bolt the antenna through the top of the cabinet Drill a larger hole through the paneling inside the cabinet to provide access Figure 8 Connect the TMS 100 cable to the back of the TMS unit in the corresponding digital input port 4 3 5 COMMUNICATIONS CHECK After the installation was complete the system needed to be checked to make sure that the cables were landed correctly and that the signal strength of the wireless connection could transfer data reliably At the time of installation field technicians could not bypass the Caltrans network firewall using an external laptop modem to access the device s system As a workaround it was decided to attach a laptop directly to the unit However the TMS 100 does not provide a COM port Instead the TMS 200 unit which includes a COM port was used as a proxy to view live data and verify that the installation was correct Using the HyperTerminal interface vehicle traffic was visually observed as it passed over the detector loops The data was then compared to the live data from the TMS 200 After the ground verification was complete the team checked that the main database was receiving the information by accessing the TMS 200 from the main terminal and verifying the connection 17 Figure 9 Communications check using a laptop and TMS 200 4 4 TMS 100
10. Card MS SQL Server Easy installation No interference with existing Controller Real time detector output polling Data collected every 15 minutes Volume counts 50 and 85 Speed Vehicle Length Classification Lane Occupancy Figure 1 TMS system architecture formerly called InfoTek 4 2 TMS 100 ALGORITHM The TMS algorithm which is used to provide long vehicle counts via each lane s inductive loop detector is proprietary to CASE Global Technologies Users can configure the algorithm parameters based on the characteristics of the site After the system is set up the long vehicle counts are uploaded to a centralized database The inductive loop detector card reports the presence of each vehicle passing over the loop The presence signal remains on while the vehicle passes over the loop The algorithm calculates the speed based on the selected reference lane and estimated average length Typically the reference lane is the left fast lane which has mostly car traffic so the average length is easier to estimate The reference lane is then used to estimate the speed of traffic in the slower lanes based on the provided scaling factors For example the traffic in the right lane where the longer vehicles typically travel might travel at 90 of the speed in the fast lane All other lanes except for the HOV lane base their speed on the reference lane and the scaling factors After the speed is estimated the ve
11. Phillips Ranch Rd North 120 LA 60 R 26 57 Eastbound east of JCT RTE 57 EB on from Diamond Bar Blvd North 671 LA 60 R 29 8 Eastbound east of JCT RTE 71 EB on from RTE 71 Rio Rancho Rd 671 LA 60 R 29 8 Westbound east of JCT RTE 71 WB on from Reservoir St 37 LA 710 7 6 North of Pacific Coast HWY JCT NB shoulder RTE 1 40 LA 710 19 1 Northbound North of Firestone INB on from EB Firestone Blvd Blvd 40 LA 710 19 1 Southbound North of Firestone SB on from EB Firestone Blvd Blvd 41 LA 710 23 28 South of JCT RTE 5 NB shoulder 435 LA 710 23 75 South of JCT RTE 60 SB shoulder 436 LA 710 27 11 North of JCT RTE 10 SB shoulder 220 LA 605 2 31 North of Carson St NB shoulder 222 LA 605 11 0 North of Telegraph Rd NB shoulder 63 3 SCHEMATIC OF DATA COLLECTION SET UP Controller Cabinet Camera Piezo Loop Automated Vehicle Classification Station 64 Controller Cabinet Camera Loop Vehicle Count Station 65 4 TMS 100 HANDBOOK The attached TMS 100 User Guide provides information on setting up the units and modifying the parameters 66 CASE Global Technologies 1 TMS 100 User Guide TMS 100 Wizard 1 User Guide Copyright O 2010 CASE Global Technologies Inc All rights reserved 3 28 2010 http www casesystemsinc com Tel 949 268 1865 18 Morgan Ste 200 A Irvine CA 92618 CASE Global Technologies TMS 100 User Guide Ta
12. SYSTEM SETUP AND CALIBRATION To obtain high quality data from the TMS system each unit must be set up and calibrated to reflect the conditions specific to the site Software setup and calibration can be performed remotely using the TMS 100 s cellular wireless capability However loop detector calibration must be performed on site to ensure that vehicles are correctly classified The TMS system and loop detector system in lane loop detectors and loop detector cards work in conjunction with each other so they must be calibrated simultaneously and iteratively The main site specific inputs are the number of lanes average vehicle length reference lane and lane factor e Average vehicle length Interior lanes with a higher volume of small vehicles will have a smaller average length compared to the outer lanes which might carry higher volumes of large vehicles e Reference lane Lane in which speeds will be referenced from Ideally the reference lane has consistent speeds the leftmost lane has the most consistent speeds in most cases e Lane scaling factor Percent decrease in speed moving away from the reference lane 4 4 1 COMMUNICATING WITH THE TMS 100 You can access the TMS 100 through the HyperTerminal interface or a web browser 18 4 4 1 1 HYPERTERMINAL INTERFACE To use the HyperTerminal interface 1 Select Start gt All Programs gt Accessories gt Communications gt HyperTerminal 2 To connect the com
13. calibration Station 098 reflected the high error rates observed in the PVR analysis caused by over counting the LVs as explained in the PVR analysis Station 037 also had similar high error rates However both these stations performed strongly in identifying SVs The most promising station was 040 which had lower LV errors than SV errors Table 16 Table 17 and Table 18 show the hourly and daily errors for each evaluation site Figure 21 Figure 22 and Figure 23 display the frequency of LV errors Note the term bin refers to the aggregated LV errors in a 5 10 or 15 range in Figs 21 22 and 23 respectively For example in Fig 20 Bin O contained the LV errors 2 5 to 2 5 The results presented in the tables and figures are from 6 00 a m to 9 00 p m The hours between 9 00 p m and 6 00 a m had low volume counts and were deemed unreliable The definition of error in each category SV and LV is as follows Error AVC volume TMS volume AVC volume x 100 Table 16 Station 037 northbound hourly and daily errors Station 037 NB Evaluation 1 Hourly Errors Total NB SV NB LV NB Average 0 14 1 76 27 73 Stdev 2 55 3 45 25 12 Station 037 NB Evaluation 1 Daily Errors Total NB SVNB LV NB Average 1 79 0 03 25 10 Stdev 5 80 5 77 9 82 40 gt o E 9 3 o w M Station 037 Evaluation 1 NB LV Error Frequency Figure 21 Station 037 no
14. connections A PPP modem driver must be installed on PC Figure 6 AT Mode CASE Global Technologies 8 TMS 100 User Guide Rear Panel and Connectors The following figure illustrates the rear panel of the TMS 100 E EA PIG 32 DIGITAL INPUT CHANNELS 8 DIGITAL OUTPUT CHANNELS Figure 7 TMS 100 Wizard 1 Modem Back 1 32 digital input channels There are 4 sets of 8 channel connectors that make up 32 channel inputs A cable with 16 pin ribbon connectors connects to one of the four TMS connectors shown above M Fora single loop configuration one cable would support up to 8 lanes See TMS Cable Mapping M For dual loop configuration a dual loop with 4 lanes would take up to 8 channels and use only one cable Any more lanes would require an extra cable See TMS Cable Mapping The cable U shaped connector connects to a detector rack where the output of the detector card is monitored Each cable wire is labeled with a number This cable is supplied by Case Global Technology i Appendix A A ground wire is also supplied which must be properly grounded for the detector output to be properly sensed CASE Global Technologies 9 TMS 100 User Guide 2 8 Digital Output Channels Similar to the 32 input channels above the 10 pin ribbon connector is connected to the TMS 100 that carries digital 8 different java programmable digital output channels 3 GSM Antenna There are var
15. for Station 098 revealed a high number of external errors that were not associated with the TMS system As a result the first set of aggregated data could not be used for the evaluation The TMS algorithm parameters for lane by lane reference speed and length were adjusted to address the misclassifications In addition the threshold for LVs was increased from over 40 feet to over 50 feet However because of limited staffing Caltrans could not calibrate the loop detector sensors to address the cross talk errors observed in Station 098 The data was collected again with improved results The TMS parameters were then fine tuned and a third set of aggregated data was collected The data was collected again with improved results The TMS parameters were then fine tuned and a fourth set of aggregated data was collected over one month for long term analysis The aggregated data was collected over four time periods 37 e Round 1 Two week collection March 14 27 using a 40 foot threshold e Round 2 Two day collection May 17 19 and 20 Day 1 used a 40 foot threshold while Day 2 used a 50 foot threshold and TMS parameter adjustments to reference speed and length e Round 3 One day collection May 26 27 and 28 using a 50 foot threshold and further adjustments to the TMS reference speed and length parameters e Round 4 One month collection March 2011 using a 50 foot threshold and further adjustments to the TMS reference speed and leng
16. if traffic speed fluctuates more then a shorter time interval would be preferred Modify Settings Listening port 12345 Destination IP null Trap distance in feet 20 v Short Collection Interval 60 seconds Stored Collection Interval 15 minutes v Figure 9 Modify Settings 15 CASE Global Technologies 16 TMS 100 User Guide Dual Loop In dual loop configuration TMS polls both the lead and trail loops in real time and provides accurate volume speed occupancy and vehicle length counts The dual loop does not require the user to provide reference lane or speed factor in writing the loop string however the user needs to define the length bin thresholds Reporting intervals are also user definable Stored Collection Interval 15 minutes Lane configuration Dual loop Y Number of lanes 2 Loop String 20 40 60 Reset after Idle Minutes Modify Settings Figure 10 Dual Loop In this example M The Stored Collection Interval is set to 15 minutes 15 minute bins Data can also be stored in 5 30 and 60 minute intervals M The loop string defines 4 bins based on the 20 40 and 60 foot thresholds B TMS will reset after 3 hours if there is no attempt to communicate with it Note that the data collector contacts the TMS 100 on a regular basis so the modem can go several days without a reset CASE Global Technologies 17 TMS 100 User
17. lengths was observed such as LV SV SV LV in both the PVR data and video data with the correct time intervals between the vehicles the PVR timestamp would be adjusted to match the occurrence in the video To better analyze the data high resolution still images were extracted from the video at 30 frames per second fps using FFmpeg software Lanes were analyzed individually If it was observed that a vehicle crossed the loop detector the classification of the vehicle and the frame in which it crossed the loop detector were noted Vehicles were classified based on the number of axles similar to the classification used by the AVC system The frame number was converted into an image timestamp using the fps conversion FrarH Times ro The image timestamps and vehicle classifications were entered into a spreadsheet containing the PVR data Table 1 shows a sample of the spreadsheet 29 Table 1 Sample spreadsheet The colored portion is PVR data the non colored portion is video data Length Video Video Range of 30 Hz Video Video Lane feet Timestamp Time Seconds Image Image Lane Class 4 71 02 13 0 00 17 2 17 495 525 510 4 9 3 32 02 17 6 00 21 6 22 645 675 652 3 6 4 54 02 19 9 00 23 9 24 705 735 715 4 6 5 64 02 27 6 00 31 6 32 945 975 944 5 9 4 84 02 31 5 00 35 5 36 1065 1095 1040 4 9 5 3 LONG VEHICLE CLASSIFICATION CRITERIA TMS classifies vehicles into two bins bin 1 includes
18. northbound and eastbound directions as main roads and numbers them 1 4 The Traffic Management group uses the lane closest to the cabinet as the main road and it numbers the lanes 1 4 see Figure 4 When using shared cabinets the Census group follows the Traffic Management nomenclature 15 B Controller Cabinet Controller Cabinet I I m Figure 4 Lane naming schemes Left Census group uses the eastbound direction Right Traffic Management group uses the lane closest to the cabinet 4 3 4 FIELD INSTALLATION The physical TMS 100 unit connects to the Input File of the cabinet and plugs into a 120v power source The TMS 100 has an 8 wire cable Figure 5 Match the cable mapping guide to the detector layout sheet to identify where on the controller landing board each cable wire goes After the cable is landed properly connect it to the corresponding digital input channel on the TMS 100 unit Each cable must be grounded and any unused wires must be wrapped with electrical tape Figure 6 Figure 5 TMS 100 8 wire cable Each wire corresponds to a single loop Map the color coded wires to the lanes beforehand similar to Figure 3 Figure 6 Land the individual input wires to the Input File according to the detector layout sheet and cable mapping plan 16 To install the antenna drill a hole through the top of the cabinet Thread the antenna bolt through the hole and connect the antenna wire to the TMS
