Home

UWB RTLS for Construction Equipment Localization

Contents

1. 6 Figure 2 3 Time Difference of Arrival Technique adapted from Ghavami et al 2004 7 Figure 2 4 Schematic Diagram of UWB Systems adapted from Zhang et al 2012b 7 Figure 2 5 Relationship between Offset Precision and Accuracy Maalek amp Sadeghpour Figure 2 6 Comparison of Cumulative Accuracy Curves for AOA vs TDOA amp AOA Maalek KSP DN 10 Figure 2 7 UWB sensors positions Al to A6 Saidi et al 2011 sse 11 Figure 2 8 UWB tags mounted at different heights nominally 1 m 2 m and 3 m Saidi et al pup 12 Figure 2 9 UWB Tracking in Lay Down Yard Saidi et al 2011 sess 13 Figure 2 10 Area Layout for Static Test in Fully Furnished Office Cho et al 2010 15 Figure 2 11 Results of Open Space Dynamic Tests Cho et al 2010 sss 16 Figure 2 12 Predetermined Five Paths for Closed Space Dynamic Test Cho et al 2010 17 Figure 2 13 Illustration of Correction Process Vahdatikhaki amp Hammad 2014 20 Figure 2 14 Flowchart of Iterative Correction Process Vahdatikhaki amp Hammad 2014 21 Figure 2 15 Flowchart of Optimization based Method Vahdatikhaki et al 2014 22 Figure 2 16 The development model Opitz et al 2004 sse 24 F
2. Select Al Select All Logging Manual Start Stop Log forthe interval of Seconds le Ready Figure A8 Outdoor Testing Configuration The general idea is if you know two points in the space you can easily determine an unknown point So we have to measure coordinates of two points w r t origin But here we will assume two sensors as two points and one of these two sensors points would be considered as origin see Figure A10 Note If you are facing issues during this configuration refer to the video at YouTube named How to use the Survey Point Finder provided by Ubisense Link https www youtube com watch v T7p0IBZtIGI 1 Position the sensors 2 Measure the height of the sensors from the ground and save in the calibration sheet see Table Al 3 Take any sensor as origin 0 0 0 and define x and y axes Let s call this sensor Point P1 4 Align another adjacent sensor such that it s either x or y will be zero let s call this sensor Point P2 5 Measure the distance of all sensors from the two points which are P1 and P2 see Figure A10 Remember that distance should be measured from the fudicial mark on the sensors Save all the distances in the calibration sheet 135 10 Open location engine and load area from the Map tab Go to the Map tab again and then go to Add Survey Points a dialog box would be opened as shown in Figure A9 Enter the coordinates of P1 a
3. esee eeeeeee eene enne tn eene nennen tense 9 Figure 4 2 Proposed Approach 1 serere lee er ettet eta tene Tree Ry eed eeeta 93 Figure 4 3 Visual IDs for Equipment Identification eese 95 Figure 4 4 Association Error Explanation seseserrensvensesnersnrsersesrennsesnessessnrsersesresnsensssversnssenssere 96 ix Figure 4 5 Design of Experiment for MSDF Case Study seen 97 Figure 4 0 Detection Results eue netu bota d ti dae NU EE 100 Figure 4 7 Pixels Conversion gris cau ted tu enu eu asit amet es 103 Figure 4 8 Measurement of Attributes of Field of View adapted from MATLAB 2014a 104 Figure 4 9 Coordinate Transformation Using First Method eee 105 Figure 4 10 Control Points ies danseren dikes cmi etae tae rains Renee na I eaea 106 Figure 4 11 Coordinate Transformation using Second Method sss 107 Figure 4 12 First 6 Frames for Data Fusion Case 1 with First Transformation Method 111 Figure 4 13 Last 5 Frames for Data Fusion Case 1 with First Transformation Method 112 Figure 4 14 Incorrect Association during Fusion Process with First Transformation Method 113 Figure 4 15 First 6 Frames for Data Fusion Case 1 with Second Transformation Method 114 Figure 4 16 Last 5 Frames for Data Fusion Case 1 with Second Transformation Method 115 Figure 4 17 Corr
4. 1 end 143 APPENDIX C DATA FUSION MATLAB CODE Coordinate Conversion Code Method 1 Construct an imref2d object given a knowledge of world limits and image size A imread Background3 rotated jpg xWorldLimits 1 3 5 yWorldLimits 5 7 0 RA imref2d size A xWorldLimits yWorldLimits xlfilename C Users umroot Documents Hassaan DataFusion DetectionResults July21 xlsx sheet TruckDetected2 xldata xlsread xlfilename sheet xldata for i 1 1length xldata xlIntrinsic xldata i 2 ylIntrinsic xldata i 3 xWorld yWorld intrinsicToWorld RA xIntrinsic yIntrinsic WorldCoordinates i xWorld abs yWorld ActualCoordinates 2474245 Error i dist WorldCoordinates i ActualCoordinates end Data Association and Position Estimation Code Q data fusion data association and position estimation UWBTruckDataPath C Users umroot Documents Hassaan DataFusion Case Study for Thesis DataForFusion UWBTruck DataForFusion July27 xlsm UWBExcavatorDataPath C Users umroot Documents Hassaan DataFusion Case Study for Thesis DataForFusion UWBExcavator DataForFusion July27 xlsm VideoTruckDataPath C Users umroot Documents Hassaan DataFusion Case Study for Thesis DataForFusion VideoTruck DataForFusion July27 xlsm VideoExcavatorDataPath C Users umroot Documents Hassaan DataFusion Case Study for Thesis DataForFusion VideoExcavator DataForFusion July2
5. Langford G O 2012 Engineering Systems Integration CRC Press Langley R B 1999 Dilution of Precision GPS World May 1999 pp 52 59 Lee W Liu W Chong P H Tay B L amp Leong W 2009 Design of Applications on Ultra Wideband Real Time Locating System 2009 IEEE ASME International Conference on Advanced Intelligent Mechatronics Lundquist C 2011 Sensor Fusion for Automotive Applications Thesis Link ping University Department of Electrical Engineering Link ping Luo R C Yih C C amp Su K L 2002 Multisensor Fusion and Integration Approaches Applications and Future Research Directions EEE Sensors Journal 2 2 107 119 Luo X O Brien W J amp Leite F 2013 Evaluating the Impact of Location Aware Sensor Data Imperfections on Autonomous Jobsite Safety Monitoring Computing in Civil Engineering 573 580 Maalek R amp Sadeghpour F 2013 Accuracy assessment of Ultra Wide Band technology in tracking static resources in indoor construction scenarios Automation in Construction 30 170 183 123 Mahfouz M Zhang C Merkl B Kuhn M amp Fathy A 2008 Investigation of High Accuracy Indoor 3 D Positioning Using UWB Technology Microwave Theory and Techniques IEEE Transactions on 56 6 1316 1330 Malik A 2009 RTLS for Dummies Wiley Massimiliano D Antonio B Gian M Daniele M Carlo M amp Fabio R 2011 Fusion of Radio and V
6. Metallic Liquid Humans Line of Sight between sensors and tags 36 Table 3 2 Effect of Number of Sensors on the UWB System adapted from Zhang 2010 Location method Number of sensors Extra information Result detecting tag required 2D horizontal position AOA 1 Known height of tag known height AOA 2 or more None 3D position TDOA AOA 2 or more None SD poor highest accuracy 3 3 Experimental Work The performance of the UWB system for construction management is evaluated by conducting three indoor and three outdoor tests Furthermore the performances of the wired and the wireless UWB systems are also compared by conducting two sets of indoor tests These indoor and outdoor tests are summarized in Table 3 3 and discussed in detail in the following sections Table 3 3 Overview of Experimental Work Location UWB System Purpose Data Processing EUR Hz Wireless Pon of No 16 Tags Performance udo Wired and Performance SCM 34 Wireless Comparison Wired and Performance Wireless Comparison pe Aree iud Wireless Marne Real N A 8 3 Excavator Outdoor Wireless Tracking Real scM and OM 8 3 Roller i Tracking Real Wireless Excavator Simple Averaging 4 19 37 3 3 1 Indoor Dynamic Tests 3 3 1 1 Indoor Wireless Dynamic Test to Evaluate Performance of Tags Design of Experiment This test was conducted at the atrium of
7. 1 0 1 2 3 a Period 1 b Period 2 Figure 3 26 Results of Simplified Correction Method S3 amp S4 Average for At 1 sec To further improve the location accuracy the OM was applied to the data The results of OM are shown in Figure 3 27 By comparing Figure 3 25 and Figure 3 27 it can be observed that the OM has not significantly changed the path obtained from the UWB system Similarly by comparing Figure 3 26 with Figure 3 27 it can be concluded that if we use two tags data then SCM produces much better results But for OM two tags are not enough as this method is useful when there are a large number of GCs therefore OM has not been applied with S3 and S4 tags 68 12 11 10 9 8 7 6 5 4 3 2 1 0 1 2 3 12 11 10 9 8 7 6 5 4 3 2 1 0 1 2 3 a Period 1 b Period 2 Figure 3 27 Results of Optimization based Method All Tags Averaged for At 3 sec 3 3 2 3 Full scale outdoor dynamic test to investigate performance of wireless UWB system on real construction site This test was conducted at a construction site in Downtown Vancouver At the site earthmoving operation was carried out with the help of two excavators Design of Experiment This experiment was designed to localize and track one excavator within the site area The total area of the site was about 36 5 m x 24 m which was surrounded by walls on two sides and by fences on the other two sides as shown in Figure 3 28 This picture was provided
8. 2008 Intelligent Monitoring Software User s Guide Sony Corporation Sony 2012 Network Camera Application Guide Sony Corporation Teizer J amp Vela P 2009 Personnel tracking on construction sites using video cameras Advanced Engineering Informatics 23 452 462 Ubisense 20132 Location Engine Configuration User Manual Ubisense Ubisense 2013b Site Manager Manual Vahdatikhaki F amp Hammad A 2014 Framework for Near Real Time Simulation of Earthmoving Projects using Location Tracking Technologies Automation in Construction 42 50 67 125 Vahdatikhaki F Hammad A amp Siddiqui H 2014 Optimization based Excavator Pose Estimation Using Real time Location Systems Automation in Construction Submitted Welch T Musselman R Emessiene B Gift P Choudhury D Cassadine D amp Yano S 2002 The effects of the human body on UWB signal propagation in an indoor environment EEE Journal on Selected Areas in Communications 20 9 1778 1782 Zeng Y Zhang J amp Genderen J 2006 Comparison and analysis of remote sensing data fusion techniques at feature and decision levels n Proceedings of the ISPRS Commission VII Symposium Enschede ISPRS Zhang C 2010 Improving Crane Safety By Agent Based Dynamic Motion Planning Using UWB Real Time Location System Thesis Concordia University Department of Building Civil and Environmental Engineering Montreal Zhan
9. GCS GPS GUI IEC IF IP IPC Two Dimensional Three Dimensional Angle of Arrival Application Programming Interface Actual Update Rate Building Information Modeling Computer Aided Design Design of Experiment Dilution of Precision Easting and Northing Coordinate System Equipment of Interest Expected Update Rate Fusion Module Field of View Frames per second Geometric Constraint Global Coordinate System Global Positioning System Graphical User Interface International Electrotechnical Commission Information Filter Internet Protocol Image Processing Component xiii ISO k NN LAN LCS LoS MDR MRM MSDF OC PTZ RC RFID RTLS SCS SIF TDoA TL UCS UM UWB VCS VM WLAN International Standards Organization k Nearest Neighbors Local Area Network Local Coordinate System Line of Sight Missing Data Rate Minimum Reset Measurements Multi Sensor Data Fusion Operational Constraint Pan Tilt Zoom Remote Controlled Radio Frequency Radio Frequency Identification Real Time Location System Smart Construction Site Static Information Filtering Time Difference of Arrival Tolerance Limit Ultra Wideband Coordinate System Ultra Wideband Module Ultra Wideband Video Coordinate System Video Module Wireless Local Area Network XIV CHAPTER1 INTRODUCTION 1 4 General Review Real time information is the essence of smart decision making In construction operations real time information about
10. o o EN UA e bo 0 10 00 60 55 50 45 40 35 30 25 20 15 10 5 0 5 10 15 20 25 30 35 40 45 50 55 60 Error Range Degrees Figure 3 39 Error Distribution for Accuracy Assessment Period 1 81 Second Period Analysis During this three minute period the excavator moved a piece of pipe from one place to another During the first minute the excavator moved forward and then waited there while a worker attached the pipe to its boom while during the second minute the excavator moved backward and then swung its boom by almost 180 Finally during the third minute the excavator was stationary while a worker was removing the pipe from its boom Initially the AUR and MDR of each tag are analyzed and presented in Table 3 19 It can be observed that for some tags the AURs are very low compared with the EUR whereas for some tags the AUR is less but satisfactory Moreover out of 10 tags the MDR for 5 tags is more than 80 and for these tags the AUR is less than 1 Hz However for the remaining 5 tags Tag 1 2 3 4 and 5 the AUR is more than 1 Hz and the MDR is also acceptable The best performance is of Tag 4 with an AUR of 1 19 Hz and an MDR of 71 65 For further analysis the data of five tags with satisfactory performance in terms of AUR and MDR are considered Table 3 19 AUR amp MDR Analysis for Period 2 Tag AUR Hz MDR 1 1 07 74 56 2 1 15 72 71 3
11. vl Ceiling 5 Floor 0 Max standard error 0 05 Note Once you are done using the UWB System set the RF Power to 0 Measure incident power and set the activity thresholds Firstly ensure that either your tags are all powered off or are far away from the cell This is because you are attempting to measure the power of the background radio noise rather than the power of the tag signal Set the Disable Radio flag on the master sensor of the cell Right click on the cell and select Incident Power Plot Leave the power plot until the Set Thresholds button becomes enabled it requires almost 1000 readings each reading requires 2 or 3 time slots so you actually have to wait for 2000 to 3000 time slots and the thresholds that will be set should be marked on the axis of the cumulative plots Press the Set Thresholds button Note Once the activity thresholds have been set be sure to uncheck the Disable Radio flag Disable sleep mode by right clicking on the Master Cell go to Properties Flags and check the Disable Sleep flag Note Once the calibration is completed uncheck the Disable Sleep flag Place a working tag at a known location and measure its coordinate w r t origin point Make sure the tag should be easily visible i e line of sight to all sensors and there should not be any type of distorting item s e g human body metallic item in between the tag and the sensors Note If you are using system in an outdoor envi
12. 09 54 The raw data from these two tests were plotted as shown in Figure 3 14 From this figure it is clear that the movement of the tag on the slope is more realistically captured by the wired system whereas for the case of the wireless system the data are too noisy 4 5 4 5 4 4 3 5 315 1 5 0 5 2 5 4 5 1 5 0 5 2 5 4 5 a Wired b Wireless Figure 3 14 Wired and Wireless UWB System II Slope Raw Data 3 3 2 Outdoor Dynamic Tests 3 3 2 1 Outdoor dynamic test for tracking movement of an excavator Design of Experiment This test was conducted on the intersection of two busy streets i e Saint Catherine and Guy in Downtown Montreal This test was designed to 1 evaluate the performance of the wireless UWB system in outdoor environment and 2 investigate whether tags are easily attachable to construction equipment using magnets Each UWB sensor was connected to the wireless bridges and they both were installed on a tripod as shown in Figure 3 15 Compact tags were prepared for attaching them to equipment by adding two magnets with each tag as shown in Figure 3 16 These tags were then attached to the excavator as shown in Figure 3 17 Results Although the system was configured properly this test was not successful because the UWB system was unable to detect the tags It was found that the connectivity between sensors was intermittent Sometimes the sensors were connected and sometimes they wer
13. 1 14 72 84 4 1 19 71 65 5 1 07 74 43 6 0 14 96 82 7 0 76 82 11 8 0 23 94 57 9 0 83 80 26 10 0 29 93 24 After the AUR and MDR analysis the movement of the excavator during the three minute period tracked by the wireless UWB system was analyzed as shown in Figure 3 40 For this 82 analysis one location was extracted from the data of the five tags with an AUR of more than 1 Hz Tag 1 2 3 4 and 5 by first averaging each tag s data over a period of 1 sec and then averaging all five tags data From Figure 3 40 the working area of the excavator can clearly be identified and it can also be observed that it was not stationary Tracked Movement UWB Covered Area T T T T gt x 37 47 57 67 106 116 126 7 y Figure 3 40 Tracked Movement of Excavator for Period 2 Furthermore the orientation of the excavator estimated by the wireless UWB system was analyzed In order to analyze the orientation the data from the three tags Tag 1 2 and 3 were processed As for each of these tags the AUR is more than 1 Hz so each tag s data were averaged over a period of 1 second This processing resulted in three different data points for each second which are the positions for 1 Tag 1 pz 2 Tag 2 pz and 3 Tag 3 p3 The expected orientation based upon these three positions is shown in Figure 3 41 In order to estimate the orientation of the excavator a scatter plot for these data point