19. spanning May 17 19 and 20 2010 The first day used a 40 foot threshold for LVs For the second day the threshold was increased to 50 feet The results for this analysis are again from 6 00 a m to 9 00 p m The following tables show the lower bound and upper bound errors of the aggregated data over a two day period for each evaluation site On the first day the TMS threshold parameter was set to 40 feet on the second day the threshold was increased to 50 feet In the results for Station 037 the second day had noticeable improvements after the length threshold had been adjusted as shown in Table 19 and Table 20 Similar improvements occurred at Station 098 as shown in Table 23 and Table 24 43 Table 19 Station 037 May 17 2010 Station 037 NB Evaluation 2 Hourly Errors Lower Bound 5 17 10 NB Total NB SV NB LV Average 0 93 1 38 21 24 Stdev 0 37 0 94 11 52 Station 037 NB Evaluation 2 Hourly Errors Upper Bound 5 17 10 NB Total NB SV NB LV Average 0 42 1 38 14 66 Stdev 0 33 0 94 7 33 Table 20 Station 037 May 20 2010 Station 037 NB Evaluation 2 Hourly Errors Lower Bound 5 20 10 Total NB NB SV NB LV Average 0 92 0 06 9 07 Stdev 0 47 0 66 7 55 Station 037 NB Evaluation 2 Hourly Errors Upper Bound 5 20 10 Total NB NB SV NB LV Average 0 37 0 06 2 61 Stdev 0 40 0 66 5 2
20. to the TMS 100 the evaluation used the solar powered TMS 300 at one of the sites that did not have adequate power The TMS 300 has similar features to the TMS 100 but offers the advantage of using sustainable energy The project also documented the TMS installation and set up procedures 10 4 EQUIPMENT INSTALLATION DOCUMENTATION 4 1 TMS LOOP DETECTION SYSTEM TMS is a loop detection application that delivers volume speed loop occupancy and length classification data over either a secure wireless GSM network or existing fiber see Figure 1 The TMS solution can use the existing inductive loop infrastructure and it does not interfere with equipment such as traffic controllers and ramp meters Traffic data is remotely and automatically collected into an Oracle Microsoft SQL Server or Microsoft Access database manual downloading is not required The system can store traffic data for one month even if the communication link is broken The application is pure Java software and runs on Java programmable wireless modems installed in the traffic cabinets CASE Global Technologies offers three versions of the TMS solution The TMS 100 was used for the majority of the evaluation The TMS 200 was used briefly to perform the final communication check after the installation The TMS 200 includes an additional serial port and an Ethernet port which allowed the technicians to easily attach a laptop It also has GPS for recording an accurate ti
21. 05 Bancroft Way Berkeley CA 94720 Phone 510 642 4522 Fax 510 642 0910 Execution Period 1 15 2009 6 30 2010 Contract Amount 250 000 Principal Investigator Professor Samer Madanat Center Director Thomas H West Executive Director CCIT Project Manager Ali Mortazavi Senior R amp D Engineer CCIT Administrative Officer Richard Kleinman CCIT Additional Researchers Will Tabajonda Graduate Student Researcher CCIT Seongkyun Cho Graduate Student Researcher CCIT John Drazin Undergraduate Student Researcher PATH Ashkan Sharafsaleh Senior R amp D Engineer PATH DISCLAIMER STATEMENT This document is disseminated in the interest of information exchange The contents of this report reflect the views of the authors who are responsible for the facts and accuracy of the data presented herein The contents do not necessarily reflect the official views or policies of the State of California or the Federal Highway Administration This publication does not constitute a standard specification or regulation This report does not constitute an endorsement by the Department of any product described herein For individuals with sensory disabilities this document is available in Braille large print audiocassette or compact disk To obtain a copy of this document in one of these alternate formats please contact the Division of Research and Innovation MS 83 California Department of Transportation P O Box 942873 Sacramento CA
22. 09 23 59 59 80 70 60 50 40 Speed mph 30 20 10 0 00 00 02 00 04 00 06 00 08 00 10 00 12 00 14 00 16 00 18 00 20 00 22 00 Minimum Mean Maximum Figure 18 Station 040 traffic flow 5 1 3 VIDEO CAMERA PLACEMENT Part of the site selection process was to find a safe and convenient location for the video cameras For maximum safety and visibility the ideal placement would have been atop an 27 overpass but because of the physical constraints of the sites the only viable option was along the freeway shoulder a short distance away from the sensors Caltrans supplied a truck with a boom lift to provide on site safety and an elevated platform for recording Figure 19 Actual placement along freeway shoulder 5 2 DATA COLLECTION The project used video recordings PVR data and aggregated data to evaluate the TMS 100 On site videos were taken to validate the TMS PVR data The video recordings were also used to ensure that the AVC data was accurate and robust and had come from highly calibrated system It was important since the AVC data would be used as the ground truth baseline Data was collected remotely from the AVC and TMS systems to analyze the aggregated data 5 2 1 VIDEO DATA COLLECTION AND PROCESSING The team videoed each site to use as a one to one comparison for validating both the AVC data and the TMS 100 data The video data helped researchers assess and identify diffe
23. 100 User Guide FAQ This section shows some of the frequently asked questions How can we minimize the amount of data usage to keep our monthly wireless bills lower Here are some pointers that will minimize the data usage 1 Make sure that UDP broadcast is not on This can be done by going to the TMS web page of each TMS 100 to make sure that no destination IP is specified This will make a big difference 2 In the task manager you can adjust the polling to every 45 minutes instead of every 15 minutes This should help some 3 Also once this is done Check the bill to see what your usage is after a month You may find that the usage is lower Basic Networking ASCII American National Standard Code for Information Interchange AT Command Set Commands issued by intelligent device to a modem to perform functions such as to initiate call to answer call or to transmit data Bandwidth The amount of transmission capacity that is available on a network at any point in time EDGE Enhanced Data rates for GSM Evolution this 3G technology allows wireless transmission of data at speeds up to 384K bit sec It s based on GSM technology and allows for high bandwidth services such as multimedia It has more support in North America than in other areas where technologies such as CDMA2000 and UMTS may be favored Ethernet International standard networking technology for wired implementations GPRS Standard for packet communicatio
24. 3 028 gt Lane 4 report gt lane 4 trail loop trail loop trail loop lead loop trail loop trail loop 3 MPH 4 ft 2005 09 21 17 08 56 567 trail loop trail loop trail loop lead loop trail loop 85 MPH 10 ft 2005 09 21 17 09 01 569 trail loop trail loop trail loop trail loop lead loop lead loop trail loop EEE E E E E PRP E E E E E E E E P ANSIW TCP IP Figure 13 Lane 4 report showing extra trail loop reports causing incorrect speed and length calculations The user types 4 at the gt prompt The user should see lead and trail loop combinations like lane 1 in the figure below Instead the user is seeing too many tailing loops for lane 4 in the figure above This information is used in remotely diagnosing detection problems ODOT HyperTerminal File Edt View Cal Transfer Help DF oS 05 Y 2 lead loop trail loop 51 MPH 15 ft 2005 09 19 16 08 14 779 report 1 lead loop trail loop 50 MPH 18 ft 2005 09 19 16 08 32 326 lead loop trail loop 43 MPH 18 ft 2005 99 19 16 08 48 993 lead loop trail loop 41 MPH 13 ft 2005 09 19 16 08 53 33 lead loop trail loop 49 MPH 15 ft 2005 09 19 16 09 16 291 lead loop trail loop 62 MPH 15 ft 2005 09 19 16 09 41 145 lead loop trail loop 42 MPH 27 ft 2005 09 19 16 09 47 884 Figure 14 Lane 1 report shows lead and trail events plus the resulting length and speed calculations CASE Global Technologies 20
25. 7 Station 040 had good results in the initial two week evaluation In the second evaluation the number of errors did not change significantly as shown in Table 21 and Table 22 This might be attributed to better TMS calibration at Station 040 44 Table 21 Station 040 May 17 2010 Station 040 SB Evaluation 2 Hourly Errors Lower Bound 5 17 10 Totals SV LV Average 3 76 3 76 4 87 Stdev 1 84 1 90 8 77 Station 040 SB Evaluation 2 Hourly Errors Upper Bound 5 17 10 Totals SV LV Average 5 15 5 15 4 87 Stdev 1 43 1 90 9 95 Table 22 Station 040 May 19 2010 Station 040 SB Evaluation 2 Hourly Errors Lower Bound 5 19 10 Totals SV LV Average 16 30 5 77 5 73 Stdev 2 41 2 48 4 50 Station 040 SB Evaluation 2 Hourly Errors Upper Bound 5 19 10 Totals SV LV Average 18 49 6 44 5 73 Stdev 1 85 2 48 10 97 Table 23 Station 098 May 17 2010 Station 098 EB Evaluation 2 Hourly Errors Lower Bound 5 17 10 Total SV LV Average 4 02 0 57 104 81 Stdev 1 05 0 43 35 39 Station 098 EB Evaluation 2 Hourly Errors Upper Bound 5 17 10 Total SV LV Average 3 54 0 57 73 63 Stdev 0 93 0 43 20 32 Table 24 Station 098 May 19 2010 Station 098 EB Evaluation 2 Hourly Errors Lower Bound 5 19 10
26. 7 152 Lu Y Vehicle Classification Using Infrared Image Analysis ASCE Journal of transportation Engineering Vol 118 No 2 March April 1989 pp 223 240 Martin Peter T Yuqi Feng and Xiaodong Wang Detector technology evaluation www mountain plains org pubs html mpc 03 154 pg1 php November 2003 Middleton Dan and Rick Parker Initial evaluation of selected detectors to replace inductive loops on freeway FHWA TX 00 1439 7 April 2000 Mimbela Luz Elena Y and Lawrence A Klein A summary of vehicle detection and surveillance technologies used in intelligent transportation systems Vehicle detector clearing house New Mexico State University Fall 2000 57 Minnesota Department of Transportation Field Test of Monitoring of Urban Vehicle Operations Using Non intrusive Technologies Final Report SRF Consulting Group INC MN May 1997 Minnesota Department of Transportation NIT Phase 2 Evaluation of Non intrusive Technologies for Traffic Detection Final Report St Paul MN September 2002 Nojima A Iwata Y and Hirano K An in vehicle information system using simple deformed map displays and two way infra red beacons Proceedings of the 1995 Annual Meeting of ITS AMERICA pp 201 206 Nooralahiyan A Y et al Trial of Acoustic Signature Analysis for Vehicle Classification Transportation research C Vol 5 No 314 1997 pp 165 177 Pursula M and Pikkarainen P A Neural Network Approach to Vehicle Classi
27. 7 MPH 15 IT 56 569 64 MPH 17 It 15 57 307 72 MPH 15 ft 15 57 605 59 MPH 18 ft 16 43 3237 TEHER 1500 16 423 701 KL MPH ltt 16 44 906 67 MPH 55 ft 16 46 47 63 MPH 15 ft 16 46 581 63 MPH 10 ft 16 48 74 7 MPH 56 Lt 16 499 5129 67 MPH 13 ft 16 49 71 MPH 15 ft 16 49 907 83 MPH 17 ft 16 50 654 56 MPH 63 ft 16 51 559 71 MPH 12 ft 16 52 9 Ka L FG NW Fa FA EA H Ln r 3 2 1 2 4 2 4 3 4 1 1 2 1 Nw Nw Ww PA PA EPA te to tt fo Connected 0 00 09 ANSIW TOPP Figure 11 PVR data via the HyperTerminal interface If vehicle speeds or lengths are inconsistent or it is observed that the loop detectors are missing vehicles or experiencing cross talk calibration is required To refine the accuracy of the system adjust the sensitivity of the detector card and the TMS software parameters until the PVR data approximates the live traffic Knowledge of the site hardware and typical traffic conditions can help with calibrating the system and deciding which inputs to change Because the speed and vehicle length calculations come from two independent systems the technician must decide whether to adjust the sensitivity of the loop detector card or the TMS software parameter For example while the innermost lane might be considered the fast lane with little to no truck traffic in most situations there are instances where the left lane might serve as an off ramp for trucks so the average vehicle length must be adj
28. 94273 0001 vi ACKNOWLEDGEMENTS The project team owes special thanks to Caltrans employees Lianna Bergan and Jesse Gonzales who were injured while helping the project team collect data along CA 710 The project team also warmly thanks the following people in no particular order who invaluably helped the project by committing their time and knowledge Anyone mistakenly forgotten please accept our deepest apologies e Frank Quon Allen Z Chen and Steve M Malkson Caltrans District 7 e Homar Noroozi and John Slonaker Division of Research and Innovation Caltrans e Mike Poursartip CASE Systems This work was supported by Caltrans Division of Research and Innovation vii Limited Deployment Pilot Project Monitoring Truck Traffic in Caltrans District 7 Final Report July 2010 Prepared by Ali Mortazavi Ph D Senior R amp D Engineer Will Tabajonda Graduate Student Linda Fogel Author Seongkyun Cho Graduate Student John Drazin Undergraduate Student Ashkan Sharafsaleh Senior R amp D Engineer For CALIFORNIA DEPARTMENT OF TRANSPORTATION california center for i ive transportation UNIVERSITY OF CALIFORNIA BERKELEY Table of Contents Executive SUMMA y aesaad ae aniar a lee ap ibi biendo U4 bogs dond degeng eeddega ne degend ey ddat iii Project lo id V Disclaimer Statement eee eee eee ee eee eee eee AS vi Acknowledgements EEE EE NE vii California Department of transportation see eee eee eee eee eee 1 T
29. CALIFORNIA CENTER FOR INNOVATIVE TRANSPORTATION INSTITUTE OF TRANSPORTATION STUDIES UNIVERSITY OF CALIFORNIA BERKELEY Limited Deployment Pilot Project Monitoring Truck Traffic in Caltrans District 7 Ali Mortazavi Ph D Senior R 8 D Engineer Will Tabajonda Graduate Student Linda Fogel Author Seongkyun Cho Graduate Student John Drazin Undergraduate Student Ashkan Sharafsaleh P E Senior R amp D Engineer CCIT Research Report UCB ITS CWP 2011 1 california center for ive transportation UNIVERSITY OF CALIFORNIA BERKELEY ISSN 1557 2269 The California Center for Innovative Transportation works with researchers practitioners and industry to implement transportation research and innovation including products and services that improve the efficiency safety and security of the transportation system CALIFORNIA CENTER FOR INNOVATIVE TRANSPORTATION INSTITUTE OF TRANSPORTATION STUDIES UNIVERSITY OF CALIFORNIA BERKELEY Limited Deployment Pilot Project Monitoring Truck Traffic in Caltrans District 7 Ali Mortazavi Ph D Senior R amp D Engineer Will Tabajonda Graduate Student Linda Fogel Author Seongkyun Cho Graduate Student John Drazin Undergraduate Student Ashkan Sharafsaleh P E Senior R amp D Engineer CCIT Research Report UCB ITS CWP 2011 1 This work was performed by the California Center for Innovative Transportation a research group at the University of California Berkeley in cooperation with the
30. Guide TMS Telnet Session A telnet session is established to look at live traffic Useful diagnostic commands include h 1 9 I and v Hyper terminal can be used to establish a telnet session See Connect To dialog below as an example of how to connect to a host address port combination using TCP IP D1 2Single Properties Connect To Settings Bos Host address 119256994774 intemet mpcingular co Port number 112345 Connect using Figure 11 Hyper terminal configuration R Once logged in using telnet the user will see the gt prompt The user can type h for help Below are some useful commands and how they work Other telnet programs such as Tera Term or Putty can be used too R Once done using telnet the user must log out using the x command If not the user will likely have trouble reconnecting for the next few minutes R if no commands are typed the telnet session will log the user off after approximately 2 minutes This is to prevent wasting system resources The user can easily re log in if disconnected R Hyper terminal icons can be saved in a folder The user only needs to double click on the icon to log in The user can type h command to prolong the session CASE Global Technologies 18 TMS 100 User Guide Diagnostic Commands R y command toggles on and off the vehicle report The vehicle report which includes the speed vehicle length and ti
31. LV errors we more normal except station 037 SB in nature centered around 0 error which was a vast improvement from the previous rounds worth of collected data Therefore the SV counts would require a multiple factor to correct for the systematic behavior while the LV counts would be ready for data analysis as obtained from the field Figures 24 29 show hourly error frequencies or distributions using only the lower bound LV data Figures 24 26 and 28 show the lower bound error frequencies in a histogram The major observation was that the LV error data was roughly normal in the distributions a vast improvement from previous methods which had no apparent shape The majority of data was contained within 25 from the calculated averages for each graph Only a few hourly error points were close to 100 error in either direction and excluded from the analysis because of their obvious outlier status Therefore the collected data was improved significantly from the previous rounds Figures 25 27 and 29 show how the LV lowers bound error data varied by time of day The graphs contained multiple points including data points small blue squares undercount mean sky blue bar over count mean green triangle absolute mean red circle and regular mean orange diamond The over count mean averaged all of the error values which were negative less than 0 while the undercount averaged all of the positive greater than 0 errors The absolute avera
32. March 2011 Hourly Errors Daily Errors Daily Hourly Average Average 2 28 0 10 2 06 Stdev 18 51 4 25 5 55 Peak Average 4 16 N A N A Station 040 NB Evaluation 4 Upper Bound LV Weekdays March 2011 Hourly Errors Daily Errors Daily Hourly Average Average 9 93 10 71 10 22 Stdev 18 40 4 53 4 74 Peak Average 10 48 N A N A Station 040 NB Evaluation 4 SV Weekdays March 2011 Hourly Errors Daily Errors Daily Hourly Average Average 5 56 5 67 5 69 Stdev 6 16 5 55 0 73 Peak Average 3 77 N A N A 50 40 30 FREQUENCY 20 10 Station 040 Evaluation 4 NB LV Error Frequency OOOO o N BIN o mM TL 10 mmm Figure 28 Station 040 NB LV error frequency 53 60 a o rr Tr eS amp Station 040 Evalution 4 NB Lowersouno sa LV Error Frequency A Overcount MEAN i eAbsolute MEAN 60 00 Undercount MEAN MEAN 40 00 z A L gt s e 20 00 gt s z x i A P S 0 00 f i t i 1 y t y H v i i si P 4 x 4 H b 20 00 n x 1 U 1 a i 8 40 00 E s 60 00 LU 80 00 7 TIME OF DAY 100 00 an a o co nm w gt on a co co o o o o o a N o o o o o o o o o o o oO o o o o o o o o o o o o o o gt gt gt gt gt o oa a v v v v v v v v v 7 En z z Figure 29 Station 040 NB LV error d
33. PN creates a private encrypted tunnel from the end user s computer through the local wireless network through the Internet all the way to the corporate servers and database WAN A communication system of connecting PCs and other computing devices across a large local regional national or international geographic area A AT modem 7 B bins 12 14 D Database 33 H hostname 11 14 HyperTerminal 37 TMS 3 6 8 12 14 16 17 18 19 22 23 26 28 29 30 33 34 inputs 3 8 IP address 10 11 14 26 loop detection 3 27 LOOP STRING 15 Modes 5 INDEX CASE Global Technologies 24 TMS 100 User Guide O occupancy 3 12 17 27 Output 9 34 35 RS 232 Port 4 36 speed factor 15 17 telnet 19 UDP broadcast 14 volume 3 W TMS 3 4 5 6 8 9 10 11 12 14 22 23 31 35 TMS 100 Modem 3 6
34. Peak Average 1 67 N A N A Station 037 SB Evaluation 4 Upp er Bound LV Weekdays March 2011 Hourly Errors Daily Errors Daily Hourly Average Average 17 93 17 99 17 12 Stdev 9 03 3 43 4 53 Peak Average 15 79 N A N A Station 037 SB Evaluation 4 SV Weekdays March 2011 Hourly Errors Daily Errors Daily Hourly Average Average 5 80 5 55 5 81 Stdev 5 68 2 07 1 99 Peak Average 5 73 N A N A 51 E v bis v Wd 00 70 Bis Wd 00 50 Wd 00 90 s LOWERBOUND Undercount MEAN AOvercount MEAN eAbsolute MEAN MEAN Wd 00 20 Wd 00 80 Wd 00 60 Station 037 Evaluation 4 SB Station 037 Evaluation 4 SB LV Error Frequency 80 0 Og Op I DEl UZ I 041 00 06 08 02 09 Og Of Og OG OL DL oz UE 0p OS 09 OL 08 06 00l DLL UZ L UE L UPL OG LOWERBOUND Undercount MEAN aOvercount MEAN Absolute MEAN amp MEAN Bin TIME OF DAY LV Error Distribution 50 30 20 Figure 26 Station 037 SB LV error frequency 40 09 00 PM 08 00 PM 07 00 PM 06 00 PM 05 00 PM 04 00 PM 03 00 PM 02 00 PM 01 00 PM 12 00 PM 11 00 AM 10 00 AM 09 00 AM 08 00 AM 07 00 AM 06 00 AM 05 00 AM 52 Figure 27 Station 037 SB LV error frequency Table 31 Station 040 NB March 2011 Station 040 NB Evaluation 4 Lower Bound LV Weekdays
35. State of California Business Transportation and Housing Agency s Department of Transportation and the United States Department of Transportation s Federal Highway Administration The contents of this report reflect the views of the authors who are responsible for the facts and the accuracy of the data presented herein The contents do not necessarily reflect the official views or policies of the State of California This report does not constitute a standard specification or regulation August 2011 EXECUTIVE SUMMARY Truck transportation affects every aspect of traffic and infrastructure Because large trucks cause the majority of damage and stress to roadways and bridges as well as contribute significantly to congestion it is important to have accurate real time data on the number of trucks on the roadways peak travel times lane use and where trucks are going on the highway system Currently Caltrans relies on data collected from Automated Vehicle Classification AVC stations and manual census counts at specific locations to monitor and manage truck traffic This technology is expensive time consuming and disruptive to install as well as costly to maintain therefore it has not been deployed widely A less expensive and easier to install wireless technology was tested in a small scale trial in 2006 with positive results The CASE Global Technologies Traffic Monitoring System 100 TMS 100 previously called the Infotek Wizard perfo
36. a significant role in many aspects of transportation planning highway operational analysis and pavement and bridge design Saito 2009 Thus it is important to select appropriate detectors to gather relevant truck traffic data Methods for collecting truck traffic data can be divided into three categories automated or electronic traffic count systems such as WIM inductive loops pneumatic detectors and AVC video surveillance and visual classification counts such as tally sheets and mechanical counting boards Saito 2009 This assessment focuses on automated and electronic systems because most truck tracking systems used by Caltrans are automated In addition video surveillance costs too much and visual manual counting requires a highly skilled labor force An inductance loop system is one of the traditional methods for automated vehicle classification U S Department of Transportation 1997 It has some advantages It is fairly immune to weather conditions Juba 1996 Because a dual loop system has relatively higher accuracy up to 96 2 depending on the algorithm Pursula 1994 it is used to acquire ground truth data to evaluate the performance of other vehicle classifiers such as video image processors Minnesota Department of Transportation 2002 Middleton 2000 infrared detectors Minnesota 2002 Middleton 2000 Minnesota 1997 and microwave radar Minnesota 1997 Middleton 2000 Saito 1995 It can provide basic t
37. abla Ene arar ARA ennai inden 2 1 Introduction and Problem Statement sss sees eee ee eee eee ee eee 5 2 BO eee E 8 3 Technical Approach isis dekade 11 4 Equipment Installation Documentation sese ee eee ee eee eee ee eee 12 4 1 TMS Loop Detection System ccococccccnnononononononononononononononononononononononononononononononononononenenenene 12 4 2 TMS 100 Algorithms iria A A AA 13 4 3 TMS 100 Installation Instructions aia aaa 14 4 3 1 SIM Card Installation Instructions criar aio 14 4 3 2 Software Installation Instructions roca di aia 15 4 3 3 Cabl e Mapping Guld e ads 16 134 aile Tae ale 17 4 3 5 Communications Enea ua dies aasiainiaavaliavascnainsadu eva aavanarnnden 18 4 4 TMS 100 System Setup and Calibration esoarvrroennnnvvrnnnnvvrsnnnnvnvnnennnnnnrennnnnsnsnsnenseseennnnes 19 44 1 Communicating with the IMS 100 iii 19 4 4 2 Calibrating the Loop Detector aaa 20 45 Solar TMS 300 liste lada 23 5 Data Collection and Evaluation Plan 25 EL Se SENNA 25 5 1 1 Physical Characteristics niinen sataseen a ERE EREE ERAN 25 SA TEEN 27 5 1 3 Video Camera Placement vos de 28 52 Daa COUCCU OM vrede 29 5 2 1 Video Data Collection and Processing 29 5 3 Long Vehicle Classification Criteria nesnice ei ea aaae aeea E Saia 31 5 3 1 TMS 100 and AVC Data Collection and Processing sese eee eee eee eee 31 6 RESUIES A ee ee kor 34 6 1 Performance Measure Parameters for Video Comparison sese ee eee eee eee eee eee eee ee 34
38. age Average 6 76 6 86 9 43 Stdev 14 90 2 54 3 71 Peak Average 9 73 N A N A Station 037 NB Evaluation 4 Upper Bound LV Weekdays March 2011 Hourly Errors Daily Errors Daily Hourly Average Average 6 13 4 49 5 10 Stdev 15 07 4 28 5 17 Peak Average 10 11 N A N A Station 037 NB Evaluation 4 SV Weekdays March 2011 Hourly Errors Daily Errors Daily Hourly Average Average 2 92 2 72 2 92 Stdev 5 75 1 53 1 53 Peak Average 5 12 N A N A 49 Station 037 Evaluation 4 NB LV Error Frequency 70 50 40 0 IL ma 2222222222282 ooo G G G GO G C3 0 010 TON N D WD OF ODDO Ny MO LO 8 8 E 8 8 Bin e e r Frequency 2 o 1 o 150 140 130 120 110 100 90 Figure 24 Station 037 NB LV error frequency 50 80 60 40 x 20 L e 5 0 20 40 60 80 amp 8 o o o o gt gt Station 037 Evaluation 4 NB JE DE 3 i S 8 8 GS gt G eo gt pp gt S LV Error WY 00 01 WY OG LL Distribution L pis FER I TIME OF DAY S 2 8 8 S 8 8 8 TU TU TU TU Figure 25 Station 037 NB LV error distribution Table 30 Station 037 SB March 2011 Station 037 SB Evaluation 4 Lower Bound LV Weekdays March 2011 Hourly Errors Daily Errors Daily Hourly Average Average 2 23 1 94 1 11 Stdev 10 45 3 76 4 15