14. 17 12 N A N A 8 58 8 55 N A N A Tag 2 16 84 15 69 22 89 12 88 10 63 8 82 23 56 10 51 Tag 3 14 18 18 01 12 25 20 09 9 71 8 60 3 67 10 57 Averaged 16 66 16 98 16 69 16 04 8 47 8 58 11 34 10 31 49 2A lA 3 3 2 2 2B 1B 1 1 0 0 0 1 2 3 4 0 1 2 3 4 a Wired UWB System b Wireless UWB System Figure 3 10 Investigation of Impact of Dilution of Precision Phenomenon 3 3 1 3 Indoor dynamic tests to compare performance of Wired and Wireless Systems ll Design of Experiment Two sets of tests were designed to compare the performance of the wired UWB system with the wireless UWB system Within each set the DoE for both tests was kept the same in order to simplify the performance comparison of both systems These sets of tests were conducted in the atrium of the 8 floor of the EV building in Concordia University s downtown campus In each set one test was conducted with the wired UWB system whereas the other test was conducted with the wireless UWB system The UWB covered area was 7 7 m x 5 7 m as shown in Figure 3 11 In the first set of tests the movement of a person was tracked who was carrying two compact tags and was following a pre defined path shown as A in Figure 3 11 whereas in the second set of tests the movement of a compact tag was tracked which was moved on a pre defined inclined straight line sloping rope shown as B in Figure 3 11 where one end of the rope was about 2 meters ab
15. 2 4 4 Simplified Correction Method Vahdatikhaki amp Hammad 2014 proposed a method to reduce the measurement errors in which sensory data is captured from the construction site and processed by the data processor This method focuses on adjusting the data according to the GCs and OCs in which it is iteratively validated that a set of GCs and OCs are satisfied for each data point The assumption of this method is that several UWB tags are installed on different parts of different pieces of the equipment and each tag has a unique ID Vahdatikhaki amp Hammad 2014 described that this method is implemented using the following steps 1 The UWB tags are grouped according to their geometric relationships with respect to the equipment to which they are attached 2 each tag s data are averaged over a short period of time At 3 if there are any missing data it will be calculated using interpolation 4 the data are corrected based upon the operational constraints and the geometric constraints and 5 several tag s data are averaged Vahdatikhaki amp Hammad 2014 further explained these steps as see Figure 2 13 in step 2 averaging over time refers to averaging a tag s location over several points in time in step 3 new data is created for the missing data points using interpolation in step 4 data correction refers to the adjustment of the tags data iteratively to ensure that a set of OCs and GCs are satisfied at every give
16. 3 9 Table 3 7 and Table 3 9 it is concluded that for indoor applications the wired system performs much better than the wireless system and the SCM has the potential to improve the accuracy of the data collected from the wireless UWB system 47 3 5 3 5 3 3 25 2 5 1 5 2 2 5 3 1 5 2 2 5 3 a 1A Averaged b 2A Averaged 4 4 3 5 3 9 3 3 2 5 2 5 1 5 2 2 5 3 1 5 2 2 5 3 c 1A SCM d 2A SCM Figure 3 9 Performance Comparison of Wired UWB System and Wireless UWB System Evaluation of Impact of Dilution of Precision For evaluating the impact of DoP the data collected from tests 1B and 2B are analyzed For data analysis each tag s data were firstly averaged over a period of 500 msec and then the data of the three tags were averaged In case of the wireless test 1 e 2B only data from Tag 1 and Tag 2 were averaged as the MDR of Tag 3 is almost 88 see Table 3 7 The tracked rotational movement of the boom of the RC crane is shown in Figure 3 10 Figure 3 10 a shows the data 48 from tests 1A and 1B where Figure 3 10 b shows the data from tests 2A and 2B It can be observed that the in tests 1A and 2A the RC crane was in the middle of the UWB covered area corresponding to Position A in Figure 3 7 whereas the in tests 1B and 2B the RC crane was near the edge of the UWB covered area corresponding to Position B in Figure 3 7 From Figure 3 10 a it can be observed that in both positions the performance of the
17. 75 80 85 90 95 Error Range Degrees Figure 3 44 Error Distribution for Accuracy Assessment Period 2 86 3 4 Summary Conclusions and Recommendations The analysis presented in this chapter evaluates the performance of the UWB system in indoor and outdoor dynamic conditions and also compares the performance of the wired and the wireless UWB system in indoor dynamic conditions Another focus of this study was to use several tags to track a single object and later on all tags data would be combined to calculate the pose of the tracked object This approach enhances the data and smoothens the tracking of movement of the tagged object This data enhancement approach can be implemented either by simple averaging or by adjusting the data according to the GCs and OCs however along with the number of tags this approach is also dependent on the MDR Furthermore this analysis also highlights the trade off associated with the selection of the mode of the UWB system i e wired or wireless In terms of accuracy of the estimated tag locations the wired mode yield better results than the wireless mode whereas in terms of spatial disruptions on the monitored area the wireless mode is preferred as it imposes minimal spatial disruption because of less required cabling Moreover it is concluded that the factors that affect the performance of the UWB system should be considered during the design phase of the experiment For example suitable
18. GCs were calculated based on the specifications of the roller provided by the manufacturer Figure 3 25 shows the data resulting from the aforementioned process By comparing the smoothness of the tracked path of the roller for both 66 periods shown in Figure 3 25 and Figure 3 23 it can be observed that the SCM has improved the accuracy to a very little extent 12 11 10 9 8 7 6 5 4 3 2 1 0 12 3 12 11 10 9 8 7 6 5 4 3 2 1 0 1 2 3 a Period 1 b Period 2 Figure 3 23 All Tags Averaged with At 3 sec In order to investigate the impact of the low quality tags i e S1 and S2 another analysis was performed after eliminating these two tags and applying the aforementioned SCM In this case At was set to 1 s as these two tags have low MDR One OC and only one GC i e the distance between S3 and S4 were used Figure 3 26 shows the results of the above mentioned process By comparing Figure 3 25 and Figure 3 26 it can be observed that this process yielded much better results and the actual pattern of movement of the roller is visible S4 Figure 3 24 Geometric Constraints for Outdoor Dynamic Test 67 12 11 10 9 8 7 6 5 4 3 2 1 0 12 3 12 11 10 9 8 7 6 5 4 3 2 1 0 1 2 3 a Period 1 b Period 2 Figure 3 25 Results of Simplified Correction Method All Tags Averaged for At 3 sec 12 11 10 9 8 7 6 5 4 3 2 1 0 1 2 3 12 11 10 9 8 7 6 5 4 3 2
19. IP Address 10 133 0 1 Dlink AP 3520 WDS AP IP 192 168 0 50 MAC FA c a Dlink AP 3520 WDS Dlink AP 3520 WDS Dlink AP 3520 WDS Dlink AP 3520 WDS IP 192 168 0 51 IP 192 168 0 52 IP 192 168 0 53 IP 192 168 0 54 MAC 78 amp MAG 6A Ss MAC 5A L MAC 82 Sensor Sensor2 Sensor3 Sensor4 IP 10 133 0 240 IP IP 10 133 0 237 IP 10 133 242 MAC EF MAC MAC D3 MAC 00 Figure A2 Connectivity Diagram Note down each sensor s position x y and z 128 Note If you are using the UWB system on a client computer then skip Step 12 It is necessary only when you will be using the UWB system on a server 12 Open Platform Control Software Application Start gt All Programs gt Ubisense 2 1 2 Platform Control in the service part start both of the following UbisenseCoreServer 2 1 UbisenseServiceController 2 1 13 Open a software application named Site Manager Start gt All Programs gt Ubisense 2 1 2 Site Manager and follow below mentioned steps these steps are critical i Load Walls by first clicking on the Areas tab and then going to Walls tab and then Load Walls See Figure A3 Here select the desired dat file which was saved in Step 3 Site Manager Tera gt s r Walls Area Regions Help EB Types H Objects Representation E Areas HU Cells Object Locations Geometry 4b Dae ROSE PP LI A 4 Figure A3 ii Draw a line inside the loaded area then go to Regions tab a
20. VIDEO DATA FOR CONSTRUCTION EQUIPMENT EOC ZANE 89 4J ROC PICTON vever 89 4 2 Comparison of UWB and Video Technologies for Construction Projects 89 4 3 Proposed Approach cpi et ie ibtd e UH deca ae bebe dua d danas 9 431 Hardware Components aao p ede eder ped I a E bn aid vens 9 482 Software Components sace dice e o inet pc Dea oae deditus 92 44 Implementation and Case Study cies oie tela texta do ded ey vao bius elati vie Dinos 96 4 4 1 Design of Experiment ioca teer eo adco dicm yu med nri deme 96 AAD Implementation ss uot dtr ee ev Ode dete aeu d edd ERA Waste dd 98 4245 Aabakken 109 45 Summary and ConclusiOls LS 117 CHAPTER 5 CONCLUSIONS AND FUTURE WORK nnerooneeneenssnrennssvenneensenesnsensenssnsennsese 119 Sa S s Rv 119 5 2 gt Research Contributions and Conclusionsusmrevenmn snniidnjnsnieinstvnv 120 5 3 Limitations and Future Work aee eerte etes tier eo orn din Re EIN Reed 121 REFERENCES ua 122 APPENDIX A UWB System Configuration User Manual sese 127 APPENDIX B Detection and Localization MATLAB Code for Image Processing 143 vi APPENDIX C Data Fusion MATLAB Code eese eee nen entren APPENDIX D List of Related Publications vii LIST OF FIGURES Figure 2 1 UWB Tags Ubisense 201392 aen icto te etti da texto dateres exta dedan 6 Figure 2 2 Angle of Arrival Technique adapted from Ghavami et al 2004
21. and concrete stairs whereas the outdoor test was conducted at a busy street in Downtown Montreal The results of the indoor test are presented in Table 3 12 It can be observed that at shorter distances i e less than 15 m the system works properly without line of sight but as the distance increases the connectivity is not possible with thick obstacles which are normally present in the indoor environment In this test setting at the distance of 27 5 m the bridge and access point were separated by some walls and stairs which include steel and concrete whereas with complete line of sight the connectivity is good up to a distance of almost 61 m Table 3 12 Results of Indoor Wireless Connectivity Test cud RE on e pore Connectivity Status 11 5 25 No Connected 25 Not Connected 2435 50 No Not Connected 100 Not Connected 33 25 Yes Connected 61 25 Yes Connected The results of the outdoor test are presented in Table 3 13 It can be observed that the connectivity is quite good up to a distance of almost 60 m whereas as the distance increases more than 60 m the connectivity is not reliable In the case where the distance is 80 m the bridge was at one side of the street and the access point was at the other side and there was a truck between the bridge and the access point 3 3 2 2 Outdoor dynamic test for tracking movement of a roller Design of Experiment This test was designed to tra
22. and then averaging image data over specific duration of time can improve the positioning accuracy of the video data The results of the MSDF approach are also dependent on the synchronization and alignment of data Our fusion success rate improved from 96 when applying the image location expressing method for the image alignment to 100 when applying the spatial transformation from control point pairs method 120 5 3 Limitations and Future Work Although in this research the prototype of the proposed MSDF approach is implemented with a simple architecture having a single camera and only one UWB cell containing four sensors and four tags the proposed system is easily scalable to a large scale system containing a network of cameras and various UWB cells where each UWB cell would be monitored by an individual camera or stereo vision camera and various construction equipment would be tagged with several UWB tags Furthermore the proposed MSDF approach is currently applied to construction equipment however it can also be applied for construction workers safety In that case each worker would be assigned a unique UWB tag and the IPC would be trained so that it can also detect persons working in the monitored area Additionally this type of MSDF approach can also be applied to manufacturing facilities or healthcare services In view of the conclusions drawn from this research and the recommendations the future efforts can be directed
23. avoid using the same images for annotation and detection different images were selected for detection out of the 30 images in each second 99 The detector was applied separately for each EoI The detector first detects the Eol in the image and then localizes it in the form of a rectangular box and outputs three parameters of the rectangular box These three parameters are 1 the pixel coordinates Xa Ya of the left top corner of the box 2 the width of the box and 3 the height of the box For the fusion of the Eol s data with its corresponding UWB data the center point of this box is then calculated using Equation 4 1 The MATLAB code for detection is attached in Appendix B width height Center Point Xa Ya x 2 yr 2 4 1 The detection results for the truck and the excavator are shown in Figure 4 6 From this figure it can be noted how the EoIs were localized in the form of a rectangular box Figure 4 6 a shows the position of the truck when it was being loaded by the excavator and Figure 4 6 b shows its position during the dumping operation Figure 4 6 c shows the position of excavator during the digging operation whereas Figure 4 6 d shows its position while it was loading the truck c Excavator Digging d Excavator Loading Figure 4 6 Detection Results 100 Furthermore the outcomes of the detector were studied There are four possible outcomes Fawcett 2006 as shown in Table 4
24. axis They also considered several non linear equations but no equation actually improved the positioning accuracy Furthermore out of five line segments the developed regression lines for three line segments were unable to improve the accuracy In this situation they assumed that the heavy metallic items bookshelf and mailboxes standing against the walls near this space may have distorted the UWB signals To overcome this issue they further divided this straight line into three 16 sections hallway entrance hallway and end of hallway Then they applied the regression equation only to the middle section of that line and after removing the outliers the raw data was applied to the other two sections With the revised error model the accuracy of that line improved significantly Office Se JeusXooq Figure 2 12 Predetermined Five Paths for Closed Space Dynamic Test Cho et al 2010 In order to validate the proposed error model Cho et al 2010 collected a new set of data and analysed with the pre determined outlier constraints and regression equations for the five pre determined paths They found that using the proposed error modelling process the positional accuracy improved by about 27 8 They also suggested that the Kalman filter and the Kalman smoother perform better with the proposed error modelling process Furthermore they recommended that the developed error modelling processes can be extended to other wireles
25. by its UM data point 4 4 Implementation and Case Study The implementation of the proposed approach is performed through a case study This section firstly discusses the design of the case study for the implementation and validation of the proposed approach then explains the implementation and finally analyzes the results of the implementation 4 4 1 Design of Experiment A case study was designed to validate the proposed approach In this case study the locations of two construction equipment a truck and an excavator were estimated during a simulated earth moving operation in the lab environment using Remotely Controlled RC excavator and truck In addition an RC crane was used as a potential obstacle for the other two equipment Figure 4 5 shows the Design of Experiment DoE for this test The crane was standing at position C for the whole duration of test while not performing any operation The excavator and the truck were involved in four earthmoving activities which were 1 digging 2 loading 3 hauling and 4 dumping In order to have a ground truth the working zones for these activities were identified and marked on the floor as shown in Figure 4 5 with black rectangles The excavator was performing digging and loading activities at position E whereas the truck was 96 involved in waiting to be loaded at position T hauling from position T to T2 and dumping at position T Each of the two equipment the exc
26. by the site engineer before the site visit As it can be seen in this figure one large excavator and one small excavator are performing earthmoving operations However when the site was visited on Monday June 23 2014 two large excavators were present in the site area along with a large crane as shown in Figure 3 31 a At that time these equipment were working on the demolition of a concrete chimney rather than performing earthmoving operations This was a setback for the UWB data collection process as the heavy metallic body of the large crane was a significant source of radio noise for the wireless UWB system The demolition process was carried out for two consecutive days and the crane left the site on the third day The test was conducted for four days i e from Monday June 23 2014 to Thursday June 26 2014 The site conditions on each day are shown in Figure 3 31 69 Figure 3 28 Site View on May 22 2014 before Visit For localizing the excavator through the wireless UWB system four UWB sensor panels were attached to the fences covering an area of about 36 5 m x 22 m Each UWB sensor panel was configured by installing a UWB sensor its corresponding wireless bridge and a cable container box on a fiberglass sheet as shown in Figure 3 29 These sensor panels were specially designed as per the discussion with the site engineer because due to safety reasons it was not feasible to install the UWB sensors on tripods w