39. ble of Contents TABLEOECONTEN TS e R NOR Eaa RR ANNA A A a NAO RAON da dt bale 2 TMS TOO MODE Heer 6 CONFIGURAT ONAN A STATUS ur 10 TMS Eeee eE e een mnik e p naa se haee a eA me ATAATA ATE RERE PETATE TA PA TAA Ta baT A CATA nr 11 CONFIGURATION AND STATUS o 13 FOOP STRING CONFIGURATION 00 db E EE 14 TMS TELNET SESSION TT 17 SINGLE LOOP CONFIGURATION CABLE MAPPING xx x x e e x e eee 20 DETEGTOR OUTPUT DIAGRAM E A 21 CASE Global Technologies 3 TMS 100 User Guide The TMS 100 Wizard 1 is a wireless GSM modem with 32 channel digital inputs 1 RS 232 port 8 digital outputs and is java programmable The TMS 100 also operates as an external wireless modem with serial connectivity The TMS 100 is an ideal solution for telemetry SCADA and remote sensory devices The TMS 100 supports two different applications the TMS Modem and TMS Figure 1 TMS 100 Modem The TMS Modem is a wireless modem It functions like a dialup modem except it transmits data over the GSM wireless network AT amp T in the US instead of phone lines The TMS is used by departments of transportation to monitor traffic along highways and main roads TMS is an intelligent loop detection software that polls the output of any inductive loop detector card and provides volume speed loop occupancy and length classification Front Panel CASE Global Technologies 4 TMS 100 User Guide The following picture shows the front panel of t
40. browser a user can see the current status of the TMS 100 Modem From the Web page the user can configure the listening port Destination IP and baud rate of the modem Once the listening port and baud rate are set the Modify Settings button is pressed The new settings are stored and the modem is rebooted before the new settings take effect This process takes about 30 seconds Any data written into the serial port of the TMS 100 would be forwarded to the destination IP through UDP this IP address uses the same port number as the listening port CASE Global Technologies 11 TMS 100 User Guide TMS TMS is used by the transportation industry to monitor vehicle traffic on major roads and highways In a dual loop configuration TMS will measure traffic volume average speed standard deviation speed lane occupancy and vehicle length bins of 0 20 feet 20 40 feet 40 60 feet and 60 feet Bins are created for either 15 minute or 1 hour time intervals In a single loop configuration The TMS measures volume and lane occupancy only The TMS application must be properly connected using a special cable to the output of 222 type detector cards The Data collector in the traffic center will automatically poll the TMS 100 to download the traffic data to a database i e Oracle MS SQL or MS Access CASE Global Technologies 12 TMS 100 User Guide CHANNELS 17 32 Figure 8 32 digital input channels Im
41. cation technology employing single loop detectors changes with traffic flow especially with the onset of congested states of traffic To assess how the TMS 100 performed in different traffic states CCIT planned to evaluate congested and free flowing traffic situations Using the Freeway Performance Measurement System PeMS database the CCIT team identified peak hours and free flow hours for the evaluation sites Figure 16 Figure 17 and Figure 18 The PeMS results showed heavy congestion for all three sites at around 4 5 p m This time period was not captured for the one to one video analysis but it was looked at during the long term analysis of comparing the TMS and AVC systems Speed mph 4 320 Lane Points 100 Observed Segment Type VDS Segment Name 717265 10 06 2009 00 00 00 to 10 08 2009 23 59 59 Days Mo Tu We _Th Fr Speed mph 0 00 00 02 00 04 00 06 00 08 00 10 00 12 00 14 00 16 00 18 00 20 00 22 00 Minimum Mean O Maximum Figure 16 Station 098 traffic flow 26 Speed mph 3 456 Lane Points 74 Observed Segment Type VDS Segment Name 717962 10 09 2007 00 00 00 to 10 14 2007 23 59 59 Days Tu We Th Speed mph 0 00 00 02 00 04 00 06 00 08 00 10 00 12 00 14 00 16 00 18 00 20 00 22 00 Minimum Mean Maximum Figure 17 Station 037 traffic flow Speed mph 3 456 Lane Points 98 Observed Segment Type VDS Segment Name 718002 10 06 2009 00 00 00 to 10 08 20
42. constraints and then further reduced because of a major vehicle accident at one of the evaluation sites so the team could only collect one to two days worth of aggregated data Although this study was limited in the number of sites evaluated the results showed the TMS 100 performed with at least 95 accuracy as noted in the 2006 pilot Beyond the high level of accuracy demonstrated in the evaluation the TMS solution offers many other benefits The TMS 100 requires little or no construction is easy to install one or two sites can be deployed in a day can use the existing infrastructure provides remote access for calibration has lower per unit costs and is inexpensive to maintain In addition having a solar powered option expands the areas that it can be installed Because the system is less expensive and easier to install more data collection points can be established to track traffic patterns better plan ramp metering manage congestion and monitor pavement conditions for maintenance This report contains the following e Installation instructions e Description of how to customize the TMS algorithm e Setup and calibration information e Evaluation methodology and results e TMS User Manual PROJECT FACT SHEET Title Limited Deployment Pilot Project Monitoring Truck Traffic in Caltrans District 7 Sponsor Caltrans Division of Research and Innovation Executing Organization California Center for Innovative Transportation 21
43. ct classifications Lin 2004 However the Caltrans TMS 100 experiment showed that vehicles with lengths over 40 feet are detected correctly Considering that most trucks are over 40 feet the TMS 100 can provide adequate truck traffic data using the current single loop infrastructure instead of the more expensive AVC or WIM systems This research evaluates whether the TMS 100 can produce accurate truck traffic data and provide a long term cost effective solution 3 TECHNICAL APPROACH To assess the accuracy of the TMS 100 on a larger scale the team selected evaluation sites based on physical characteristics traffic volume and the availability of AVC stations The original goal was to compare a minimum of a week s worth of TMS data to the AVC data However before testing the TMS 100 the data collected by the AVC systems needed to be validated to establish a baseline To validate the AVC data the team videotaped the traffic for a 30 minute period to use as a one to one comparison The 30 minute videotape data was used to verify the robustness of the AVC data This process involved installing video cameras at each evaluation site and then manually matching the recorded traffic to the AVC data Another separate video recording was also used to evaluate the TMS data by manually matching the recorded traffic to the TMS data Using the video recordings both the AVC and TMS systems were further calibrated and fine tuned as necessary In addition
44. ctor technology was confirmed in a small scale 2006 demonstration project as part of the Innovative Corridors Initiative To evaluate the accuracy of the TMS 100 three existing Caltrans AVC sites in two districts were selected to use as a comparison Both systems collected data for one week on the number of vehicles longer than 40 feet The results were very promising with the less costly TMS 100 performing with at least 95 accuracy when compared to the AVC stations The system evaluated consisted of TMS 100 intelligent GPRS EDGE wireless modems and integrated programmable software that was able to post process the collected data to provide traffic information The system has 32 digital inputs and 8 digital outputs and can support up to 28 loops 14 loops in a dual loop configuration It collects traffic data from loop detectors and applies algorithms to the collected data in real time The device summarizes volume 8 occupancy and long vehicle counts into typical 15 minute bins The post processed results are relayed to traffic management centers via a cellular network TMS 100 works with existing polling systems and the data collecting software is compatible with Oracle MS Access and MS SQL databases TMS 100 is used in Caltrans District 10 as a controller in traffic monitoring stations Darter 2007 Usually a single loop assumes the vehicle length when it measures speed or classifies vehicles This method of assumption can cause incorre
45. d effective length lowered the error rates Many lessons were learned from the initial poor results from the first round of evaluation The analysis of the errors at Station 098 led to the adjustments made during the second third and fourth evaluations which improved system performance The errors observed showed the connection between the TMS unit and the equipment in the field Ensuring that the loop detector sensors are calibrated is just as important as adjusting the TMS software parameters It also demonstrated that the calibration process is iterative and that both systems might need to be adjusted simultaneously Lastly our analysis supports that the system can provide a viable cost effective solution for Caltrans to assess truck traffic volume based on these characteristics e Ability to use the existing widely deployed single loop detectors and infrastructure e Ease of installation e Capability of using various communication channels e Remote access 56 8 REFERENCES Carlson B Clearing the congestion vision makes traffic control intelligent Advanced Imaging February 1997 pp 54 57 Darter Michael T Stephen M Donecker Kin S Yen Bahram Ravani and Ty A Lasky Overview of Caltrans District 10 IRIS Demonstration Design AHMCT Research Report UCD ARR 07 12 31 02 December 31 2007 Dillenburg J Lain C Nelson P C Rorem D Design of the ADVANCE traffic information center Proceedings of the 1995 Annual Me
46. e the system Measure of accuracy level Average COTY Ma HL ting Mers T Mert Rate for correct identification long vehicle classification of given population Total _Correct_LV_Detection _ My Hit_Rate Testing LV Set Size METL Rate for incorrect identification of given population Total Missed LV Detection M c M Missed Rate SS AS AS i Testing LV Set Size MGTL Rate for misidentifying one population for the other Total False Alarms _ My False_Alarm_Rate Testing Alert Set Size Mgrs Rate for identifying accuracy level distance away from true value Saito 2009 B e gt r A EA A Mert ML E 6 PE C TY M GTL Table 8 Sample of measured parameters Station Classification Output Errors P Ground Truth 6 2 AVC VALIDATION RESULTS Because the AVC data was used as ground truth it needed to be validated to ensure that it was accurate Of the three sites chosen for evaluation analysis was performed only on stations 098 and 037 While collecting video data at Station 040 NB it was discovered that the wires leading to the piezo detectors were shorter than required resulting in low line resistance in the system This resistance caused abnormalities in the PVR data The issue was resolved by adding lead wire and recalibrating the system 34 Table 9 Table 10 Table 11 and Table 12 show the lane by lane AVC validation results Only truck driving lanes were anal
47. e then classified as either LV 40 feet and longer or SV For the AVC system this involved using an algorithm to identify classes 8 14 as LVs see Table 2 for the AVC classes Table 7 shows the processed AVC data classified into two bins Table 3 Sample AVC PVR data Vehicle Speed LPL Number Date Time Array Number Class MPH Feet of Axles 22 12 2009 00 02 05 Lane 3 1 9 39 55 5 22 12 2009 00 02 13 Lane 3 3 9 39 52 5 22 12 2009 00 02 17 Lane 3 5 9 32 57 5 22 12 2009 00 02 25 Lane 3 9 9 37 66 5 22 12 2009 00 02 36 Lane 3 13 9 38 56 5 Table 4 Sample TMS PVR data Lane Length ft Speed MPH Occupancy ms Timestamp lane 2 ft 14 MPH 64 220 ms 01 56 2 lane 4 ft 8 MPH 63 165 ms 01 56 1 lane 3 ft 17 MPH 63 265 ms 01 56 2 lane 1 ft 13 MPH 68 195 ms 01 56 7 Table 5 Sample aggregated AVC data collected directly from the system Class 1 Class Class Class 4 Class 5 Class 6 Class 7 Class Class 9 Class 10 Class 11 Class 12 Class 13 Class 14 Class 15 00 00 N B 1 North N B 2 North N B 3 North N B 4 North Table 6 Sample 15 minute aggregated TMS data collected directly from the system Each column shows lane by lane vehicle classification where C1 is lt 40 feet and C2 is gt 40 feet 0 10 21 747 107 0 12 0 0 1 7 0 0 1 0 0 0 1 31 0 0 0 0 0 0 0 1 0 0 0 0 0