27. changed to about 36 5 m x 20 m as shown in Figure 3 30 Moreover ten UWB tags were attached to the excavator through magnet The positions of tags on the excavator are shown in Figure 3 32 The EUR of the tags was set to 4 19 Hz as almost 32 tags were present in the monitored area and SIF was used with all the default settings except MRM which was set to 3 Furthermore an IP camera was also installed on the site to have a complimentary source of data for visual validation of the results of the UWB system The data from both data sources were recorded for almost two days Performance Analysis The data are analyzed in the 2D x y plane In order to demonstrate the analysis method two separate short periods of the test were analyzed both of three minute duration The first period was on Day 4 from 12 52 PM to 12 54 PM when the excavator was stationary and not performing any operations whereas the second period was also on Day 4 from 11 27 AM to 11 29 AM when the excavator was not stationary and performing an operation 71 Fiberglass Sheet UWB Sensor Wireless Bridge Cable Container 1 EET T ANTES Figure 3 29 UWB Sensor Panel First Period Analysis Initially the AUR and MDR of each tag are analyzed and presented in Table 3 17 It can be observed that for some tags the AURs are very low compared with the EUR whereas for some tags the AUR is less but satisfactory Moreover out of 10 tags the MDR for 5 ta
28. effect No Yes Line of sight and occlusion issues Provides location with Provides location with error more training Training required No Yes Cost of deployment High Low Configuration at site Difficult Easy Tagging issues e g battery replacement Yes No 90 4 3 Proposed Approach The proposed MSDF approach consists of two sensory data sources which are 1 UWB RTLS and 2 video The high level concept of the proposed approach as shown in Figure 4 1 is that each construction equipment is tagged with UWB tag s and the UWB sensors would localize these tags On the other hand the video of the monitored area would be recorded using an IP based surveillance camera The UWB system and the camera are connected to a server through Local Area Network LAN Several software components are running on the server which are 1 UWB Module UM for the configuration of the UWB system and recording UWB tags positions 2 Video Module VM for recording the video of the monitored area and applying image processing technique for extracting the location of construction equipment and 3 Fusion Module FM for fusing data from the UWB RTLS and video The FM would then provide the required accurate location of the construction equipment This location can be extracted using either of the afore mentioned technologies i e UWB RTLS and image processing however that information would not be as accurate and timely as the
29. of the P4 who was moving randomly without following a specific path 41 6 6 5 5 4 4 3 3 2 2 1 1 0 0 3 4 5 6 7 3 4 5 6 7 a Slim Tag b Compact Tag Figure 3 3 Tag s Performance Comparison for P1 7 7 6 6 5 5 4 4 3 3 2 2 2 3 4 5 6 7 2 3 4 5 6 7 a Slim Tag b Compact Tag Figure 3 4 Tag s Performance Comparison for P2 42 6 6 5 5 4 4 3 3 2 2 1 a Slim Tag b Compact Tag Figure 3 5 Tag s Performance Comparison for P3 5 5 5 5 4 5 4 5 3 5 3 5 2 5 2 5 1 5 1 5 3 4 5 6 7 3 4 5 6 7 a Slim Tag b Compact Tag Figure 3 6 Tag s Performance Comparison for P4 3 3 1 2 Indoor Dynamic tests to compare Performance of Wired and Wireless Systems I In order to compare the performance of wired UWB system with the wireless UWB system two sets of indoor dynamic tests were conducted with two tests in each set These tests in which the movement of a remote controlled RC crane was tracked were conducted in a lab environment 43 wa Design of Experiment The Design of Experiment DoE for both set of tests was kept the same in order to simplify the performance comparison of both systems The boom of the RC crane was moved in a circular path around its center of rotation The ground truth was the controlled movement of the RC crane from which the center point and the radius of the circle were measured Furthermore it was analyzed whether the tracked movement of the boom of the RC crane was smo
30. proposed approach can provide due to its capability to overcome the limitations of the individual technologies as discussed in Section 4 2 I N wu x 7 LAN AN Pan Tilt Zoom Camera Software Components on Server UWB Video Fusion Module Module Module Figure 4 1 Proposed Approach Overview 4 3 1 Hardware Components The hardware components of the proposed system are required for actively perceiving all the events happening within the monitored area and for recording these events The UWB RTLS and camera are required for perceiving events and the server is used for recording these events 91 4 3 1 1 UWB RTLS The major hardware components of the UWB system are tags sensors and network components as discussed in Section 2 2 The tags are attached to the construction equipment and each tag is associated with its corresponding equipment in the system Several UWB sensors are strategically placed on the edges of the monitored area These sensors require power networking and timing cable connections depending upon the required mode of the UWB system i e wired or wireless All sensors are connected to the server through a network switch 4 3 1 2 PTZ Camera An IP camera is required for recording the video of the events happening in the monitored area This IP camera is connected with the server through a network switch A Pan Tilt Zoom PTZ camera can be used as it can cover larger area by linking it
31. rate Train the system using annotated frames and negative images Adjust PTZ to cover desired FoV Start recording Apply detector Save pixels in Split video to frames i Image Data pixels Fusion Module for each time step i Transform M and Vj into GCS and save in M H and v Association satisfies type and threshold conditions Estimate position p y of equipment j using available data sources Apply k NN Calculate Algorithm with Euclidean k 1 distances Update P Figure 4 2 Proposed Approach 93 Data Alignment Data Association Position Estimation 4 3 2 2 Video Module The VM basically performs two tasks 1 video recording and 2 image processing During the software settings for video recording the camera is registered the frame rate is set according to the UWB tag update rate and the PTZ camera is adjusted to cover the desired FoV Whereas for the image processing software settings include frame annotation and training of image processing software For the data collection the recorded video is split into frames and then passed to the Image Processing Component IPC which then detects the equipment using Histograms of Oriented Gradients HOG technique Dalal amp Triggs 2005 and saves its pixel coordinates which are in Video Coordinate System VCS The results are saved in a matrix named v which is then passed to the FM The limitation of the VM in the proposed appro
32. standard deviation is satisfactory Furthermore the standard deviation for Tag 7 corresponds to an error of about 0 5 meter in both directions The same conclusion can be drawn by visually inspecting the data points from these tags as shown in Figure 3 33 From this figure it can be observed that the data 173 points for Tag 2 are very scattered whereas the data points for Tags 3 4 and 5 are more concentrated Lastly the data points for Tag 7 are also scattered but not as scattered as the data points of Tag 2 Table 3 17 AUR amp MDR Analysis for Period 1 Tag AUR Hz MDR 1 0 16 96 29 2 1 28 69 53 3 3 20 23 55 4 2 09 50 18 5 2 38 43 56 6 0 20 95 10 7 1 84 56 28 8 0 33 92 58 9 0 16 96 95 10 0 07 98 15 Table 3 18 Mean amp Standard Deviation Analysis for Period 1 Tag Mean Position m Standard Deviation m X y X y 2 55 14 107 06 1 37 0 92 3 56 24 112 87 0 13 0 18 4 55 57 112 33 0 13 0 17 5 54 81 113 33 0 28 0 49 7 52 32 115 84 0 52 0 55 After this analysis the orientation of the excavator for this three minute duration estimated by the wireless UWB system was analyzed For analyzing orientation the data from the same five tags were processed As for each of these tags the AUR is more than 1 Hz so each tag s data were averaged over a period of 1 second for the whole duration of 3 minutes Tags 3 and 4 were i
33. state from the associated measurements whereas identity estimation stage categorizes the object from which the measurements originated They further discussed major techniques and algorithms used for each fusion stage which are summarized in Table 2 1 Zeng et al 2006 explained fusion process at feature and decision levels In the feature level fusion feature extraction is performed in order to yield a feature vector from the observation of each sensor After the data association stage where feature vectors are sorted into meaningful groups these feature vectors are then fused and an identity declaration is made based upon the joint feature vector Whereas in the decision level fusion each sensor performs independent processing to produce a declaration of identity and then the declarations of identity from each sensor are subsequently combined via a fusion process The Kalman Filter KF is a mathematical tool used for estimating the instantaneous state of a linear dynamic system and filtering out the noise by using measurements linearly related to the state but corrupted by white noise Grewal amp Andrews 2008 It is mostly used for the control of complex dynamic systems such as continuous manufacturing processes aircraft ships or spacecraft The Kalman Filtering is an iterative and recursive process which consists of two sub processes the time update and the measurement update In the time update process a prior estimate is comp
34. tag update rate for construction safety applications would be 1 Hz Additionally appropriate values for the RF frequency and power of the wireless bridges should be selected based on the distance between the bridges and the environment conditions The following conclusions are drawn from the analysis presented in this chapter 1 The data from the wireless UWB system should be enhanced using a suitable data enhancement method in order to accurately track the movement of the tagged object as discussed in Section 3 3 1 2 and Section 3 3 1 3 however high MDR restricts the applicability of data enhancement methods and also degrades data 2 The wireless UWB system has high MDR compared with the wired system The reason is that it uses only AOA estimation technique which reduces the number of readings which are required for the filter to calculate the location Additionally the wireless bridges are a vital component of the wireless UWB system and their precise configuration is essential as discussed in Section 3 3 2 1 87 3 The calibration process is less controllable in construction sites and small angular errors in calibration result in larger positioning errors due to the large scale of construction sites To maximize the utility of the wireless UWB system on construction projects it is recommended that 1 Using more tags on each piece of equipment can provide more data so that if some tags have high MDR the other tags dat
35. that in this case tag S2 has the worst performance with an MDR of 73 91 whereas tag S4 has the best performance with an MDR of 23 93 Then the MDR was analyzed at the second level which means if the data is missing for the whole second then it is considered as missing data Table 3 16 shows the MDR analysis at the second level In view of the analysis presented in Table 3 14 Table 3 15 and Table 3 16 it is concluded that the performance of tags S1 and S2 is unsatisfactory and in contrary the 59 performance of tags S2 and S3 is reasonable One explanation of this is that having been placed higher on the equipment S3 and S4 had better visibility than S1 and S2 22 98 m o 2 Sensor 3 _ 2 Un B Origin 9 0 a Sensor 1 Sensor 4 Figure 3 18 Area Settings for Outdoor Dynamic Test S 3 S b Tag Positions a During Compaction Process Figure 3 19 Tracked Roller for Outdoor Dynamic Test 60 Table 3 14 AUR Analysis for Outdoor Dynamic Test AUR Hz Tag Real Test Static Test S1 2 34 8 29 2 2 30 8 09 S3 4 53 8 31 4 6 56 8 31 Table 3 15 MDR Analysis msec for Outdoor Dynamic Test Missing Data Rate msec Duration min Period S2 S3 To further investigate each tag s performance in terms of the logged data control charts were drawn for the time difference between two consecutive readings for the first 500 d
36. the equipment and workers can certainly assist in reinforcing the safety and improving the overall efficiency The availability of real time information is also the basis for the concept of Smart Construction Site SCS which aims at improving the overall safety sustainability and efficiency of a construction project by making the real time information about the project available to all the stakeholders in order to enable them to make right decisions at the right time Zhang et al 2009 describes SCS as an intelligent integrated setup where 1 the information about the entire environment is acquired from the sensors attached to moving objects 2 equipment s path is automatically planned and 3 every stakeholder including the staff members has intelligent assistance from various agents providing information and decision making strategies The advancements in Real time Location Systems RTLSs such as Radio Frequency Identification RFID and Global Positioning System GPS have enabled researchers to investigate the applicability of these systems to automate the on site data collection process Ultra Wideband UWB technology a type of RTLS has been investigated by several researchers for the identification localization and tracking of construction resources The UWB technology has the potential to track and visualize construction resources on site and increase the awareness level of the construction staff in near real time UWB RTLS provid
37. the location of the tags captured by the wireless UWB system has a lot of variation Additionally the error distribution for this accuracy assessment was investigated as shown in Figure 3 44 For investigating this error distribution different error ranges were defined each having a length of 5 and then it was calculated how many times the error occurred within each range From Figure 3 44 it can be observed that an error within the range of 20 to 15 occurred for the maximum of 8 4 Furthermore it can also be noted that 57 of the error is distributed within the range of 25 to 25 84 UWB Covered Area B Orientation UWB Covered Area Orientation UWB Covered Area Orientation 7 37 47 57 67 7 3T 47 ST 67 7 37 47 57 67 96 96 96 106 106 106 116 116 116 126 126 126 a Second 30 b Second 31 c Second 32 UWB Covered Area li Orientation UWB Covered Area li Orientation UWB Covered Area li Orientation 7 37 47 57 67 7 37 47 57 67 7 37 47 57 67 96 96 96 106 106 106 116 116 116 p p 126 126 126 d Second 166 e Second 167 f Second 168 Figure 3 42 Scatter Plots for Orientation of Excavator Period 2 85 a First Minute b Last Minute Figure 3 43 Excavator Position 0 5 Probability of Occurrence 95 90 85 80 75 70 65 60 55 50 45 40 35 30 25 20 15 10 5 0 5 10 15 20 25 30 35 40 45 50 55 60 65 70
38. to convert Xe yc from cm to m and also to remove the systematic error These values for removing the systematic error were found after some trial and error analysis After this transformation some GCS coordinates were visually validated and it was noted that this transformation method has produced better results than the previously applied transformation method Figure 4 10 Control Points Table 4 7 Coordinates of Control Points VCS Coordinates GCS Coordinates Control Points pixels cm a 457 376 176 275 b 995 376 268 275 c 494 261 176 337 d 943 261 268 337 106 Figure 4 11 Coordinate Transformation using Second Method _ 1 Xecs oo 4 6 604 yc Jos oo 4 7 After the coordinate transformation the transformed locations from the UM and the VM were analyzed for synchronization This analysis showed that the data from the UM was lagging by a period of 5 seconds because the data from the UWB system and the camera were recorded using two different computers and the time difference between these two computers was 5 seconds So the UWB frames were advanced by an interval of 5 seconds and then these frames were passed on to the next fusion stage for association with the video frames 107 Data Association and Position Estimation Three association cases were identified based on the outcomes of the detector from the VM which are descr