48. e vehicle distribution by class for lanes 3 4 and 5 LV Distribution by Ground Truth Class for Lanes Count AVC LV Counts and Lengths 3 4 and 5 Veh Count Veh Count Avg Length Stdev Length Class 8 11 17 54 3 4 Class 9 90 82 63 8 3 Class 10 0 0 N A N A Class 11 2 2 61 2 0 Class 12 0 0 N A NA Class 13 1 0 N A N A Class 14 6 6 55 3 6 Class 15 28 N A N A Total LV Counted 110 107 AVC Class 15 counts are not included in the LV count Error has unknown distribution between LVs and SVs 39 6 3 2 TMS AGGREGATED DATA COMPARISON RESULTS After performing the one to one analysis it was determined that increasing the threshold length to 50 feet would reduce some of the definition errors It was also observed that class 15 errors might account for a significant number of over counted LVs The class 15 errors in the first round of collecting aggregated data had a 30 70 split between LVs and SVs The second third and fourth evaluations show all or nothing assignments with all class 15 vehicles classified as LVs The upper bound has all the class 15 vehicles as LVs and the lower bound has none of the class 15 vehicles as LVS 6 3 2 1 FIRST ROUND OF AGGREGATED DATA COMPARISON 1 2 WEEK RESULTS The aggregated data collected in the first round during the weeks of March 14 27 2010 was done prior to discovering that the loop detector sensors and TMS parameters required
49. es bracket assembly pole cap conduit and cable attachment 30 minutes e Installing foundation and conduit 2 hours e Installing pull box 45 minutes e Pulling loops through the base pole includes installing the pole to base and grounding rod 1 hour e Installing boxes on pole and soldering the DLC wire 1 hour e Installing pedestrian pad 1 hour When using rapid set concrete installing a TMS 300 box can be completed in three to four hours 22 Figure 12 Left Cabinet interior of a solar TMS 300 Right Field installation 23 5 DATA COLLECTION AND EVALUATION PLAN Three sites were selected to evaluate the accuracy of the TMS 100 in classifying vehicles CCIT s plan was to conduct the evaluation in two phases occurring simultaneously The first phase focused on performing a one to one comparison of the TMS per vehicle records PVRs to 30 minutes of traffic captured via video The goal for the second phase was to compare 60 minute bins of TMS aggregated data to AVC aggregated data over a period of two weeks The AVC data was used as a ground truth baseline However the evaluation did not go as planned In the process of comparing the TMS PVR data to the video data it was discovered that some of the single loop detector sensors as well as the TMS units were not calibrated rendering the aggregated data that was already collected as useless After the TMS 100 units were recalibrated which took away time from the evaluation t
50. eting of ITS AMERICA pp 321 328 Harlow Charles and Shiquan Peng Automatic vehicle classification system with range sensors Transportation research part C Emerging Technologies Volume 9 Issue 4 August 2001 pp 231 247 www sciencedirect com science visited June 4 2009 Harvey Bruce A Glenn H Champion Steven M Ritchie and Craig D Ruby Accuracy of truck traffic monitoring equipment Technical report GTRI Project A 9291 Communications and network division Information Technology and telecommunications laboratory Georgia Tech Research Institute June 1995 www fhwa dot gov ohim atme atme htm visited June 30 2009 Juba M Succeeding with video detection Traffic Technol Int Oct Nov 1996 pp 33 36 Kell J H Fullerton I J and Mills M K Traffic Detector Handbook FHWA IP 90 002 Federal Highway Administration 1990 Kistler Instruments Corporation Piezo electric force transducers Design amp Use Amherst NY 2000 Klein L A Sensor Technologies and Data Requirements for ITS Norwood MA Artech House 2001 p 549 Kunigahalli R A multidimensional geometric approach to detection of freeway congestion Proceedings of the 1995 Annual Meeting of ITS AMERICA pp 597 604 Lin Wei Hua Joy Dahlgren and Hong Huo Enhancement of Vehicle Speed Estimation with Single Loop Detectors Journal of the Transportation Research Board No 1870 TRB National Research Council Washington D C 2004 pp 14
51. fication with Double Inductive Loops Proceedings of the 17 ARRB conference Part 4 1994 pp 29 44 Ritchie S G Abdulhai B Parkany A E Sheu J B Cheu R L and Khan S I A comprehensive system for incident detection on freeways and arterials Proceedings of the 1995 Annual Meeting of ITS AMERICA pp 617 622 Rouphail N M Dutt N Estimating travel time distributions for signalized links model development and potential its applications Proceedings of the 1995 Annual Meeting of ITS AMERICA pp 623 632 Safaai Jazi Ahmad Siamak A Ardekani and Majid Mehdikhani A low cost fiber optic weigh in motion sensor SHRP ID UFR 90 002 Strategic Highway Research Program National Research Council Washington D C 1990 Saito M and R Patel Evaluating the efficacy of a microwave traffic sensor in New York City s freeway and street network Proceedings of the Transportation Congress of ASCE October 1995 pp 1952 1963 Saito Mitsuru and Thomas G Jin Evaluating the accuracy level of truck traffic data on state highways Prepared for Utah Department of Transportation research division UDOT 09 02 February 2009 Skszek Sherry L State of the Art Report on Non Traditional Traffic Counting Methods FHWA AZ 01 503 October 2001 Taylor Brian and Art Bergan The use of dual weights elements double thresh hold to improve accuracy on weigh in motion systems and the effect of accuracy on weigh station sorting
52. ge ignored the errors sign when averaged while the regular mean did not ignore the sign Also each hour contained 23 error points because there were 23 weekdays in the data set From the three figures the absolute errors did not vary substantively by hour However the peak hours had slightly larger absolute errors compared to the non peak hour errors This can 48 be explained by the TMS dependence on the loop detectors for accurate vehicle counts but may also be from other location dependent parameters Therefore the system was less precise during peak hours but not drastically Another observation from the figures was that the peak hour errors had significantly wider spreads compared to the non peak hour errors Also most of the absolute errors were within 25 error and the regular average errors were closer to 0 error still The non peak hours distributions were similar in terms of the spread and the multiple averages More data still needs to be collected to understand the reasons behind the error disparity between non peak hours and peak hours as well as why some hours had a narrow spread and others larger ones Note the term bin refers to the aggregated LV errors in a 10 range in Figs 24 26 and 28 respectively For example in Fig 24 Bin O contained the LV errors 5 to 5 Table 29 Station 037 NB March 2011 Station 037 NB Evaluation 4 Lower Bound LV Weekdays March 2011 Hourly Errors Daily Errors Daily Hourly Aver
53. gs with AVC and TMS 100 data It should be noted that since those 30 minute video recordings provide a very small sample size to be considered a global ground truth the ground truth used for this analysis was AVC data that were collected around the clock At the beginning we used the 30 minute video recoding to make sure the AVC data is accurate and robust and the data is coming from a very highly calibrated system that would later on be considered as the ground truth for our overall analysis Table 8 shows the following measured parameters Mets Ground truth short vehicle count measured through the video camera Meri Ground truth long vehicle count measured through the video camera Mss Number of vehicles that were correctly classified as a short vehicle Mu Number of vehicles that were correctly classified as a long vehicle Ms Number of short vehicles classified as a long vehicle because of a system error M s Number of long vehicles classified as a short vehicle because of a system error Mse Number of short vehicles erroneously counted as a long vehicle or not counted because of non system related external errors My Number of long vehicles erroneously counted as a short vehicle or not counted because of non system related external errors The external errors are caused by the following e Loop detector cross talk error e AVC class 15 error e Definition error 33 The following performance parameters were used to analyz
54. he TMS 100 RS 232 PORT STATUS LEDs Figure 2 TMS 100 Modem Front Shows both RS 232 Port and Status LEDs The TMS 100 s front panel is consisting of M Status LED s Reset Button Select Button RS 232 Port 1 RS 232 Serial port The TMS 100 Modem supports the following DB 9 pins PIN DB9 Description DCD Data Carrier Detect RX Receive Data TX Transmit Data Not Used Ground Not Used RTS Request To Send CTS Clear To Send Not Used W COIN OD OT AWIN Serial Port Baud rates The TMS 100 supports the following baud rate 1200 2400 4800 9600 19200 28800 38400 57600 and 115200 2 Status LED s The TMS 100 is equipped with Six LEDs on the front panel as described in the table below CASE Global Technologies 5 TMS 100 User Guide LED Description LED Description Off very week signal Power On when powered up RSSI Blinking adequate signal Solid very strong signal Blinks as data is transferred Mode 0 See TMS 100 Modes bss to serial port Mode 1 Appendice A GSMR On when is register on GSM network 3 Reset Button To reset the TMS 100 insert a straightened paperclip into the Reset hole to press the button Keep pressing and power cycle off and on the device Wait for at least 5 seconds to release the button Then wait for the device to finish booting 4 Select But
55. he aggregated TMS and AVC data had to be recollected The sensors were not recalibrated The combination of needing to redo the second phase and losing staffing because of the accident that occurred resulted in not being able to fulfill the second phase see 1 for pictures of the accident Instead of two weeks of aggregated data the team could only collect aggregated data for one to two days for each evaluation site However one month four weeks of data was collected later in March 2011 which was analyzed in the fourth phase The following sections describe the evaluation tasks 5 1 SITE SELECTION The original pilot program called for 26 TMS 100 installation locations with several locations having AVC stations Of these 26 sites 3 were selected because of their physical characteristics traffic characteristics and AVC availability e Station 098 eastbound along Route 60 Figure 13 e Station 037 northbound along Interstate 710 Figure 14 e Station 040 southbound along Interstate 710 Figure 15 5 1 1 PHYSICAL CHARACTERISTICS The sites were chosen based on whether they had working sensors provided good data offered large shoulders for safety and had good visibility Google Earth and the Street View feature were also used to observe the physical site 24 Google ENE Wn ve MOEN gt fer ogle a NE EN 25 5 1 2 TRAFFIC CHARACTERISTICS Previous studies have shown that the performance of vehicle classifi
56. he daily hourly error was comprised of an average of 15 hourly errors in a day Therefore the number of data points incorporated in the averages for the tables below is as followed 23 errors for the daily averages 23 errors for the daily hourly averages and 345 errors for the hourly averages 47 Tables 29 30 31 presented the computed averages standard deviations and peak averages using hourly daily daily hourly average errors for the 2 sites The hourly averages produced wider distributions with less precise averages than the previous two methods This suggested that hourly data contained too many deviant errors which were large but not outliers an inappropriate statistical method for longer studies From the tables the daily and daily hourly error averages produced roughly the same averages and distributions Therefore these methods would be a viable option for further study With smaller standard deviations multiple data collections spanning many months using the two different daily error statistics would obtain similar results without significant deviations barring unforeseen circumstances The averages for the SV data were small less than 10 in absolute terms and within Caltrans limits However all SV error data contained systematic error which is occurrence of some internal or external characteristic found in all the data shifting the observed mean from the actual mean and may even cause the distribution to be altered as well The
57. hicle length is calculated based on the duration of the loop presence signal Each vehicle is binned into one of two length classifications long vehicles and all other vehicles The user can adjust the length threshold Through experimentation during this project it was determined that a 50 foot threshold minimized misclassifications all vehicles calculated over 50 feet are classified as long vehicles Long vehicles typically are 60 65 feet and the vast majority of other vehicles are significantly under 50 feet so the 50 foot threshold provided a good enough gap The algorithm uses a collection interval to calculate the estimated speed The estimated speed is recalculated for each lane based on the supplied time interval The collection interval selected for this project was 1 minute For example if the speed in lane 3 is calculated at 58 MPH between 11 00 and 11 01 58 MPH is then used as the speed estimate for the length 12 calculations from 11 01 to 11 02 when the next speed estimate is calculated Longer or shorter time intervals can be used but there are tradeoffs to consider A 30 second interval provides a smaller sample set to estimate the speed but is more responsive to changes in traffic speed A 2 minute interval offers a larger sample set but is less responsive to speed changes It was noted that long vehicle count results are best when there is a steady flow of traffic Count measurements during heavy traffic over the induc