39. via EKF in each resolution level and fused the outcomes for improved estimation of the trajectory Their proposed approach not only performs a vision task in a robot but also provides it with a simulated human vision 31 Gustafsson et al 2002 developed a framework for positioning navigation and tracking problems using particle filters which consists of a class of motion models and a general nonlinear measurement equation for the position They presented a general algorithm which is parsimonious with the particle dimension They described how the technique of map matching is used to match an aircraft s elevation profile to a digital elevation map and a car s horizontal driven path to a street map They showed that the accuracy in both cases is comparable with satellite navigation e g GPS but with higher integrity They also argued based on their simulations how the particle filter can be used for positioning based on cellular phone measurements for integrated navigation in aircraft and for target tracking in aircraft and cars 2 8 Summary The literature review presented in this chapter focused on the UWB RTLS and MSDF technologies and their applications specifically in the domain of construction management UWB RTLS is an effective technology for localizing construction equipment on construction sites Several researchers have investigated the applicability of UWB RTLS for construction management however a thorough examination of th
40. were out of range whereas the RTS was able to track the workers The system used by Saidi et al 2011 was a UWB only based on TDOA and did not use AOA Furthermore out of the two sets of experiments conducted by Saidi et al 2011 one set was conducted in an open space field whereas the other set was conducted in a construction lay down yard The real construction environment normally include both indoor and outdoor conditions however this research only focuses on the outdoor conditions of the construction environement because the indoor conditions are more challenging in terms of establishing a ground truth measurement due to the obstacles and the limitations in the power output of the UWB system used They also assumed the conditions to be ideal if they have minimal obstacles and reflections and have a good medium for RF signal propagation Figure 2 9 UWB Tracking in Lay Down Yard Saidi et al 2011 Cho et al 2010 analysed the reliability of the wireless UWB system s data for tracking assets in indoor construction sites They conducted static and dynamic tests in various building spaces They also developed an error model to minimize the positioning errors of wireless UWB system using some statistical techniques including regression analysis outlier detection and Kalman 13 filtering While conducting these indoor tests they kept at least one receiver in direct LoS from any location of the monitored area The
41. with the UWB system to adjust the FoV according to the position of the tracked equipment 4 3 1 3 Server The server is a high speed computer which is required for running the different software components of the system The server would collect the data from the UWB system and the IP camera The server must be connected to the same network via the network switch to access the UWB sensors and the IP camera 4 3 2 Software Components The three modules of the software components are described in detail in Figure 4 2 and further explained in the following sections 4 3 2 1 UWB Module The UM is used for the configuration of the UWB system and recording the UWB tags locations All UWB tags along with their corresponding equipment are registered in the UM The tag update rate is set in the UM according to the DoE After the UWB software settings the UM records the UWB tags positions which are in the UWB Coordinate System UCS in a matrix named u Furthermore the UM also saves the type and ID of the equipment with which the tag was attached in u and then passes it to the FM 92 UWB Module Register tags in the system Software Setting Set the update rate Configure the monitored area and sensors positions Start recording Data Collection Save locations in u UWB Data locations Reject association Video Module Video Recording Image Processing Register camera Annotate frames Set the frame
42. within 20 cm For the former experiment they used an industrial total station to measure the locations of sensors whereas for the latter one they used a differential GPS with a measurement error of 20 cm to 30 cm They positioned six UWB sensors see Figure 2 7 where line of sight LOS to all sensors was available throughout the coverage area and mounted three UWB tags spaced at 1 meter intervals with the highest tag at 3m see Figure 2 8 on a fiberglass pole Then they established a ground truth model and placed multiple benchmarks within the open space field 10 Figure 2 7 UWB sensors positions A1 to A6 Saidi et al 2011 To collect the data they moved the UWB tag pole from one benchmark to the next At each benchmark the data were collected for one minute and the UWB tag pole was placed at 48 benchmarks This procedure was repeated twice first the calibration items which are the locations of the UWB sensors and reference tag were measured with the total station whereas for the second time the calibration items were measured with differential GPS receiver It took them almost 10 hours to setup and collect data for each of the above experiments Saidi et al 2011 also highlighted that the case in which the locations of the calibration items were measured with the total station i e with the accuracy of 1 mm represented the ideal setup procedure for the system which might not be achievable in the field due to practi
43. z N A Candidate 2 x y 2 N A Select the survey point group to which to add the new point or type in a new group name Enter a new survey point name It must be unique to the group selected above Names currently in the group are listed for convenience 137 Sensor 4 P Sensor 3 23 14 13 Sensor 1 Master x Point 1 M ud Origin Two other distances would be required which are Coordinates of d 0 distance from Sensor I to Point 1 Point 1 0 0 z Point 2 d 0 Z d 0 distance from Sensor 2 to Point 2 12 Figure A10 Outdoor Testing Configuration Table A1 Distance Value Description di5 Distance from Point 1 to Sensor 1 d23 Distance from Point 2 to Sensor 1 dig Distance from Point 1 to Sensor 4 da4 Distance from Point 2 to Sensor 4 di Distance from Point 1 to Sensor 2 dii Distance from Point 2 to Sensor 1 dii Distance from Point 1 to Sensor 1 0 dz Distance from Point 2 to Sensor 2 dit Distance from Point 1 to Tag dzt Distance from Point 1 to Tag 139 Table A2 Data Cables Orange 40 m Green 23m Yellow 85m Blue 80m Timing Cables All 3 60m Power Cables Orange 30m Yellow 20m Yellow 30m 140 Wired Connectivity For TDOA normal connection of timing cables is from any timing cable socket on the master sensor to the input timing cable socket to
44. 00 000 m within the lay down yard positioned the UWB sensors at the boundary of the yard and tagged several construction workers and machines with UWB tags They did not consider the height z coordinate of the tracked item person in this set of experiments They synchronized the timestamps of UWB system with a construction robotic total station RTS within 1s and registered both location measurement systems i e UWB and RTS to a common coordinate system They mounted a UWB tag with a 1 Hz update rate and a mini RTS prism on a construction worker s helmet and collected data without interruption for 32 min and 14 s or 1287 position points with both systems where they used the RTS measurements as ground truth They calculated the location errors by calculating the difference between the UWB and RTS measurements and found that almost 47 of all errors were less than 1 25 m whereas 87 were within 2 5 m They defined an unusual activity if at an UWB update rate of 1 Hz the difference 12 between one location reading and the next is greater than 2 5 m as the worker might be jumping or falling They also proposed that if this type of unusual activity happens the data might be fed to any alert system They also proposed the optimization of UWB covered area as it will reduce the installation cost along with impacting the tracked resources Furthermore from this experiment they also noted that at a distance of greater than 100 m the UWB signals
45. 2 associated ut vl amp ue v2 elseif utV1 gt utV2 truckPosition i ut v2 2 excPosition i uetvl 2 associated ut v2 amp ue v1 end case3 case3 1 145 APPENDIX D LIST OF RELATED PUBLICATIONS Journal Papers Vahdatikhaki F Hammad A amp Siddiqui H Submitted 2014 Optimization based Excavator Pose Estimation Using Real time Location Systems Automation in Construction Rafiee M Siddiqui H Hammad A amp Zhu Z To be submitted 2014 Improving Indoor Security Surveillance by Fusing Data from BIM UWB and Video Automation in Construction Conference Papers Siddiqui H Vahdatikhaki F amp Hammad A 2014 Performance Analysis and Data Enhancement of Wireless UWB Real time Location System for Tracking Construction Equipment n Proceedings of the 21st International Workshop on Intelligent Computing in Engineering 2014 Cardiff UK Rafiee M Siddiqui H amp Hammad A 2013 Improving Indoor Security Surveillance by Fusing Data from BIM UWB and Video n Proceedings of the 30th International Symposium on Automation and Robotics in Construction Montreal 146
46. 28 2 7 Applications of Multi Sensor Data Fusion MSDF has numerous industrial applications This section mainly discusses the MSDF applications in the domain of construction management Some other applications of MSDF are also discussed in this section 2 7 1 Applications in Construction Management The impact of data imperfections on construction process monitoring and the benefits of the data fusion approach for construction management have been investigated by several researchers Luo et al 2013 explored the effects of location aware sensor data imperfections e g erroneous or missing data on the autonomous jobsite safety monitoring and investigated methods to reduce the impacts of the sensor data imperfections on the jobsite safety system They found that the imperfections of the location data collected from various location aware sensors strongly affects the safety monitoring system Furthermore they suggested the data fusion approach to reduce the sensor data imperfections and to improve the performance of the jobsite safety monitoring system Chi amp Caldas 2012 presented an automated image based safety assessment method for earthmoving and surface mining operations They evaluated the image based data collection devices and algorithms for safety assessment and also discussed the analysis techniques and rules for monitoring the safety violations They found that the applied safety rules enabled automated violation detection and
47. 3 These outcomes are described as 1 true positive T if the Eol is present in the image and it is detected correctly 2 true negative 7 if the EoI in not present in the image and nothing is detected in the image 3 false positive F if the Eol is present in the image and something else is detected and 4 false negative F if the EoI is present in the image and nothing is detected In this case study there is no true negative outcome i e T 0 because in all the sampled images for detection both EoIs were present Table 4 3 Description of Detector Outcomes Outcome Eol Detected True Positive Ty Yes Yes Correctly True Negative Tn No No False Positive Fy Yes Yes Incorrectly False Negative Fn Yes No The outcomes of the detector are shown in Table 4 4 It can be observed that for the truck T was satisfactory 64 whereas on the contrary the 7 for the excavator was very low i e 17 One reason for this low 7 can be the similarity in the color of several parts of the excavator and the background of the image Table 4 4 Analysis of Detector Outcomes Outcome Excavator Truck True Positive Tp 17 64 True Negative T 0 0 False Positive Fy 3 15 False Negative Fn 80 21 Furthermore the performance metrics Powers 2011 Fawcett 2006 of the detector were calculated which are 1 Recall which is the proportion
48. 5 software application for recording data The tags come in various form factors Figure 2 1 depending on the asset to be monitored e g for tracking people slim tag Figure 2 1 a is used whereas for tracking equipment compact tag Figure 2 1 b is used The sensors are installed at the boundaries of the monitored area which forms a cell and the higher the number of sensors in a cell the better the accuracy of the tag s position estimated by the UWB system Each sensor gathers two types of information from the signal received from the tag the angle of the signal and the time when the signal is received Maalek amp Sadeghpour 2013 The UWB system utilizes two positioning techniques to estimate the tag s position depending on the information received by the sensors which are Angle of Arrival AOA and Time Difference of Arrival TDOA In the AOA technique the angle of the arrived signal is measured at several sensors by routing the main lobe of a directional antenna or an adaptive antenna array Each measurement forms a radial line from the sensor to the tag For 2D localization the location of the tag is defined at the intersection of two directional lines of bearing as shown in Figure 2 2 Ghavami et al 2004 In the TDOA technique the difference in the arrived signal s time at two different sensors is calculated Then each time difference is converted to a hyperboloid with a constant distance difference between the two sensors where t
49. 7 xlsm Sheet 1 excel sheet number uwbtruck xlsread UWBTruckDataPath Sheet uwbexcavator xlsread UWBExcavatorDataPath Shee videotruck xlsread VideoTruckDataPath Sheet videoexcavator xlsread VideoExcavatorDataPath Sheet t Data Association no video item availavle one video item availavle both video items availavle 144 end ut uwbtruck i ue uwbexcavator i vl videotruck i v2 videoexcavator i i associated if sum v1 0 amp amp sum v2 0 truckPosition i ut excPosition i ue casel casel l associated no association elseif sum vl 0 amp amp sum v2 0 utVl dist ut vl ueVl dist ue v1 if utV1 lt ueV1 truckPosition i ut vl 2 excPosition i ue associated ut vl elseif utVl gt ueV1 truckPosition i ut excPosition i uetvl 2 associated ue v1 end case2 case2 1 elseif sum vl 0 amp amp sum v2 0 utV2 dist ut v2 ueV2 dist ue v2 if utV2 ueV2 truckPosition i ut v2 2 excPosition i ue associated ut v2 elseif utV2 gt ueV2 truckPosition i ut excPosition i uetv2 2 associated ue v2 end case2 case2 1 elseif sum vl 0 amp amp sum v2 0 end utVl dist ut v1 utV2 dist ut v2 if utV1 lt utV2 truckPosition i ut vl 2 excPosition i uetv2
50. D b 3D Figure 2 6 Comparison of Cumulative Accuracy Curves for AOA vs TDOA amp AOA Maalek amp Sadeghpour 2013 As all the variables for these experiments were simulated in an indoor environment and all tracked items were in a static mode the nature of real construction site which is mostly outdoor and highly dynamic can affect the UWB system s performance significantly Saidi et al 2011 also conducted several experiments to evaluate the static and dynamic performance of a UWB RTLS Their focus was to design the testing of this type of RTLS for personnel applications in open space and in realistic construction conditions Moreover they also developed a mathematical static model for estimating position errors of this system Saidi et al 2011 also identified twenty three factors that influence the accuracy of the UWB system which include the calibration error hardware antenna type receiver orientation and the tags roll pitch and yaw angles They also suggested that the effect of the orientation yaw angle of the UWB tag is one of the most important factors They designed the open space experiments to evaluate firstly the 3D errors and secondly the sensitivity of the 3D errors to inaccuracies in the measured positions of the sensors Within this set of experiments two experiments were conducted the first one with the sensor locations known to be within 1 mm and the second one with the sensor locations known to be
51. Figure 3 35 Schematic View of Orientation of Excavator Excavator image is taken from Google 2014 78 UWB Covered Area Bi Orientation UWB Covered Area B Orientation UWB Covered Area Bi Orientation 106 106 106 116 116 116 126 126 a Second 1 b Second 2 c Second 3 126 UWB Covered Area H Orientation UWB Covered Area HB Orientation UWB Covered Area B Orientation 106 106 106 116 116 116 126 126 d Second 178 e Second 179 f Second 180 126 Figure 3 36 Scatter Plots for Orientation of Excavator Period 1 79 Furthermore in order to assess the accuracy of the wireless UWB system on a construction site an analysis was conducted based on the angle between the lines formed by joining Tags 2 and 3 and Tags 5 and 7 as shown in Figure 3 37 The distance between Tag 1 amp Tag 2 d 2 and Tag 1 amp Tag 3 d 3 were measured using a measuring tape at the time of installation of tags on the excavator as 2 63 m and 3 65 m respectively 10 8 6 Cc co diz e HOED 7 I5 5 9 Figure 3 37 Angle Calculation for Accuracy Assessment Excavator image is taken from Google 2014 The expected angle 0 between lines and is calculated using Equation 3 5 which resulted in an angle of 35 78 1 di 0 tan 3 5 13 As the data from Tag 5 and Tag 7 are better than the data from Tag 1 and T