58. ibration on the systematic error locations to extract the accurate vehicle count Therefore most LV error data will become more accurate and precise as more weekday data is collected Looking into the data s variance by time of day it was shown that the peak hours were slightly more error prone than the rest of the day but the underlying cause was indeterminate Overall the TMS varied slightly from the time of day the data was collected The comparison results showed that TMS performance matches Caltrans expectations and the results were improved with longer evaluation times and will improve with longer intervals and further calibration of the AVC and TMS The TMS unit was adaptable to varying situations The installation sites chosen ranged from a simple three lane highway to a six lane highway with on ramps With the ability to adjust input parameters from a remote location the TMS unit can be calibrated with ease to adapt to these physical situations and other dynamic ones For instance if the flow of traffic is known to shift between seasons input parameters can be adjusted remotely to better reflect what is on the 55 road at that time With the AVC system adjacent to several TMS locations it is also possible to validate the changed parameters The unit s ability to be calibrated was matched by the success of its calibration as seen during the adjustments made between the four evaluation periods Adjusting the classification threshold an
59. ious options as to the GSM antenna depending on the signal strength In areas with powerful signal RSSI 60 a small antenna is adequate and can function will inside a traffic cabinet In areas with weaker signal RSSI 80 an antenna is external is mounted If the signal is really weak RSSI 95 a high gain antenna is required Case Global Technology resells compatible antennas 4 DC 12V Power Connector The TMS 100 modem can either receive TCP IP connections from a host computer or it can connect to a host computer to provide wireless connectivity for dialup only applications TMS 100 Modem receiving TCP IP connections from a host computer application TCP IP enabled application running on a host computer can connect to the TMS 100 Modem cabled to the serial device The serial device could be a utility meter a traffic counter changeable message sign an irrigation system or any other device with an RS 232 serial port The application using TMS 100 s IP address and port can connect either over the internet or a custom APN Once a connection is established data can be transferred between the application and the serial device via the TMS 100 Modem 3 i For information on Custom APN s go to att com or perform a Google search CASE Global Technologies 10 TMS 100 User Guide Configuration and Status By typing the DNS Cingular hostname lt phone number gt internet mycingular com of the hostname in the web
60. istribution After referring to the previous evaluation rounds of error analysis the fourth round of data collection produced normal LV frequency data centered near 0 error expect Station 37 SB a significant improvement in accuracy from previous rounds which had no apparent shape The range of the hourly errors decreased significantly from the previous evaluations a significant improvement in precision The two different daily errors had smaller average spreads still The SV errors and LV Station 037 SB errors contained systematic error shifting the mean error measured and possibly the spread as well but not outside of Caltrans range tolerances Overall the TMS system after further calibrations with a longer LV transition length was shown to provide accurate and reliable LV and SV counts over the course of a month Therefore the system was performing within its expected operation and further calibration would not be needed unless more precise data was needed 54 7 CONCLUSION While this limited deployment project had its challenges it also had many successes The Caltrans team was able to install 20 working TMS units in the field see 2 despite the constraints and challenges The TMS unit proved versatile and offered a solar solution when there were power issues at a site A series of unexpected circumstances in addition to funding and staffing constraints limited the evaluation to three sites The CCIT researchers and Caltrans perso
61. mated and electronic traffic count systems such as AVC and Weigh In Motion WIM is not extensive enough to provide the data necessary to address the issues around truck traffic Truck sensors and piezoelectric traffic counters have been installed at only a select number of sites because the equipment is too expensive for large scale deployment Because the sensors must be placed in the pavement installation is time consuming and disruptive requiring road closures that cause traffic delays and added congestion Additionally the complexity of the equipment requires substantial ongoing maintenance costs Radar detection alternatives have also proven to be costly and resulted in inaccurate data The core objective of this project was to evaluate whether the InfoTek Wizard system a less expensive and easier to install wireless technology could provide the same or better information on truck traffic The system collects data from existing loop detectors The post processed results are relayed to traffic management centers via a GSM wireless network or optional Ethernet or fiber network The device uses the Traffic Monitoring System Algorithm to classify vehicles If the new technology proves accurate it offers the following benefits e No construction costs e Ability to deploy one to two sites per day e Lower per unit costs e More data collection points e Option to use the existing infrastructure including installed single loops and cabinets
62. mestamp and location The TMS 300 runs on solar power so it can be used in areas where power is inconsistent or not available The TMS 300 was used at one of the evaluation sites On average the TMS 100 requires less than 1 watt of power Beyond energy savings TMS 300 is also a good alternative for new loop detector stations and road construction sites that require fast installations Like the TMS 100 it provides real time data using cellular wireless communication It can also be reset remotely TMS can be integrated with existing controllers and loop detectors The TMS cable is connected directly to the loop detectors and the GSM antenna mounts on top of the traffic cabinet for wireless connectivity The system can withstand temperatures ranging from 13 F to 162 F 25 C to 72 C and relative humidity from 5 to 95 TMS collects data via a wireless modem with sends it to a database located on a remote server The data can then be post processed at the remote site For single loop configurations TMS summarizes volume occupancy and long vehicle counts into 15 minute bins The system can be calibrated remotely and customized to send real time alert notifications such as changing traffic conditions or predefined thresholds 11 Traffic Cabinet Traffic Center Wireless Interne VPN GSM Data Frame Relay gt InfoTek Data Collector Output Cable ner Network Tunnel with TCP UDP IP Oracle Detector MS Access
63. mestamp of traffic in all lanes R 1 9 shows both the loop and vehicle reports one lane at a time This diagnostic is not available for lanes 10 14 The v command will still show these lanes RB Letter L Toggles on and off the loop report The v Command Example SLO HyperTerminal File Edit View Call Transfer Help D 83m gt v Vehicle Report true gt lane 3 72 MPH 35 ft 2 15 55 972 lane 72 MPH 13 ft 15 56 252 lane 64 MPH 34 ft 15 56 290 lane 77 MPH 15 ft 15 56 569 lane 64 MPH 17 ft 1557307 lane 72 MPH 15 ft 15 57 605 lane 59 MPH 18 ft 16 43 247 lane 77 MPH 15 ft 16 43 701 lane 67 MPH 17 ft 16 44 906 lane 67 MPH 55 ft 16 46 417 lane 63 MPH 15 ft 16 46 581 lane 63 MPH 10 ft 16 48 74 lane 77 MPH 56 ft 16 49 129 lane 67 MPH 13 ft 16 49 171 lane 77 MPH 15 ft 16 49 907 lane MPH 17 ft 16 50 654 lane MPH 63 ft 16 51 559 MPH 12 ft 16 52 9 2 4 1 2 3 2 1 2 4 2 4 3 4 1 1 2 ER NNNNNNNNNNNNNNNNN ANSI TCP IP Figure 12 v command example For single loop systems the length and speed are not available The loop reports displays loops as vehicles pass over them This is useful in diagnosing incorrect wiring or faulty loops CASE Global Technologies 19 TMS 100 User Guide The Single lane command In this example lane 4 is used to provide further analysis DOT HyperTerminal File Edit View Call Transfer Help Dae 55
64. ng calibration Next the accuracy of the system needs to be assessed by observing oncoming large vehicles with known lengths and comparing the observations with the per vehicle records PVRs from the TMS 100 unit via the HyperTerminal When a vehicle with a known length is seen downstream of the sensor check the PVR as the vehicle passes by If the vehicle length is consistent by type or class of vehicle no further calibration is required The process can be simplified by checking one lane at a time Two things to watch for are missing vehicles and cross talk an incident in which a vehicle triggers a count in the adjacent lane Missing vehicles can occur if the loop detector is too deep in the asphalt so that it might not detect a passing vehicle In this situation increase the loop detector s sensitivity However increasing the sensitivity might also increase the average detected vehicle length because the effective size of the detector might increase Cross talk can occur if the loop detector is too shallow in the asphalt or its sensitivity is set too high In this situation the sensitivity must be lowered However lowering the sensitivity might also decrease the average detected vehicle length because the effective size of the detector might decrease 20 SLO Hyper Terminal File Edit Yew Call Transfer Help Dae as DD E gt v Vehicle Report true 3 72 MPH 35 ft 2 15 55 972 72 MPH 13 ft 5 56 252 64 MPH 34 ft 56 290 7
65. ng to and from the ports of Los Angeles and Long Beach and 16 sites on State Route 60 which carries a significant amount of truck traffic into and out of the Los Angeles basin To evaluate the TMS 100 CCIT used video data as well as data from the existing AVC systems along I 710 and CA 60 as a comparison To use the AVC data as a ground truth baseline CCIT needed to first confirm that the AVC detectors were valid and accurate To assess both the AVC and TMS 100 data the deployment sites were videotaped to provide a one to one comparison This document describes the methodology and procedures involved in validating the AVC system and evaluating the TMS 100 data using both the AVC and video data The report also documents how to install and set up the TMS equipment and software Although the intent was to evaluate 26 sites because of unexpected circumstances and funding and time constraints the project could test only three sites for a limited period of time One problem was that the initial AVC data collected for the purpose of comparison was not usable because the sensors were not correctly calibrated Caltrans personnel needed to perform the calibration However because of work furloughs and a shortage in staffing there was not enough time to complete all the tasks for the original number of stations The collection time also needed to be shortened Then part way through the project the CCIT researchers and Caltrans staff were involved in a mul
66. nnel were involved in a serious multicar collision while collecting video data As a result the only video data collected was at Station 098 The video data analysis for Station 098 revealed that the loop detector sensors in that station were not calibrated so the data was not usable Three weeks of AVC and TMS data were collected after the incident as well as one month worth of data in 2011 The initial comparison of TMS and AVC ground truth data showed large values of error due to the poor calibration of TMS units However the team was able to complete two more rounds of comparing AVC and TMS data with the TMS parameters calibrated and fine tuned after each round The third round consisted of one day of collected data Because the TMS parameters were then calibrated properly for the site better data was collected The final round consisted of one month of collected data in early 2011 The hourly errors for LVs overlapped near 0 error with a narrower distribution than the previous evaluations an improvement in accuracy and precision The two daily errors produced similar average ranges as well as narrower distributions compared to the hourly errors However weekend errors varied more than weekday s errors and had wider distributions because of low LV counts and were ignored With more error data it was determined that SV and Station 037 SB LV counts incorporated systematic error while the other two LV data groups were normal requiring further cal