52. Fill the missing data using interpolation Optimization based Correction Calculate the pose N m 7 Operation N I finished Figure 2 15 Flowchart of Optimization based Method Vahdatikhaki et al 2014 2 5 Data Fusion Integration of data from multiple sources resulting in reliable and feature rich information is Nature s approach Creatures interpret signals from multiple sensors to judge the surrounding environment For example the human brain interprets signals from the five body senses sight sound smell taste and touch with knowledge of the environment to create and update a dynamic model of the world which allows humans to interact with the environment and make decisions about present and future actions Elmenreich 2002 Data fusion a multidisciplinary field is the process of integrating data or information in order to estimate the state of a system or an entity This integration enhances the confidence improves reliability and reduces ambiguity of measurements for estimating the state of entities in engineering systems It also enhances the completeness of fused data that is required for 22 estimating the state of engineering systems Shahandashti et al 2011 Data fusion has three general goals increasing the 1 completeness 2 conciseness and 3 correctness Completeness measures the amount of data conciseness measures the uniqueness o
53. UWB RTLS for Construction Equipment Localization Experimental Performance Analysis and Fusion with Video Data Hassaan Siddiqui A Thesis in Concordia Institute for Information Systems Engineering Presented in Partial Fulfillment of the Requirements for the Degree of Master of Applied Science Quality Systems Engineering at Concordia University Montreal Quebec Canada September 2014 Hassaan Siddiqui CONCORDIA UNIVERSITY School of Graduate Studies This is to certify that the thesis prepared By Hassaan Siddiqui Entitled UWB RTLS for Construction Equipment Localization Experimental Performance Analysis and Fusion with Video Data and submitted in partial fulfillment of the requirement for the degree of Master of Applied Science Quality Systems Engineering complies with the regulations of the University and meets with the accepted standards with respect to originality and quality Signed by the final examining committee Dr M Mannan Chair Dr N Bouguila CIISE Examiner Dr A Bagchi External Examiner BCEE Dr Amin Hammad Supervisor Approved by Chair of Department or Graduate Program Director Dean of Faculty ABSTRACT UWB RTLS for Construction Equipment Localization Experimental Performance Analysis and Fusion with Video Data Hassaan Siddiqui Construction sites are well known for their dynamic and challenging nature Several researchers are investigating the application of vari
54. UWB results Table 3 4 Tag IDs Tag IDs Person Slim Compact PI 010 000 084 205 020 000 101 222 P2 010 000 084 202 020 000 108 122 P3 010 000 084 228 020 000 059 088 P4 010 000 084 195 020 000 059 098 38 Compact Tag N 4 gt A Workstation c Position of Sensor 1 amp 2 d Position of Sensor 3 amp 4 Figure 3 1 Test Settings Performance Analysis For analyzing the performance of the wireless UWB system and comparing the performance of the slim tags with the compact tags firstly the Actual Update Rate AUR and the Missing Data Rate MDR of all tags were analyzed The AUR and the MDR are calculated using Equations 3 2 and 3 3 respectively In Equation 3 2 At is the time difference between two consecutive readings 39 1000 PE mean At 3 2 siop ed s MOIS EDD EON RE Test Duration sec 100 6 3 Origin Saem 0 0 0 Y Figure 3 2 Area Settings Sensor 2 8 72 m Sensor 3 Table 3 5 lists the AUR and MDR of both types of tags for each person It can be observed that the AURs are different from the EUR which was 16 Hz Especially for the compact tags the maximum AUR is 2 44 with an MDR of 84 88 for P2 which is considerably lower than the EUR whereas for the slim tags the maximum AUR is 12 99 with an MDR of 19 25 95 for P2 40 It is prominent that the AUR of compact tags is consi
55. WB Server a Wired System b Wireless System Figure 2 4 Schematic Diagram of UWB Systems adapted from Zhang et al 2012b 2 3 Applications of UWB RTLS in Construction Management Although UWB RTLS has several industrial applications the focus of this section is to highlight the applications of UWB RTLS in construction management As not much literature is available in this domain therefore some related literature is reviewed in detail 7 Maalek amp Sadeghpour 2013 evaluated the performance of UWB RTLS under certain conditions which occur very often on a real construction site They conducted seven different experiments to assess the accuracy of location estimated by the UWB RTLS For each experiment they simulated various construction site scenarios which are related to 1 the presence of metallic items within the monitored area 2 UWB signal blockage 3 metallic items tracking 4 wireless mode of UWB system 5 tracking multiple items and 6 the effect of number of UWB sensors total of 8 sensors To measure the accuracy of data Maalek amp Sadeghpour 2013 used the Distance Root Mean Squared DRMS Equation 2 1 method for 2D accuracy whereas Mean Radial Spherical Error MRSE Equation 2 2 method was used for 3D accuracy These methods are different from the average of Euclidean distances between the actual location and the estimated location as they provide a single value to represent the accuracy a
56. a can be used for positioning that piece of equipment however adding more tags limits the EUR 2 Using more UWB sensors can provide more visibility and therefore more readings to the filter for accurately calculating the locations This approach can also be used in the cellular architecture For example rather than making one cell containing 8 sensors two cells each containing 4 sensors can solve the problem of the limited EUR 3 The timely availability of the surveying team is very important for accurate system calibration which requires effective coordination with the site team and management In addition to overcome the limitations imposed on the performance of the wireless UWB system by the harsh environment of the real construction sites it is anticipated that fusing data from a complimentary sensory data source e g video can further enhance the localization of the construction equipment on construction sites as will be discussed in CHAPTER 4 88 CHAPTER 4 FUSING UWB AND VIDEO DATA FOR CONSTRUCTION EQUIPMENT LOCALIZATION 4 1 Introduction Providing real time information to all stakeholders of a construction project as discussed in Section 1 1 requires the perception of various aspects of a construction project Perceiving information about the aspects of a project requires the integration of multiple technologies for overcoming the technical limitations of the individual technologies by improving the performance of t
57. ach is that it can identify the type of equipment but it cannot identify its specific ID in case the site has several pieces of equipment of the same type e g several trucks However equipment can be identified by adding visual IDs labels on the equipment For example recently a project in downtown Montreal named Roccabella is visually identifying multiple cranes by labelling them as shown in Figure 4 3 4 3 2 3 Fusion Module The MSDF paradigm as discussed in Section 2 6 has four stages see Table 2 1 In the proposed approach the FM will iteratively work for three stages of the DF model which are 1 Data Alignment 2 Data Association and 3 Position Estimation At the data alignment stage the FM first transforms the UWB locations and image pixels at each time step i from UCS and VCS to the Global Coordinate System GCS and saves them to u and v respectively The transformation of image pixels from VCS to GCS can be done using either MATLAB s image location expressing technique MATLAB 2014a or MATLAB s spatial transformation from control point pairs MATLAB 2014b At the data association stage firstly the Euclidean distances between the positions from the two data sources are calculated and the k NN algorithm is applied to associate each equipment UWB location from u with its nearest neighbor within v The information about the type of equipment from the UM and VM serves as an additional input for
58. ag 3 the angle between lines and 5 is considered as the actual angle 04 and is compared with 6e The calculation of 0 was performed in three steps using the individual UWB tag s data which were averaged over a period of I sec These steps are 1 calculate the angle of l2 a with the local x axis 2 calculate the angle of 157 B with the local x axis 3 calculate 0 a B This process is shown in Figure 3 38 Moreover this process was performed for the whole three minute period s data Finally the mean and the standard deviation of the error e between 0 and 0 0 0e were calculated which were found to be 19 83 and 17 88 respectively From these values it can be observed that the location of the tags captured by the wireless UWB system has a lot of variation Additionally the error distribution for this accuracy assessment was investigated as shown in Figure 3 39 For investigating this error distribution different error ranges were defined each having a length of 5 and then it was calculated how many times the error occurred within each range From Figure 3 39 it can be observed that an error within the range of 10 to 15 occurred 80 for the maximum of 16 67 Furthermore it can also be noted that 85 5 of the error is distributed within the range of 5 to 40 Figure 3 38 Actual Angle 0 Calculation for Accuracy Assessment gt QI gt a Probability of Occurrence
59. area configuration in the Site Manager Application Open this application go to the Cells tab select the Area and check if there are any errors at the bottom of window If there are any remove all of them by selecting them one by one from the drop down list and then pressing the Remove Object button 142 APPENDIX B DETECTION AND LOCALIZATION MATLAB CODE FOR IMAGE PROCESSING o Detection of EoI in the images detector Vision Cascade0bjectDetector C Users umroot Documents Hassaan DataFusion Ca se Study for Thesis From Soltani TrainingAndAnnotationForTruck TruckHassan xml detector Vision CascadeObjectDetector C Users umroot Documents Hassaan DataFusion Ca se Study for Thesis From Soltani TrainingAndAnnotationForExcavator ExcavatorHassan xml xlfilename C Users umroot Documents Hassaan DataFusion DetectionResults July21 xlsx offset 2 for changing the row of excel sheet for i 3610 30 6580 imgfile sprintf C Users umroot Documents Hassaan DataFusion ImagesforDataFusion Tes tMay22 For Detection 04d jpg i I imread imgfile bbox step detector I try x bbox 1 1 y bbox 1 2 w bbox 1 3 h bbox 1 4 bboxPolygon x y X W y xtw yth x yth catch exception bboxPolygon 0 0 0 0 0 0 0 0 end xlWriteData i bboxPolygon xlRange sprintf A Sd offset sheet Excavatorbg xlswrite xlfilename xlWriteData sheet xlRange offset offset
60. as noted that the MSDF approach has the potential to better localize construction equipment by overcoming the limitations of UWB RTLS iii ACNOWLEDGEMENT First and foremost I would like to express my sincere gratitude to my supervisor Dr Amin Hammad for his intellectual support encouragement and patience His advice and criticism was my most valuable asset during my studies The joy and enthusiasm he has for his research was contagious and motivational for me even during tough times of this pursuit Besides my supervisor I would like to thank Mr Faridaddin Vahdatikhaki for his valued assistance in performing the tests and processing the test data I am also very thankful to Mr Mohammad Soltani for his appreciated support in the implementation of image processing I would also like to thank Mr Pierre Vilgrain from PCL Constructors Westcoast Inc for his support in the realization of the outdoor test In addition the support from all of the research group members was very invaluable My sincere thanks also go to Ms Mahsa Rafiee Mr Shayan Setayeshgar and Mr Ali Motamedi for their help in various aspects of this research I dedicate this thesis to my wife my parents and my brothers for their endless encouragement which is the essence of this accomplishment iv TABLE OF CONTENTS Eistof FMS CS eter edu ehe Eee tL Ec Eu EE NE EI ug vii Es xi Esben xiii CHAPTER D INTRODUCTION ed dates eese edet ET 1 Wid General Re
61. ata points as shown in Figure 3 20 Based on the EUR settings the ideal time difference between two consecutive readings of each tag is 119 25 milliseconds which corresponds to the EUR of 8 3 Hz A minimum acceptable threshold of 50 i e 4 15 Hz is assumed for the interval So the Tolerance Limit TL for this control chart is set to 119 25 x 2 238 5 milliseconds From Figure 3 20 a it can be observed that for tag S1 there are more than 40 data points exceeding the TL whereas also for tag S2 more than 40 data points exceeds TL as shown in Figure 61 3 20 b From Figure 3 20 c it can be observed that almost 40 data points exceeds the TL for tag S3 while for tag S4 almost 20 data points exceeds TL Figure 3 20 d Moreover along with the points exceeding TL the maximum time difference between two consecutive readings is also very critical From Figure 3 20 it can be observed that tag S2 has the highest difference between two consecutive readings which is 2756 msec and tag S4 has the best performance in this regard To investigate whether or not the low update rate and the high ratio of missing data were because of the performance of tags a static indoor test was conducted using the same tags and same EUR setting using the wired UWB System Table 3 14 shows the AUR of the four tags for the static test It can be observed that the AUR for each tag is very much close to the EUR Accordingly it can be observed that the tags are no
62. avator and the truck were tagged with two UWB tags UWB sensors were placed at the corners of the room as shown in Figure 4 5 A wired UWB system was used in this test to minimize the MDR The EUR of the UWB tags was set to 8 38 Hz which is suitable for the low number of tags An IP camera was placed near the wall in between sensor 1 and sensor 2 as shown in Figure 4 5 The frame rate for video recording was set to 30 fps The video and UWB data were recorded for 3 minutes Sensor 4 Sensor 3 Door T o oc UTE L Y E Camera Position Sensor 1 M X Metallic Whiteboard Sensor 2 4 16 m Figure 4 5 Design of Experiment for MSDF Case Study 97 4 4 2 Implementation 4 4 2 1 UWB Module The UWB tags EUR and sensors positions were configured using the Ubisense Location Engine Platform Ubisense 2013a whereas for tags registration and the monitored area settings Ubisense Site Manager Ubisense 2013b was used Furthermore for recording the data from the UWB sensors a logging application was used which was developed in house in C using the APIs from the manufacturer The AUR and MDR of the logged data were analyzed and presented in Table 4 2 It can be observed that the AUR values for all four tags are the same as the EUR and the MDR values are also very good To simplify the synchronization of the UWB and video data for fusion the UM would pass 1 location per second for each equipment to the FM Therefore th
63. c N C Magdic A Podbreznik P amp PSunder M 2008 Automated construction activity monitoring system Advanced Engineering Informatics 22 493 503 124 Rodriguez S 2010 Experimental Study on Location Tracking of Construction Resources Using UWB for Better Productivity and Safety Thesis Concordia University Concordia Institute for Information Systems Engineering Montreal Rosner B 1975 On the detection of many outliers Technometrics 17 2 221 227 Rosner B 1983 Percentage points for a generalized ESD many outlier procedure Technometrics 25 2 165 172 Saidi K S Teizer J Franaszek M amp Lytle A M 2011 Static and dynamic performance evaluation of a commercially available ultra wideband tracking system Automation in Construction 20 5 519 530 Shahandashti S M Razavi S N Soibelman L Berges M Caldas C H Brilakis I Zhu Z 2011 Data Fusion Approaches and Applications for Construction Engineering Journal of Construction Engineering and Management 137 10 863 869 Shahi A Cardona J Haas C West J amp Caldwell G 2012 Activity Based Data Fusion for the Automated Progress Tracking of Construction Projects Construction Research Congress ASCE Shoelson B 2013 Cascade Training GUI Retrieved Aug 01 2014 from Mathworks http www mathworks com matlabcentral fileexchange 39627 cascade training gui specify ground truth Sony
64. cal considerations Saidi et al 2011 defined the 2D and 3D measurement errors as the Euclidean distance between the coordinates estimated by the UWB system and measured with the total station They found that the average 2D and 3D errors were 0 087 m 0 010 m and 0 466 m 0 040 m respectively where the averages of the standard error of the mean are represented by or intervals As the 3D error is significantly larger than the 2D error they suggested that several sensors must be mounted at different heights at either equal or close to equal distance to each other to minimize the 3D error In addition they also noted that the 3D error decreases as tag elevation increases However they found no correlation between tag elevation and the 2D error Also they noted that the mean error decreases with the decrease in elevation whereas the standard deviation remains relatively constant at the three elevations Furthermore they noted that the error is low at the center of the coverage area which was expected based upon the manufacturer s specifications 11 Figure 2 8 UWB tags mounted at different heights nominally 1 m 2 m and 3 m Saidi et al 2011 Saidi et al 2011 conducted the second set of experiments in a lay down yard see Figure 2 9 which was for steel components of a power plant to evaluate the dynamic performance of the system under realistic construction conditions They selected an active work zone of about 1