67. ns utilizing Global Standard for Mobility GSM infrastructure GPS Global Positioning System GSM Global System for Mobile Communications Standard for digital communications HyperTerminal HyperTerminal is a communications program bundled with multiple versions of the Microsoft Windows operating system It is a tool used when connecting to other computers bulletin board systems BBSs and a host of other Internet related services 8 Windows Vista does not include HyperTerminal CASE Global Technologies 23 TMS 100 User Guide LAN Local Area Network A collection of devices connected to enable communications between themselves on a single physical medium Modem Modulator Demodulator Modem is a device that converts digital signals into analog signals and vice versa Used for conversion of signals for transmission over telephone lines Network A number of devices connected to enable the device to communicate with any other device over a physical medium Packet A collection of data transmitted over a digital network in a burst SIM Subscriber Identity Module Telnet Telnet is a TCP IP application that enables a user to log in to a remote device TCP IP Transmission Control Protocol Internet Protocol VPN A type of technology designed to increase the security of information transferred over the Internet VPN can work with either wired or wireless networks as well as with dial up connections over POTS V
68. of trailing loop Values can range from 10 to 20 feet E Reporting mode allows the user to select between 15 minutes and 1 hour reporting bins The user can also specify the various UDP broadcasting intervals UDP is off by default E The number of lanes can range between 1 and 14 lanes CASE Global Technologies 14 TMS 100 User Guide LOOP STRING CONFIGURATION Single Loop The single loop algorithm estimates the speed and length of vehicles traveling over a single loop The single loop speed algorithm uses an estimation formula which requires the user to provide the average length of a vehicle traveling in the lane Since traffic in the inner lanes composes almost entirely of cars and small trucks vehicle lengths are much easier to predict For this reason the single loop requires the user to provide the average length of the vehicle in the inner lane Once the speed is calculated in the inner lane the user can scale the speed to the outer lanes by applying a speed factor Typically traffic speeds are 5 percent slower from the inner lane to the outer lanes The user can factor this speed difference on a per site basis Example Let s imagine a 3 Lane freeway we know that the average speed in the fast lane is around 75mph and since there are generally only a few small trucks in the fast lane the average length of the vehicles in this lane is 20ft Using this lane as a reference we assume that the speed factor is 1000 N
69. ote that the Speed factors are in 10 of a percent so 951 is 95 1 Keeping that in mind we can calculate the average speed in the second lane If the speed is approximately 5 lower in the second lane the speed factor can be calculated as 1000 1000 x 0 05 950 Following the same logic the speed factor In the 3 lane would be 1000 1000 x 0 1 900 CASE Global Technologies TMS 100 User Guide Lane configuration Single loop Y Number of lanes 3 Loop String 40 20 1 1000 1 950 1 900 Ex 40 20 1 1000 1 1000 Defines bin 0 40ft 41 feet 20ft Avg vehicle length Lane1 Ref Lane 1 Reset after Idle Minutes 30 Lane Factor 1000 Lane2 Ref Lane 1 Lane Factor 1000 Setting the best time Interval Speed is calculated in short time intervals say every 30 seconds 1 minute or 2 minutes The current speed of traffic calculated is based on the speed calculated from the previous time interval For example if the user wants TMS to calculate speeds every 1 minute vehicle speed and vehicle lengths at 12 31 20 are based on the speed of traffic from 12 30 to 12 31 If average traffic speed is more constant then the user may decide to use a 2 minute time interval If traffic speed is more dynamic the user may decide to use a 30 second time interval to calculate speed Being able to use a 2 minute time interval is more desirable since the average speed calculated is derived from a larger sample size However
70. plates 2 and piezoelectric sensors 10 Martin 2003 But based on the lifecycle cost the order is reversed piezoelectric sensors are the least costly at 4 424 per lane for a 12 year life followed by bending plates at 4 990 and load cells at 7 296 Fiber optics is becoming recognized as an alternative for replacing traditional WIM methods because of its lower cost higher accuracy absence of electromagnetic interference and ease of installment Safaai Jazi 1990 Martin 2003 Usually AVC combines several detectors such as two loop detectors with a piezo sensor between the loops or two piezo sensors with a loop detector in between A loop piezo loop LPL combination is commonly recommended According to a 48 hour test done by Harvey 1995 there was not a significant difference in accuracy between an LPL combination 76 6 and a PLP system 75 3 but piezo sensors cost more While a traditional dual loop system costs 1 510 annually Martin 2003 a PLP costs 4 424 Taylor 1993 Considering the time span between 2003 and 1993 when these estimations were gathered the cost gap between dual loop and PLP could even be larger A potential cost effective method is the use of wireless modems and programmable software to collect data and transmit it directly to a database The TMS 100 was developed in 2005 to provide a less expensive solution to detect long vehicles using single loop sensors The efficacy of using this single loop dete
71. portant The cable mapping is how lanes in the roadway are store collected and stored into the database Different agencies follow their one lane naming conventions Some may want to assign lane 1 to the fast lane on the freeway Others may want lane 1 to be the closest lane to the traffic cabinet Each TMS cable must be grounded It is OK to have unconnected TMS cable wires For example if there is a 3 lane dual loop configuration wires 7 and 8 will not be connected For more on how The TMS cable is connected to the output of the speed detector refer to CASE Global Technologies 13 TMS 100 User Guide Configuration and Status By typing the IP address of the hostname in the web browser a user can see the current status of the TMS application See browser page below To Modify Settings the user can configure the number of lanes dual or single loop trap distance in feet only applies to dual loop and listening port for telnet and data collector downloads and UDP broadcast interval Once the connection port and baud rate are set the Modify Settings button is pressed The new settings are stored and the TMS 100 will reboot before the new settings take effect This process takes about 30 seconds Modifying Setting E Listening port TCP IP listening port used by the telnet application that logs into TMS E Trap distance is set to the number of feet separating the speed trap center of leading loop and center
72. puter via the wireless GSM modem from the drop down menu select TCP 3 Enter the IP address of the TMS 100 unit 4 Ifthe network has a firewall the firewall port might need to be changed The port must be changed in the HyperTerminal TMS 100 unit If HyperTerminal is not installed 1 Inthe Control Panel select Add or Remove Programs Click Add Remove Windows Components Click Accessories and Utilities and then click Details Click Communications and then click Details Select HyperTerminal and then click OK N Pan 4 4 1 2 WEB BROWSER To access the TMS 100 system from a web browser enter the IP address of the particular TMS unit in the address bar The address is http lt IP address of the modem gt 4 4 2 CALIBRATING THE LOOP DETECTOR The TMS 100 unit relies on the input from the loop detector system The loop detectors must be calibrated to obtain quality data As a baseline set the loop detector sensitivity to mid level settings Detector cards vary from cabinet to cabinet and some cards might need adjustment For this limited deployment project Caltrans replaced the existing cabinet loop detector cards to reduce the possibility of errors It is recommended that technicians bring several new cards during installation in case problems arise or cards are broken 19 Figure 10 Example of a detector card Sensitivity varies between card manufacturers so consult the user manual of each card type duri
73. raffic data such as volume presence occupancy speed headway and gap and it has a lower lifecycle cost compared to some nonintrusive detectors Assuming a lifecycle of 15 years the average annual cost for one unit of a typical loop system including installation and maintenance is 1 510 Martin 2003 However a loop system also has also some limitations It causes disruptions during installation Martin 2003 Victoria 2004 is inconvenient to repair Carlson 1997 Victoria 2004 and is prone to pavement failures Carlson 1997 Victoria 2004 damage caused by snow removal equipment Luz 2000 Victoria 2004 and miscalculating axles as separate vehicles Carlson 1997 Pneumatic road tube detectors are the oldest and most widely used method for counting axles to classify vehicles U S Department of Transportation 1997 Skszek 2001 because of their simplicity lower cost Elena 2000 and ease to install However these detectors cannot count occupancy or presence and they can be severely influenced or damaged by heavy snow snow removal equipment rain and extreme temperatures They sometimes also miss low speed vehicles Martin 2003 and vary greatly in accuracy according to the circumstances Piezoelectric sensors bending plates load cells and capacitance mats are traditional WIMs They vary in accuracy with load cells being the most accurate 1 of standard deviation in the confidence interval followed by bending
74. rent errors and provided insight about the system CCIT used a Cannon HV20A video camera connected to a laptop to record directly to hard disk using the HDVSplit program This program records the video start time in the filename To ensure clear image quality for data analysis CCIT used a shutter speed of 1 500 Figure 20 28 Figure 20 Effects of shutter speed on image quality Left Shutter off Center 1 120 second shutter speed Right 1 500 second shutter speed Because the placement of the video camera was a slight distance away from the cabinets to synchronize the camera and PVR data the video camera operator and the AVC TMS 100 operator communicated via cell phone to begin the data collection simultaneously Starting both data recording processes simultaneously resulted in a low offset time The TMS PVR data is recorded via the HyperTerminal using the Transfer gt Capture Text option The video equipment did not have the capability to place timestamps within the recording To easily compare the video to the AVC and TMS PVR data which is marked with the actual time timestamps were periodically inserted into the video by filming the current time on a cellular phone Both the PVR system and video recording system had a slight lag between the start time and actual recording To synchronize the data an offset was calculated using the video timestamp and matching it to the vehicle combination For example if a sequence of vehicle
75. rmed with at least 95 accuracy compared to the existing AVC system This collaborative study between CCIT and Caltrans evaluated whether the TMS wireless technology could provide reliable real time data on a broader basis The evaluation focused on District 7 which encompasses the heavily trafficked areas around the Los Angeles and Long Beach ports Although TMS units were installed at 20 sites only 3 were used in the study due to unforeseen circumstances that shortened the evaluation period The TMS solution uses a wireless modem with programmable software It collects traffic information from the loop detectors and applies an algorithm to the data in real time The processed results are relayed to traffic management centers via a cellular network The software is compatible with several major database systems The company also offers a solar powered unit TMS 300 that can be used in locations that do not have accessible A C power The scope of the study involved installing the TMS 100 units customizing the TMS algorithm adjusting the calibration devising a system to use on site video taken simultaneously as ground truth data to determine accuracy and comparing the TMS data to the data of the existing AVC system To evaluate the TMS 100 the intent was to collect 60 minutes of TMS and AVC aggregated data over a period of two weeks and then compared those results However the available staffing and time were limited due to Caltrans budgetary
76. rthbound LV error frequency Table 17 Station 040 southbound hourly and daily errors Station 040 SB Evaluation 1 Hourly Errors Total SB SV SB LV SB Average 16 12 16 81 3 12 Stdev 4 52 4 85 15 96 Station 040 SB Evaluation 1 Daily Errors Total SB SV SB LV SB Average 14 39 15 02 2 75 Stdev 5 69 5 75 4 51 41 40 35 30 25 20 Frequency 15 10 L Station 040 Evaluation 1 SB LV Error Frequency T LD 9 60 E I E T iN o o e 7 Y mom mo IG 35 E ou y S O M ON oO mn a More Bin Figure 22 Station 040 southbound LV error frequency Table 18 Station 098 eastbound hourly and daily errors Station 098 Evaluation 1 Hourly Errors Total EB EB SV EB LV Average 1 68 0 18 84 61 Stdev 4 27 4 49 39 97 Station 098 Evaluation 1 Daily Errors Total EB EB SV EB LV Average 2 63 0 90 77 26 Stdev 3 43 3 31 11 31 42 Station 098 Evaluation 1 EB LV Error Frequency gt M E w 3 EP w en M Figure 23 Station 098 eastbound LV error frequency 6 3 2 2 SECOND ROUND OF AGGREGATED DATA COMPARISON TWO DAYS RESULTS As part of the second evaluation the TMS software parameters for lane by lane reference speed and length were adjusted to address the issues with over counting LVs This evaluation occurred over a period of two days