65. ccuracy is 10096 whereas out of the two periods in which the excavator was not detected for one period the accuracy was 30 while for the other period the accuracy was 0 102 4 4 2 3 Data Fusion Module The FM was implemented in MATLAB according to the design described in Section 4 3 2 3 and Figure 4 2 The MATLAB code for the FM is attached in Appendix C Data Alignment In the data alignment stage the coordinate transformation is performed For the association of data all data have to be in the same GCS In this case study the UCS is considered the same as the GCS which is the Cartesian coordinate system as shown in Figure 4 5 Thus no transformation is required for the UWB data However the coordinates from the VM are in pixels so they should be transformed from the VCS to the GCS For this transformation firstly the pixels Xa Ya in the VCS were converted to another pixels coordinate system Xp yp to avoid the unwanted space present in the video as shown in Figure 4 7 This conversion is performed using Equation 4 5 Xp yp Xa 240 Ya 134 4 5 0 0 Xa 240 134 Xb Ya MI a Unwanted Space b Conversion Figure 4 7 Pixels Conversion After this conversion the new pixel coordinates xy yo were then transformed into the GCS using two different coordinate transformation methods Firstly the MATLAB s image location expressing technique MATLAB 2014a was used This method requires two at
66. checking the correctness of the 94 association Furthermore a threshold value for the distance between the input UM and VM positions of equipment is used to accept or reject the association Visual IDs Figure 4 3 Visual IDs for Equipment Identification The data association error is explained in Figure 4 4 The distance between the actual location and the UWB estimation is the error from the UWB system euwg whereas the distance between the actual location and the estimation based on image processing is the error from the image processing system yideo and the total error rotai is the sum of these two errors which is much larger than both of them So to avoid wrong association a threshold 6 can be defined which would be the maximum of ywp and video such that if ero gt then the association should be rejected 95 UWB Location o Total pad Cows io Actual Location Video E Video Location Figure 4 4 Association Error Explanation At the position estimation stage at each time step i the position of equipment j p is estimated by averaging the associated items from u and v In the case where both matrices do not contain locations of all equipment then the positions of the equipment are estimated using the available data For instance for equipment if the data from the UM is available but no data is available from the VM then the position of that equipment is estimated just
67. ck the movement of a roller which is operating on a construction site The duration of this test was 29 5 minutes Four UWB sensors were installed at the edges of the site covering an area of 22 98 m x 14 035 m as shown in Figure 3 18 Although two tags are enough for tracking the roller s movement four tags i e 7 S S and S4 were attached to the 58 roller to provide data redundancy as shown in Figure 3 19 b The EUR of tags was set to 8 3 Hz as total 14 tags were present in the monitored area and SIF was used with all the default settings except MRM which is set to 3 because of the wireless setting of the UWB system in this test Table 3 13 Results of Outdoor Wireless Connectivity Test Distance RF Power Line of E m Sight Connectivity Status 12 5 Connected 25 Connected 0 1 Yes 50 Connected 100 Connected 12 5 Not Connected 25 Connected 60 Yes 50 Connected 100 Connected 50 Not Connected 80 No 100 Connected Performance Analysis The data are analyzed in the 2D plane For a better analysis the duration of the test is divided into six five minute long periods The AUR of each tag is analyzed and presented in Table 3 14 It can be observed that the AUR are very different from the EUR tag S4 has the best performance whereas tag S2 has the worst performance Furthermore firstly the MDR was analyzed as shown in Table 3 15 It can be observed
68. collected from these tests For applying SCM one OC and three GCs were considered which are described in Table 3 8 The speed of rotation of boom is 13 3 sec Zhang 2010 which is converted from sec to m sec using Equation 3 4 where r is the radius of rotation and its value is 0 66 m 2nr Speed of Rotation z Sis 3 4 G33 Table 3 8 OC and GCs for Indoor Dynamic Tests I Constraint Description Value OC Speed of rotation of boom 0 16 m sec GC I Distance between Tag 1 and Tag 2 4 cm GC 2 Distance between Tag 2 and Tag 3 8 cm GC 3 Distance between Tag 1 and Tag 3 4cm The results of the SCM are shown in Figure 3 9 c and Figure 3 9 d From Figure 3 9 c it can be observed that the SCM has not enhanced the data from the wired UWB system whereas from Figure 3 9 d it is clear that the data from the wireless UWB system is enhanced by the SCM This is because the data from the wired system was already good Moreover the accuracy of the tests 1A and 1B is analyzed and compared as shown in Table 3 9 For calculating the accuracy the mean difference between the radii of the expected rotation and the UWB estimated rotation is calculated and analyzed along with its standard deviation It can be observed that for the wireless system the SCM has improved the accuracy whereas for the wired system the SCM has not produced better results Therefore in view of the analysis presented in Figure
69. derably lower than that of slim tags and their MDR is also much higher than that of the slim tags Therefore in terms of AUR and MDR the slim tags have better performance than the compact tags One reason for this lower performance can be that the compact tags have omnidirectional antenna but the slim tags have unidirectional antenna and the compact tags were held in a position that they were not facing the sensors rather they were facing up whereas the slim tags were facing the sensors Table 3 5 Update Rate and Missing Data Rate Analysis Actual Update Rate Hz Missing Data Rate ue Slim Tags Compact Tags Slim Tags Compact Tags PI 8 82 2 36 45 19 8532 P2 12 99 2 44 19 25 84 88 P3 10 66 2 17 33 70 86 59 P4 12 34 2 28 23 22 85 82 Moreover the movements of the four persons estimated by both tags were analyzed Figure 3 3 shows the movement of P1 which was almost in a straight path along the vertical axis By comparing Figure 3 3 a and Figure 3 3 b it can be observed that although both tags were in close vicinity but the movements estimated by both tags are very much different the reason for this might be the antenna type and the direction of tags as discussed earlier in this section The movement of P2 was also almost in a straight path but along the horizontal axis as shown in Figure 3 4 while P3 was moving almost in a circular path as shown in Figure 3 5 Figure 3 6 shows the movement
70. e 4 1 compares the major advantages and limitations of UWB RTLS and image processing specifically in the domain of construction management It can be observed that the 89 UWB RTLS has the disadvantages of limited update rate high missing data and relatively higher cost for monitoring multiple equipment workers in a large and densely populated construction site However these limitations are complimented by the image processing technology Similarly the limitations of the image processing technology imposed by a construction site such as having a significant amount of visual clutter variability in photometric visual content with the passage of time the presence of occluding and moving obstacles Teizer amp Vela 2009 3D localization and off line processing are complimented by the advantages of the UWB RTLS Consequently the proposed MSDF approach overcomes these limitations of each individual technology and has the potential to localize the construction equipment even when one of these data sources is not fully available or deprived Table 4 1 Comparison of UWB amp Image Processing Technologies for Construction Projects Required Features UWB Image Processing Localization 3D Mostly 2D Identification of specific equipment Yes No Real time processing Yes No Update rate Limited High Missing data High Low Coordinate system Global Pixels Multipath and radio noise effect Yes No Weather and light conditions
71. e UWB data was processed by firstly averaging each tag s data over a period of 1 second and then taking the mean of both tags averaged data of the same equipment For instance for the truck after averaging each tag s data over 1 second period the mean of the data from tags Truck I and Truck 2 was calculated Table 4 2 AUR amp MDR Analysis for MSDF Case Study Tag AUR Hz MDR Truck 1 8 37 0 14 Truck 2 8 38 0 00 Excavator 1 8 37 0 11 Excavator 2 8 38 0 00 4 4 2 2 Video Module The IP camera was configured using the Sony Network Camera SNC toolbox application Sony 2012 For video recording the RealShot Manager application Sony 2008 was used This application records the video in the cam format The recorded cam files were then converted to avi files using the same application To perform image processing the video should be splitted into images Therefore a code was written in MATLAB which takes the video file as 98 input in wmv format and split it into images according to the frame rate of the video The avi files were converted into wmv file The resolution of the resulted splitted images was 1920 x 1080 pixels For image processing the HOG technique was implemented using a MATLAB application named Cascade Train GUI Shoelson 2013 Version 1 0 of this application was used which requires Image Processing Toolbox Computer Vision System Toolbox and Control System T
72. e accuracy of the estimated locations would be affected The performance of the UWB system is also sensitive to the orientation and the measurement of the locations of sensors and the location of the calibration tags Finally it is essential to assess the environment where the UWB system would be used The RF noise present in the environment could affect the accuracy of estimated locations Furthermore the materials of objects which are to be tagged and the objects which are present in the environment have impact on the performance of the UWB system 35 Table 3 1 Factors affecting UWB System Category Factor Connection type dependent Wired Cable connections Wireless Line of sight between bridges RF frequency of bridges RF power of bridges Distance between bridges Connection type independent Tag Type Compact Slim Tag Settings Expected update rate vs Actual update rate Filtering algorithm and parameters Total number of tags used in the test Strategic placement of tags elevated tag gives better result Dilution of Precision System Settings Number of sensors Size and geometry of cell Measurement of location and orientation of sensors Quality of calibration and measurement of location of Tag Environment RF Noise Object to be tagged Metallic Non metallic Humans Objects present in the monitored area
73. e critical than the wired system As for the wired system all sensors are connected through the timing and data cables whereas for the 33 wireless system appropriate settings of wireless bridges are essential because of the additional issues related to the stability of the communication between the sensors The RF power and RF frequency of the wireless bridges should be selected according to the environment as the RF frequency might receive interference from the existing Wi Fi networks The effect of wireless bridges is investigated in detail in Section 3 3 2 1 It is also important to select the right type of tags for each environment The compact tags are suitable for tracking equipment whereas for workers slim tags are preferable Furthermore appropriate tag settings can improve the performance of the UWB system The update rate of tags 1s critical and should be selected based upon the total number of tags present in the UWB covered area For the UWB system used in this research Ubisense 2013 each second is divided into 153 time slots where the length of each time slot is 7 453 msec The highest update rate which can be selected is 33 54 Hz which requires four time slots Slot Interval 4 To achieve this update rate a maximum of four tags should be present in the UWB covered area As the number of tags increases the update rate will decrease in order to allow the system to log all tags location For example if the update rate
74. e data were not logged consistently simple averaging did not yield better results Therefore interpolation is used to fill the missing data Figure 3 23 shows the plot of the location based on the above method Figure 3 23 a can be compared with the individual tag data shown in Figure 3 22 By comparing Figure 3 23 a and Figure 3 23 b it can be observed that the location of the roller has changed over time For instance the roller was compacting vertically during period 1 and horizontally during period 2 Similar patterns were observed from other periods of the test which are not shown here for brevity 65 12 11 10 9 8 7 6 5 4 3 2 1 0 12 3 a SI b S2 12 11 10 9 8 7 6 5 4 3 2 1 0 12 3 12 11 10 9 8 7 6 5 4 3 2 1 0 12 3 c S3 d S4 Figure 3 22 Raw Data of All Tags for Period 1 Data Enhancement Although the above analysis shows some valuable information about the compaction performed by the roller there are some errors and outliers in the data To enhance the data SCM is applied The averaging period At was set to be 3 s For the correction process only one OC is used which is the speed of movement of the roller The threshold speed is calculated based on the average speed u and the standard deviation 6 of S4 as this tag has the highest AUR and the least MDR Threshold Speed u 2o 1 165 m s Along with one OC six GCs were used which are shown in Figure 3 24 These
75. e not One reason 55 for the connectivity issue can be the interference from Wi Fi signals of nearby coffee shops The wireless bridges were operated at 2 4 GHz radio frequency with 25 RF power Lessons Learned Through this experience it was learned that the wireless bridges should be operated at 5 GHz radio frequency so that the connectivity is not affected by other wireless systems and the power level should be adjusted according to the site layout and site conditions Furthermore it was also concluded that the tags were easily attachable to the equipment using magnets and they did not cause any problem during the operation of equipment UWB Sensor and Wireless Wireless Bridge Access Point fee Work Station Figure 3 15 UWB Settings for Outdoor Test 56 Figure 3 16 Compact tags with magnet for Construction Equipment a Right b Left Figure 3 17 Tag positions on excavator Investigation of the Effect of Wireless Bridges As the results of this test were not satisfactory therefore another set of tests was conducted to evaluate the impact of wireless bridges on the overall performance of wireless UWB system In this set of tests the wireless bridges are tested in indoor and outdoor scenarios with varying distances and obstacles The indoor test was conducted on the 8 floor of the EV building 57 Concordia University s Downtown Campus which contains drywalls concrete structure
76. e system software development Washington D C Department of Defence Dong X amp Naumann F 2009 Data fusion resolving data conflicts for integration Proceedings of the VLDB Endowment 2 2 1654 1655 Elmenreich W 2002 An Introduction to Sensor Fusion Research Report Vienna University of Technology Institut fur Technische Informatik Austria El Omari S amp Moselhi O 2011 Integrating automated data acquisition technologies for progress reporting of construction projects Automation in Construction 20 6 699 705 Fawcett T 2006 An introduction to ROC analysis Pattern Recognition Letters 27 8 861 874 122 Ghavami M Michael L amp Kohno R 2004 Ultra Wideband Signals and Systems in Communication Engineering John Wiley amp Sons Google 2014 3D Warehouse Retrieved June 20 2014 from https 3dwarehouse sketchup com index html Grewal M S amp Andrews A P 2008 KALMAN FILTERING Theory and Practice Using MATLAB Third ed John Wiley amp Sons Inc Gustafsson F Gunnarsson F Bergman N Forssell U Jansson J Karlsson R amp Nordlund P J 2002 Particle Filters for Positioning Navigation and Tracking IEEE Transactions on Signal Processing 50 2 425 437 Hall D L 1992 Mathematical Techniques in Multisensor Data Fusion Artech House IEEE EIA 1998 Industry implementation of international standard ISO IEC 12207 New York NY
77. e wireless UWB system for real construction environment is missing in the literature Moreover two data enhancement methods are also reviewed which would be used to minimize the erroneousness of the UWB data Furthermore it was found that several researchers have studied fusion of data from multiple sources for various aspects of construction management and their study showed that this technique has strong potential in the domain of construction management Pertaining to the limitations of the UWB technology and the potentials of the MSDF technology we propose MSDF based approach for localization of construction equipment We believe that the combined usage of the UWB technology with the image processing based equipment detection and localization can effectively locate construction equipment on construction sites by applying an accurate data fusion model 32 CHAPTER 3 EXPERIMENTAL PERFORMANCE ANALYSIS OF UWB RTLS 3 1 Introduction As discussed in CHAPTER 2 several researchers investigated the performance of the UWB RTLS for construction projects However a comprehensive research that analyzes the performance of the UWB system specifically wireless under dynamic conditions in both indoor and outdoor environments is missing in the literature Thus the objectives of this chapter are to 1 evaluate the factors that affect the performance of the UWB system 2 analyze and compare the performance of the wired and the wireless UWB sys