77. served only in the three lane system analysis where one lane would count an LV and the adjacent lane would count an SV milliseconds after The high values for the hit rate and false alarm rate for Station 098 see Table 14 indicate that the majority of the errors originated from the cross talk Because of the reduced Caltrans staffing it was not possible to do an on site adjustment to address the cross talk issue Table 15 shows the results of comparing the AVC PVR and video data 38 Table 13 Video data analysis results for Station 098 Station 098 Evaluation Total Lane 3 Lane 4 Lane 5 Total LV counted 177 51 103 23 Total LV identified as LV 132 Total SV identified as LV 45 Total cross talk LV 25 11 14 0 LV caused cross talk vehicles 23 SV caused cross talk vehicles 2 SV identified as LV 43 12 27 4 LV identified as LV 109 28 62 19 TMS missed LV 1 1 0 0 Total ground truth LV 110 29 62 19 Missing LV Lane 3 Class 8 Table 14 Comparing TMS PVR and video data TMS performance parameters for Station 098 TMS Performance Station 098 Total Lane 3 Lane 4 Lane 5 LV accuracy 61 55 60 83 Hit rate 99 96 100 100 Missed rate 1 3 0 0 False alarm rate 62 79 66 21 From cross talk 22 38 23 0 From definition error 39 40 43 21 Table 15 Comparison of AVC PVR data and video data Larg
78. th parameters The results are described in the following sections 6 3 1 TMS PVR DATA COMPARISON RESULTS Based on what we had learned from the video analysis performed during the AVC validation the TMS evaluation was more in depth in terms of identifying specific types of errors and analyzing the three truck lanes as a system in which each lane s sensor can interact with its adjacent lane Table 13 and Table 14 show the results of the video analysis for Station 098 The PVR analysis revealed two types of errors that were not associated with TMS 100 and the TMS algorithm 1 Definition errors Definition errors occurred in the process of converting the Caltrans 15 bin axle based vehicle classification collected from AVC to the TMS 2 bin length based format Many vehicles approaching the 40 foot threshold especially class 7 vehicles were often observed as being slightly longer than 40 feet and thus classified as LV This definition error occurred 45 times during the observation It was also observed that the AVC PVR data had the average vehicle lengths for LVs higher than the selected threshold of 40 feet Therefore the team hypothesized that adjusting the threshold value for LVs above 40 feet would reduce this error 2 Cross talk errors These errors were caused by LVS triggering a count in the adjacent lane This cross talk error was also seen in the AVC validation of Station 098 lane 3 Indications of this type of error were ob
79. ticar collision while collecting video data at one of the stations The team sustained both minor and major injuries causing the available staffing and resources at the coordinating Caltrans office to be further reduced As a result the CCIT team could only obtain video data for one station The long term evaluation comparing the TMS system to the ground truth AVC data continued but that happened at a decelerated pace and with several modifications Despite these obstacles the project was able to confirm the positive results of the initial pilot and demonstrate the performance accuracy of the TMS 100 The project also had the opportunity to install the TMS 300 which uses solar power because two of the sites selected did not have adequate power The TMS 300 offers an alternative for locations that do not have accessible A C power as well as for new loop detector stations and road construction sites that require fast installations 2 BACKGROUND The detection and classification of vehicles are critical for the efficient operation of streets and highways Harlow 2001 including signal and traffic control Rrouphail 1995 Klein 2001 Dillenburg 1995 ramp metering Kunigahalli 1995 Kell 1990 incident referencing and mapping Ritchie 1995 Klein 2001 Darter 2007 driver information system Nojima 1995 and road management and design Vehicle classification is especially critical for tracking trucks because truck traffic plays
80. tive loop can be accurate but long vehicle count calculations degrade as the traffic stops 4 3 TMS 100 INSTALLATION INSTRUCTIONS Setting up the TMS 100 includes installing the SIM card and software and creating a cable mapping guide to attach the wires These tasks can be performed off site The tasks involved at the site include installing the physical unit and then performing a communications check to verify the data link It is recommended that the TMS 100 be connected to the GSM network prior to field installation For more information about setting up and using the TMS 100 see the TMS 100 User Guide in 2 4 3 1 SIM CARD INSTALLATION INSTRUCTIONS 1 Remove the front and rear screws 2 Slide back the top cover Caution Do not touch the exposed processor 3 Slide down the SIM card holder 4 Make sure that the jumpers are set properly 5 Break the SIM chip from the plastic card Caution Do not touch the metal contact of the SIM card 6 Insert the SIM card and lock it into the holder 7 Reassemble the unit 8 Tag the TMS 100 with the APN address telephone IP address and location 9 Power on the TMS 100 13 LIII Z LE Je Om MEF LM H Figure 2 Inside view of a TMS 100 Unit 4 3 2 A 8 9 SOFTWARE INSTALLATION INSTRUCTIONS Connect the computer to the TMS 100 with the serial cable Use a pin to select MODE 1 on the TMS 100 Run the Nokia Configurator software on the computer Browse
81. to select the software Click gt to upload the software Select Main and click Select Set up the indicator lights as follows Click M2M System Click System Parameters Select GPRS always ON Select Connection 1 Select TCP GPRS Enter the GPRS access point Enter the APN name for example ct7mobile2 dot ca gov For the gateway port enter 12345 Click Read Parameters Close the window Exit and save the parameter settings Disconnect the serial cables and power cables Connect the power cable again The system registration might take a few moments Fro sansnenopv 10 Access the box through the HyperTerminal 14 4 3 3 CABLE MAPPING GUIDE Before installing the device in the cabinet a cable mapping guide is needed to show how each TMS 100 wire connection maps to a detector loop see Figure 3 Creating the cable mapping guide prior to installation would help streamline the installation process Station XX XX Direction TMS 100 Cables Lanes Channels 1 8 Brown Lane 1 E B White Lane 2 E B Violet Lane 3 E B Blue Lane 4 E B Green Lane 5 W B Yellow Lane 6 W B Orange Lane 7 W B Red Lane 8 W B Figure 3 Sample cable mapping guide The cable mapping guide corresponds to the lane nomenclature at the specific site Within Caltrans District 7 the Traffic Management group and the Census group use different lane naming schemes The Census group designates the
82. ton TMS 100 should be preconfigured to startup in the correct mode depending on the application In case the user needs to change the TMS 100 s mode read on The TMS 100 can be configured to run in 4 different modes The mode is selected either by pressing the select button in the front of the modem or by setting the default mode using a jumper on the circuit board inside the TMS 100 Also the select button left of the status LED s allows the user to toggle between modes A paper clip or other pointy object is required CASE Global Technologies 6 TMS 100 User Guide TMS 100 Modes TMS Mode Mode 0 OFF Mode 1 OFF M Required by TMS B TheRS 232 Port is not active Figure 3 TMS Mode TMS 100 Modem Mode Mode 0 ON Mode 1 OFF M Java serial port mode B Used when running TMS 100 Modem application l o s 9 gem d InfoTek ara 00r Figure 4 RS 232 TMS 100 Modem CASE Global Technologies 7 TMS 100 User Guide Configuration Mode Mode 0 OFF Mode 1 ON This mode is used for storing the factory setting of the modem This is done before shipping the modem from Case Global Technology Factory settings include E Loading Java firmware firmware can also be loading remotely over TCP IP E Configuring APN s Figure 5 Configuration Mode for Loading Java AT Mode Mode 0 ON Mode 1 ON M Sets up modem for AT modem M This is used for dialup PPP
83. usted to reflect the truck traffic During calibration the system needs about a one minute settling period to accept the adjustments and gather new data 21 4 5 SOLAR TMS 300 INSTALLATION Station 041 one of the two sites experiencing power issues was determined to be a good candidate to evaluate the solar powered TMS 300 Prior to installation the location of the solar panel must be coordinated with Caltrans For the solar panel to charge it cannot be obstructed by trees or buildings The solar panel is set at 196 degrees facing southwest at a 45 degree tilt based on the location s latitude Another consideration is the location of the loops and the pull box to determine the length of the conduit to connect the loops to the TMS 300 and terminate the loops at the NEMA enclosure After the location is selected Dig Alert also known as Underground Service Alert USA must be informed so that they can inspect the area for underground utilities and other existing infrastructure The USA permit takes about three days Either hand or machine digging can be used depending on the regulations and agencies involved such as Caltrans or the state environmental agency Hand digging has easier permit requirements The approximate field installation times according to CASE Global Technologies are e Drilling enclosures 45 minutes e Drilling and welding fitting to foundation and drilling poles 45 minutes e Assembling solar panel includ
84. vehicles that are under 40 feet and bin 2 indicates vehicles that are longer than 40 AVC uses the Caltrans 15 bin vehicle classification system Classes 1 14 are assigned according to the vehicle length and number of axles These classes are divided into two main categories short vehicle SV and long vehicle LV Class 15 is reserved for when a vehicle cannot be classified For instance under certain circumstances such as when vehicles are changing lanes driving offset from the center of the lane or have trailers the AVC system cannot classify the vehicle Table 2 shows how the AVC data was converted to the TMS two bin classification system Table 2 Converting Caltrans vehicle classifications to TMS bins Lower Upper Length Vehicle Length Bound Bin Vehicles Included in the Class Bound lt lt or Class Equivalent 1 Passenger Vehicles O ft 13 ft 1 and 2 1 Single unit Trucks 13 ft 40 ft 3 4 5 6 and 7 2 Combination Trucks 40 ft 61 ft 8 9 and 10 2 Multi trailer Trucks 61 ft 120 ft 11 12 13 and 14 5 3 1 TMS 100 AND AVC DATA COLLECTION AND PROCESSING Two types of data were collected from the AVC and TMS systems e PVR data time stamped lane by lane as shown in Table 3 and Table 4 e Aggregated data time stamped lane by lane as shown in Table 5 and Table 6 30 To analyze the data it was first reorganized and excess information was removed The vehicles wer
85. yzed Caltrans had reported that the AVC system would perform with accuracies above 95 Station 037 a three lane highway had better results than station 098 a five lane highway with more complex lane interaction The results show that the accuracy of the AVC system varies by lane with an overall accuracy of 90 or above The AVC system tends to undercount vehicles when compared to the video data Table 9 AVC validation data Station 037 northbound lane 2 Station 037 NB Lane 2 SV LV Errors Class 15 Cross talk Missing Ground SV 675 652 0 23 13 0 10 Truth LV 57 0 51 6 4 2 0 System Accuracy 96 LV Hit Rate 89 LV Miss Rate 11 LV False Alarm 0 For Station 037 lane 3 Table 10 the AVC counted 375 SVs but only 366 SVs were actually observed This over count occurred primarily when an LV neared the adjacent lane and its sensor mistakenly counted an SV 35 Table 10 AVC validation data Station 037 northbound lane 3 Station 037 NB Lane 3 sv LV Errors Class 15 Cross talk Missing Ground SV 366 365 0 1 1 0 0 Truth LV 157 4 153 0 0 0 0 System Accuracy 99 LV Hit Rate 97 LV Miss Rate 2 LV False Alarm 0 Table 11 AVC validation data Station 098 eastbound lane 5 Station 098 EB Lane 5 sv LV Errors Class 15 Cross talk Missing Ground SV 303 291 0 12 3 3 6

Download Pdf Manuals

Related Search

Related Contents

平成23年度採用の市職員を募集  Philips InRange Bluetooth smart leash AEA1000  KDC-X991 - Kenwood  GEBRUIKERSHANDLEIDING USER MANUAL  DIGITAL Server 1200 Series Installation Guide  Sony Compact stereo PC headset  取扱説明書 - 三菱電機  INSTALLATION & SERVICE MANUAL  帆 取扱説明書 (ァイベッ ドライ ト>  Printed Documentation  

Copyright © All rights reserved. Failed to retrieve file