78. ect Association during Fusion Process with Second Transformation Method LIST OF TABLES Table 2 1 Fusion Stages amp Techniques Smith amp Singh 2006 Hall 1992 28 Table 3 1 Factors affecting UWB Systema uie en enne dose Neben ecran een sd ae idera 36 Table 3 2 Effect of Number of Sensors on the UWB System adapted from Zhang 2010 37 Table 3 3 Overview of Experimental Work esses eere 37 Table 3 4 Tag Deu ee 38 Table 3 5 Update Rate and Missing Data Rate Analysis sse 41 Table 3 6 Description of Indoor Dynamic Tests I 44 Table 3 7 AUR and MDR Analysis for Indoor Dynamic Tests I eee 46 Table 3 8 OC and GCs for Indoor Dynamic Tests I sese 47 Table 3 9 Accuracy Analysis for Tests 1A and 2A rrreneronnrrnnvrenvrerrerennesnnvseneseeresesressnesenesseresenn 49 Table 3 10 AUR and MDR Analysis for Indoor Dynamic Tests II eeesss 22 Table 3 11 AUR and MDR Analysis for Slope Test rsrrronrrrrnrrrvrnrrnnrrnrrrrnrrrernrrnrernreesnrerrrsernn 54 Table 3 12 Results of Indoor Wireless Connectivity Test rrnnarernronnrrnrrrrnrrrernrrrrernrrrrnrerrnnernn 58 Table 3 13 Results of Outdoor Wireless Connectivity Test sss 59 Table 3 14 AUR Analysis for Outdoor Dynamic Test sse 61 Table 3 15 MDR Analysis msec for Outdoor Dynam
79. eo recording The objective of the utilization of a complimentary sensory source is to overcome the limitations of individual sensor technology UWB RTLS and to improve the accuracy of the localization of construction equipment The following conclusions are drawn from the present research 1 2 3 4 5 The data from the wireless UWB system should be enhanced using a suitable data enhancement method in order to accurately track the movement of the tagged object as discussed in Section 3 3 1 2 and Section 3 3 1 3 however high MDR restricts the applicability of data enhancement methods and also degrades data The wireless UWB system has high MDR compared with the wired system The reason is that it uses only AOA estimation technique which reduces the number of readings which are required for the filter to calculate the location Additionally the wireless bridges are a vital component of the wireless UWB system and their precise configuration is essential as discussed in Section 3 3 2 1 The calibration process is less controllable in construction sites and small angular errors in calibration result in larger positioning errors due to the large scale of construction sites The results of the MSDF approach can be affected by the error propagation phenomenon as the two system components i e the UWB positioning and the image processing have some intrinsic inaccuracies Applying image processing to more frames per second
80. eory to the materials dislocation detection and found this method is well suited for this problem where both uncertainty and imprecision are inherent to the problem They also found that data fusion helps to improve the accuracy and precision of the location estimations They also indicated the potential for the proposed model to improve location estimation and movement detection Rebolj et al 2008 also presented an automated construction activity monitoring system based on a combined method consisting of three components image recognition based tracking BIM based material tracking and a communication environment supporting mobile computing They 30 found that the proposed concept is capable of ensuring timely information for well timed reactions to unexpected events on construction sites Moreover the MSDF approach has also been investigated for indoor security surveillance by Rafiee et al 2013 They presented a fully automated indoor security solution for intruder tracking that fused data from three data sources i e UWB RTLS surveillance cameras and BIM They found that the MSDF approach is suitable for indoor security applications and also appropriate for other types of applications Dibitonto et al 2011 also proposed a hybrid people tracking system based on the fusion of data from UWB and computer vision to achieve better accuracy and reliability for people tracking 2 7 2 Other MSDF Positioning Applications Lundqui
81. er and the Kalman smoother and analyzed the estimated path This corrected path is shown in Figure 2 11 a and by analyzing the corrected path they observed that the data was distributed over a wider range due to extreme noisy data points which they call outliers In this test they identified 13 points 0 3 of the total points as outliers using the Rosner s algorithm and then removed the 13 outliers They observed that although removing the outliers slightly improves the paths created by the Kalman filter and Kalman smoother the outliers between paths were not detected by the algorithm Through this analysis they concluded that the outlier algorithm should be independently applied to each path with its own Rosner s test values rather than all the paths as a whole Figure 2 10 Area Layout for Static Test in Fully Furnished Office Cho et al 2010 For the closed space dynamic test Cho et al 2010 used five pre determined straight paths see Figure 2 12 and the tag which was elevated by 104 cm was carried by a human along all the pre determined paths at a normal walking speed The location of the tag was updated every 10ms They collected data sets for four cycles and estimated and removed the outliers 15 individually by each path and each cycle They found that each path in a different cycle showed a different rate of outlier detection and on average about 9 of the points were detected as outliers for each cycle Through th
82. erpret wrong knowledge of the real world or poor business decisions can be made Therefore it 1s crucial for data integration systems to resolve conflicts from various sources and distinguish true values from false ones Dong amp Naumann 2009 Smith amp Singh 2006 investigated three major concerns that need to be addressed in order to facilitate MSDF 1 Architecture 2 Sensor management and 3 Algorithms Architecture refers to the way in which sensor nodes connect and share information sensor management refers to the way in which sensors are placed to maximize coverage of an area for different tactical goals and algorithms refers to the way in which data integration should be performed 26 Smith amp Singh 2006 and Hall 1992 explained four fusion stages for refinement of object data from raw form to meaningful information These stages are 1 data alignment 2 data association 3 position estimation and 4 identity estimation Data alignment stage aligns the data from different sources into a common frame of reference This can be conversion of coordinates from one system to another for example conversion of Cartesian coordinates to latitude longitude and height above sea level or conversion of polar coordinates to Cartesian or vice versa Data association stage compares sensory measurement and distinguishes from which target each measurement originates and classifies them Position estimation stage estimates the target s
83. es several advantages over other RTLS including long and reliable range accurate real time positioning and capability to handle the multipath issue Rodriguez 2010 However a thorough investigation of the performance of the UWB system under uncertain conditions of a construction site is still required Therefore this research is intended to realize the challenges of the construction projects and investigate the applicability of the UWB system for construction projects under dynamic conditions Furthermore the distinct nature of each construction project and the challenges they offer the uncertain and highly dynamic conditions of a construction site and the diversity of the construction equipment impose enormous challenges Therefore depending on a single 1 technology or system to deliver the required accurate information in a timely manner becomes unreliable Some research has been done to utilize multiple independent technologies under a Multi Sensor Data Fusion MSDF framework to cope with the challenges of the construction environment MSDF technique is recognized for overcoming the limitations of the individual sensing technologies by combining the sensory data from multiple sources Rafiee et al 2013 Elmenreich 2002 Luo et al 2002 Therefore this research also intends to overcome the limitations of the UWB RTLS by using image processing data as a complimentary sensory Source 1 2 Research Objectives The objectives of t
84. f object representations in the integrated data and correctness measures the correctness of data i e whether the data conform to the real world or not Dong amp Naumann 2009 Data Fusion is applied in various modern systems like air traffic control surveillance systems defense systems etc These systems are commonly developed in accordance with different industrial and governmental standards Data fusion requires dealing with simultaneous engineering processes i e one has to work with multiple developers simultaneously on the embedded software items resource management and the hardware items e g sensors and effectors over extended time Opitz et al 2004 Although this integration process can be managed by the application of formal methods these methods have some limitations too Formal methods are generally at the abstract level but systems and data integration mostly requires in depth knowledge of the systems under consideration Various formal methods and international standards have been developed to integrate data from various systems and sources Opitz et al 2004 investigated that how the software development standards can be tailored for specific data fusion processes and highlights some of the widely used international standards Opitz et al 2004 further explains that ISO IEC 12207 is one of the commonly accepted international standards which was prepared by a joint technical committee of the International Organiza
85. g C Hammad A amp Bahnassi H 2009 Collaborative multi agent systems for construction equipment based on real time field data capturing Journal of Information Technology in Construction ITcon 14 204 228 Zhang C Hammad A amp Rodriguez S 2012a Crane pose estimation using UWB real time location system Journal of Computing in Civil Engineering 26 5 625 637 Zhang C Hammad A Soltani M Setayeshgar S amp Motamedi A 2012b Dynamic virtual fences for improving workers safety using BIM and RTLS In Proceedings of the 14th International Conference on Computing in Civil and Building Engineering Moscow 126 APPENDIX A UWB SYSTEM CONFIGURATION USER MANUAL 1 Place all sensors in position minimum 3 sensors are required and power them up 2 Power up the D Link bridges and connect bridges with sensors via Ethernet Cable as shown in Figure Al Power cable wr 2 2 G Oo i 9 o P Data out v ox Network port Cross over cable Straight Cable a Figure Al Note If you are using system in an outdoor environment where it is hard to measure position of sensors w r t origin point 0 0 0 then follow instructions under section Outdoor Testing Configuration and skip point 4 3 Measure coordinates of room area write them in notepad and save as a dat file add space between the different 8 coordinates 4 Point out an origin point 0 0 0 in the room area and meas
86. ghlights the contributions and concludes the findings This chapter also includes the recommendations for the usage of the UWB RTLS on real construction sites and highlights the future research directions CHAPTER2 ITERATURE REVIEW 2 1 Introduction In this chapter the previous research on UWB RTLS and Multi Sensor Data Fusion MSDF technologies are reviewed Also the applications of these technologies in construction management are discussed This literature review is aimed to investigate the capabilities and applicability of the UWB RTLS and the MSDF techniques for improving the safety and productivity of construction projects This chapter is organized as follows Section 2 2 reviews the UWB RTLS technology Section 2 3 reviews the applications of UWB RTLS in construction management Section 2 4 reviews the data enhancement techniques for enhancing UWB data Section 2 5 reviews data fusion models Section 2 6 examines and compares the MSDF techniques Section 2 7 highlights the applications of MSDF in construction management and Section 2 8 summarizes the reviewed literature 2 2 Ultra Wideband Real Time Location System RTLS provides the information in real time about the location of assets Malik 2009 describes RTLS as a system which enables users to manage and analyze the information regarding where assets or people are located Malik further explains that an RTLS consists of the following parts 1 tags which are attached to
87. gs Tag 1 6 8 9 and 10 is more than 90 and for these tags the AUR is less than 1 Hz However for the remaining 5 tags Tag 2 3 4 5 and 7 the AUR is more than 1 Hz and the MDR is also acceptable The best performance is of Tag 3 with an AUR of 3 20 Hz and an MDR of 23 55 One explanation for this inconsistency between tags performance can be that during this period the excavator was near to the two sensors S1 and S2 and its side where the tags with the higher AUR were attached was facing these two sensors providing more visibility This explanation is also visually validated by the recorded video as shown in Figure 3 34 72 20m Sensor 3 e ug w g 9c Sensor 1 Sensor xD Figure 3 30 UWB Covered Area for Full Scale Outdoor Test For further analysis the five tags with satisfactory performance in terms of AUR and MDR are considered As during this three minute period the excavator was stationary its tags coordinates are expected to be at the same point for the whole duration Therefore statistical processing was applied to the data from these five UWB tags Table 3 18 presents the mean position and the standard deviation of the tags x and y coordinates From this table it can be noted that the standard deviation for Tag 2 is high i e an error of more than a meter in the x direction and an error of almost a meter in the y direction whereas for tags 3 4 and 5 the
88. gs the MDR is more than 50 Position 2 Tags Position 1 Figure 3 8 Design of Experiment for Indoor Dynamic Tests I Table 3 7 AUR and MDR Analysis for Indoor Dynamic Tests I 1A 1B 2A 2B Tag AUR MDR AUR MDR AUR MDR AUR MDR Hz Hz Hz Hz 1 24 05 29 26 32 31 6 92 0 00 100 00 17 10 69 41 2 28 85 15 16 29 97 13 59 15 50 56 28 16 83 54 61 3 30 05 11 61 31 44 9 30 8 23 76 40 20 57 88 04 Furthermore to analyze the movement of the RC crane s boom each tag s data were firstly averaged over a period of 500 msec and then the data of the three tags were averaged In case of the wireless test i e 2A only data from tag 2 and tag 3 were averaged as no data were logged from tag 1 The tracked rotational movement of the boom of the RC crane is shown in Figure 3 9 a and Figure 3 9 b In these figures the red circle shows the expected path of the RC crane s boom and the black line shows the actual movement of the RC crane s boom as localized by the UWB system From Figure 3 9 a it can be observed that the movement of the boom of the RC crane as tracked by the wired system followed a linear pattern However in contrast boom s movement tracked by the wireless system is too noisy and it cannot be certainly concluded that which path was followed by the boom Figure 3 9 b 46 To further improve the accuracy of the data SCM is applied to the data
89. h First Transformation Method 111 A Co Truck UWB Truck Video a Truck UWB Truck Video a Truck UWB Truck Video x Exc UWB m Exc Video x Exc UWB m Exc Video xExc UWB Exc Video 7 7 6 6 I 5 5 4 4 3 3 2 2 1 1 0 0 1 2 3 4 0 1 2 3 4 0 1 2 3 a Frame 57 b Frame 58 c Frame 59 a Truck UWB Truck Video a Truck UWB Truck Video x Exc UWB a Exc Video x Exc UWB s Exc Video 7 6 __ __ 5 4 3 2 1 0 1 2 3 4 0 1 2 3 4 d Frame 60 e Frame 61 Figure 4 13 Last 5 Frames for Data Fusion Case 1 with First Transformation Method 112 a Truck UWB a Truck Video Truck UWB a Truck Video x Exc UWB m Exc Video x Exc UWB s Exc Video IT BR E ET m EIS a Frame 22 b Frame 24 a Truck UWB 4 Truck Video Truck UWB a Truck Video x Exc UWB sExc Video x Exc UWB s Exc Video 7 5 4 3 2 1 0 0 1 2 3 4 c Frame 88 d Frame 89 Figure 4 14 Incorrect Association during Fusion Process with First Transformation Method 113 Truck UWB Truck Video Truck UWB Truck Video a Truck UWB Truck Video xExc UWB 5Exc Video xExc UWB 5 Exc Video xExc UWB 5s Exc Video a Frame 23 b Frame 45 c Frame 46 Truck UWB Truck Video 2 Truck UWB Truck Video Truck UWB Truck Video x Exc UWB s5Exc Video x Exc UWB 5 Exc Video xExc UWB s Exc Video d a 1 2 3 4 d Frame 47 e Frame 50 f Frame 56 Figure 4 15 First 6 Frames for Data Fusi
90. he independent systems through fusing data from multiple sources The uniqueness in the nature of each construction project the large size and dustiness of a construction site and the variety and density of the workforce and equipment present within it make it very challenging to deliver real time accurate information using a single technology Therefore towards the integrated systems approach and in order to overcome the limitations of the UWB RTLS which are summarized in Section 3 4 an MSDF approach is proposed in this research to ensure that the required information is available for improving the safety and productivity of a construction project The proposed MSDF approach is designed to fuse the data from two sensory data sources which are the UWB RTLS and image processing based on video recording This chapter is organized as follows Section 4 2 compares the UWB and video technologies for construction projects the proposed MSDF approach for construction projects is presented in Section 4 3 Section 4 3 2 2 presents the implementation and case study through which the proposed approach is validated and finally the summary and conclusions are given in Section 4 5 4 2 Comparison of UWB and Video Technologies for Construction Projects Each of these technologies i e image processing and UWB RTLS has some inherent limitations and the challenging nature of construction projects further limits the performance of each technology Tabl
91. he location of the tag is the intersection of the two corresponding hyperboloids as shown in Figure 2 3 Ghavami et al 2004 AOA has advantage over the TDOA as it does not require synchronization of the sensors nor an accurate timing reference Ghavami et al 2004 however TDOA requires more cabling for accurate timing reference a Slim Tags b Compact tags Figure 2 1 UWB Tags Ubisense 2013a Position Xo Yo Receiver 2 Receiver I Figure 2 2 Angle of Arrival Technique adapted from Ghavami et al 2004 The UWB system supports two modes of communication between sensors with each other and with the host computer which are the wired and the wireless as shown in Figure 2 4 The wired mode Figure 2 4 a in which all sensors are connected with the timing cables localizes the tags using both positioning techniques AOA amp TDOA whereas the wireless mode Figure 2 4 b works only with AOA since the timing cables used for estimating TDOA are replaced with the wireless bridges Position Hyperbola dj 2 Receiver I Receiver 2 Receiver 3 Hyperbola d 3 Hyperbola d 3 Figure 2 3 Time Difference of Arrival Technique adapted from Ghavami et al 2004 Timing Cable 1 I wireless Hi Bridge N N A 7 se N ad ig Wireless P j Access Point N b sv x UWB Pes N x Sensor witcl 1 y UWB Server U
92. his research are to 1 Evaluate the impact of the factors affecting the performance of wired and wireless UWB systems in construction projects through indoor and outdoor testing 2 Investigate the possibility of improving the construction equipment UWB tracking by leveraging the data from video processing 1 3 Thesis Organization This research is presented as follows Chapter 2 Literature Review this chapter reviews the literature about the UWB RTLS and MSDF technologies along with their applications in construction management Furthermore two data enhancement methods are also reviewed which are useful for improving the accuracy of the data from the UWB RTLS Chapter 3 Experimental Performance Analysis of UWB RTLS this chapter evaluates the factors that affect the performance of the UWB system and analyzes the performance of the wired and the wireless UWB systems for indoor and outdoor construction environments under dynamic conditions through several experiments Chapter 4 Fusing UWB and Video Data for Construction Equipment Localization in this chapter an MSDF based approach is proposed for the localization of the construction equipment by fusing data from two sensory data sources which are the UWB RTLS and camera The implementation of the proposed approach is also presented in this chapter along with the its validation through a case study Chapter 5 Conclusions and Future Work this chapter summarizes the present work hi
93. ibed in Table 4 8 Table 4 8 Data Association Cases Case Description 1 Both Eols detected in video module 2 One Eol detected in video module 3 No Eol detected in video module The data association was performed according to the cases described in Table 4 8 For the first case where both Eols were detected in the VM for each frame each equipment s UWB data point was associated with its nearest video data point using the k NN algorithm After this association the position of that equipment was estimated using Equation 4 8 Moreover for the second case where only one Eol was detected in the VM the detected equipment s video data point was associated with its nearest UWB data point and then the position of that equipment was estimated using Equation 4 8 whereas the position of the unassociated equipment was estimated just by its UWB data point Finally for the third case where none of the equipment was detected in the VM the positions of both pieces of equipment were estimated just by their UWB data point uf ot fet pei 4 8 These data association and position estimation processes were performed twice firstly using the VM data transformed through the first coordinate transformation method and secondly using the VM data transformed through the second transformation method Nonetheless in both cases the same UM data was used 108 4 4 3 Analysis For the analysis of the fusion results i
94. ic Test 61 Table 3 16 MDR Analysis sec for Outdoor Dynamic Test 62 Table 3 17 AUR amp MDR Analysis for Period 1 essen 74 Table 3 18 Mean amp Standard Deviation Analysis for Period I sss 74 Table 3 19 AUR amp MDR Analysis for Period 2 sese nennen 82 Table 4 1 Comparison of UWB amp Image Processing Technologies for Construction Projects 90 Table 4 2 AUR amp MDR Analysis for MSDF Case Study errrrnrnrvrnrrnrrrnnrrnnrrrrrnrrrernrersrsrrrnernn 98 Table 4 3 Description of Detector Outcomes ise aient et eet rn to tbe ihe ea orna Inna EA Rn a 101 Table 4 4 Analysis of Detector Outcomes eec eterni onn vice deen nex n 101 Table 4 5 Performance Metrics for Detector zones ec osea qi tdt liue 102 Table 4 6 Values of Attributes of Field of View cccscccccnceutseecsseccstesntsoncesdconssnstesoretedonceieseneess 104 Table 4 7 Coordinates of Control Ports eie rn tabe a e HARI RR ARS RXR IAS PURA Ra EXE gn 106 Table 4 8 Data Association CASS cio aetate y e eter eiie gehe uen EAT Tis Co a RE NEIN 108 Table 4 9 Data Fusion Cases Occurrence varccicc sccececdtvsnccondedadescvvasystvas codecsrasnastsdecsbesstantesvadesores 109 Table 4 10 Data Association Results eene tnter eite tnnt cerasnastsdedebeestaoisevadesotes 110 xil LIST OF ABBREVIATIONS 2D 3D AoA API AUR BIM CAD DoE DoP ENCS Eol EUR FM FoV fps GC
95. ideo Localization for People Tracking 4m 11 Proceedings of the Second international conference on Ambient Intelligence Springer MATLAB 2014a Expressing Image Locations Retrieved July 23 2014 from Mathworks http www mathworks com help images image coordinate systems html MATLAB 2014b Infer spatial transformation from control point pairs Retrieved August 10 2014 from Mathworks http www mathworks com help images ref cp2tform html Nazar M S 2009 A Comparative Study of Different Kalman Filtering Methods in Multi Sensor Data Fusion In Proceedings of the International Multiconference of Engineers and Computer Scientists Hong Kong IMECS Opitz F Henrich W amp Kausch T 2004 Data fusion development concepts within complex surveillance systems In Proceedings of 7th International Conference on Information Fusion Stockholm Powers D 2011 Evaluation from precision recall and F measure to ROC informedness markedness and correlation Journal of Machine Learning Technologies 2 1 37 63 Rafiee M Siddiqui H amp Hammad A 2013 Improving Indoor Security Surveillance By Fusing Data From BIM UWB And Video Jn Proceedings of the 30th International Symposium on Automation and Robotics in Construction Montreal Razavi S N amp Haas C T 2010 Multisensor data fusion for on site materials tracking in construction Automation in Construction 19 1037 1046 Rebolj D Babi
96. igure 2 17 High Level Architecture of Multi Sensor Data Fusion System 26 Figure 3 1 Test Setting Sirenin erret eee deeds eee qu EO C REA Ner d aedes 39 Fre 3 2 c Area S EUIS NS 40 Figure 3 3 Tag s Performance Comparison for Pl ea SE 42 vii Figure 3 4 Tag s Performance Comparison for P2 sss nennen 42 Figure 3 5 Tag s Performance Comparison for P3 sos erneute iste Decus iip 43 Figure 3 6 Tag s Performance Comparison for P4 ous eerie cen teris dedii ies 43 Figure 3 7 Area Settings for Indoor Dynamic Tests I essere 45 Figure 3 8 Design of Experiment for Indoor Dynamic Tests L seen 46 Figure 3 9 Performance Comparison of Wired UWB System and Wireless UWB System 48 Figure 3 10 Investigation of Impact of Dilution of Precision Phenomenon 50 Figure 3 11 Area Settings for Indoor Dynamic Tests IL eee 51 Figure 3 12 Wired and Wireless UWB System II Raw Data eene 53 Figure 3 13 Wired and Wireless UWB System II Averaged Data sess 54 Figure 3 14 Wired and Wireless UWB System II Slope Raw Data sssss 55 Figure 3 15 UWB Settings for Outdoor Test sseenvrrnrronrvvrnrnrnenvennrensrerseerarnrneneesvesnseensserssnreenee 56 Figure 3 16 Compact tags with magnet for Construction Equipmen
97. ion of boom would be 2 35 3 30 for tests 1A and 2A and 44 2 18 0 91 for tests 1B and 2B 3 fully extend the boom of the RC crane 4 position the boom to position 0 see Figure 3 8 5 rotate the boom clockwise until it reaches position 1 6 rotate the boom anti clockwise until it reaches position 2 4 16 m Sensor 2 Metallic Whiteboard X Sensor 1 M Position B Y 7 32m Position A Door Sensor 3 Sensor 4 Timing Cable Data Cable D Sensor with orientation yaw Figure 3 7 Area Settings for Indoor Dynamic Tests I Performance Comparison of Wired and Wireless UWB Systems Initially the consistency of the logged data is analyzed for both set of tests Table 3 7 shows the AUR and MDR of all three tags for both sets of tests From Table 3 7 it can be observed that the performance of the wired system is quite reasonable whereas the wireless system s results are poor In the two tests with the wired system i e 1A and 1B each tag s AUR is very close to the 45 EUR and the MDR is relatively low For example the maximum MDR is 29 26 of Tag I in test 1A However in the tests with the wireless system i e 2A and 2B the AUR is considerably lower than the EUR and the MDR is high In test 2A no data was logged for Tag 1 and for the other two tags the AUR is low whereas in test 2B the MDR for tag 3 is almost 88 It can be noted that for all two tests with the wireless system for all ta
98. is analysis they concluded that the outlier algorithm works better when applied to an individual path B Raw data Adjusted data 8 7 i Observed Points i Observed Points True Path 7 True Path 6 Path by Kalman Filter 6 Path by Kalman Filter 5 Path by Kalman Smoother 5l Path by Kalman Smoother 4 i 4 e g 3 S 2 22 1 EE 0 0 1 1 2 2 3 3 A B A meters a Raw Data Analysis with Kalman Filter b Outliers Removed Figure 2 11 Results of Open Space Dynamic Tests Cho et al 2010 To further improve the wireless UWB system s positioning accuracy Cho et al 2010 proposed an error modelling process This error model was based on the closed space dynamic test where the five straight paths were determined During developing the error model they used single regression analysis to find a best fitting line from the scattered data They separately analysed the x and y values of the collected data Firstly they compared the differences between the observed positions and the estimated positions Then they calculated the accuracy based on the distance between the two positions To estimate regression equations for each line segment each line segment corresponds to each straight path they considered several possible linear and non linear regression lines In this process they selected a line equation only if it improved the overall positioning accuracy in conjunction with the line equation of the other
99. is set to maximum but eight tags are present in the UWB covered area the update rate would automatically be decreased to 16 Hz Another concern when setting the update rate is the moving velocity of the tagged objects Objects with high velocity need more frequent updates to accurately track their traces Therefore it is essential to select a suitable number of tags with an appropriate update rate based upon their velocity Zhang et al 2012a Equation 3 1 presents the formula for calculating update rate Uodace Kate ge uu 8 1 pene ON Ede Slot Interval 7 453 Strategic placement of tags is also very important as elevated tags yield better performance Maalek amp Sadeghpour 2013 Saidi et al 2011 and the phenomenon of Dilution of Precision also has a strong impact on the location accuracy as explained in Section 2 3 Maalek amp Sadeghpour 2013 Another significant factor related to tag settings is filtering The data from the UWB sensors are filtered to remove the noise and minimize the location errors The UWB system used in this research supports four types of Information Filters IF which estimate a tag s current position by using its previous motion Ubisense Location Engine Configuration User Manual 2013a The four variants of IF are 1 information filter 2 fixed height 34 information filter 3 static information filter and 4 static fixed height information filter Each variant of the IF has a number of parame
100. ithin the site area as done in the previous tests The UWB workstation was setup on the second floor of the existing building to avoid the expected rainy 70 weather Two UWB sensors were powered by two separate power generators whereas the other two sensors were powered by cables extending from the existing building The measurement of the sensors position was done with the help of surveying team who used a total station The surveying team provided the sensors positions in the Easting and Northing Coordinate System ENCS The values were transformed to a local coordinate system by subtracting 1400 from all coordinates After the installation of the sensor panels the wireless UWB system was calibrated At that time the surveying team was not available therefore the calibration tag s position was measured using a measurement tape This measurement was not easy as the excavators and the crane were performing scheduled tasks Once this measurement was done it was tried to calibrate the wireless UWB system but the UWB sensors were unable to detect the calibration tag One reason for this was the presence of heavy construction equipment in the site and another reason was that with this UWB covered area one sensor s view was blocked by a metallic storage room Therefore to avoid this obstacle this sensor panel was relocated to another part of the fence and then the calibration was performed again After this relocation the UWB covered area was
101. ls that will be loaded for the selected area Location Cell 00001 Figure A6 Errors Cell Geometry Cell 00002 does not have an extent v Add Extent Figure A7 14 Open another software application named Location Engine Configuration Start gt All Programs gt Ubisense 2 1 gt Location Engine Configuration i Go to Log tab and check for errors ii Go to Sensors and Cells tab then load the desired area from Map tab 131 iii iv vl vii viii ix Make a new Cell by first going to the Sensor and Cells tab and the going to the Cell tab and then click on New If some unnecessary Cells are already present delete them first The Location Engine will detect the new sensors automatically and they would be under the Available Sensors section Click on the newly created cell drag the desired sensors one by one from the Available Sensors section and drop them in the space at right These sensors would now appear under the newly created cell Position the sensors one by one by adding their x y z coordinates which were measured in Step 4 by right clicking on each sensor Properties add X Y Z Choose the Master Sensor by right clicking on the Master sensor Properties Flags and check the following flags Master Timing Source All other Sensors should be Wireless Sync Right click on the Sensors one by one go the Properties Flags and check the Wireless Sync Slave flag Go to Senso
102. ment purposes They proposed a layout of an IT platform designed to facilitate automated data acquisition from construction sites to support efficient time and cost tracking and control of construction projects Furthermore they assessed the suitability of various automated data acquisition technologies i e bar coding Radio Frequency Identification RFID 3D laser scanning photogrammetry multimedia and pen based computers for their use in tracking and controlling construction activities They also proposed a model that integrates with the automated data acquisition technologies a planning and scheduling software system a relational database and using AutoCAD to generate progress reports that can assist project management teams in decision making Razavi amp Haas 2010 studied the MSDF approach for on site materials tracking in construction They used a data fusion model in an integrated solution for automated identification location estimation and dislocation detection of construction materials The data sources considered for their MSDF model were various physical sensors different location estimation algorithms location contexts from automated data collection technologies Received Signal Strength Indicator Positional Dilution of Precision time and BIM site map layout and drawings 3D models Their particular focus was dislocation detection as it can be used to detect multi handling of materials They applied Dempster Shafer th
103. n close vicinity and Tags 5 and 7 were in close vicinity therefore the data from these two pairs were averaged This processing resulted in three different data points for each second which are 74 the positions for 1 Tag 2 p2 2 Tags 3 amp 4 p3 4 and 3 Tags 5 amp 7 ps 7 The expected orientation based upon these three positions is shown in Figure 3 35 In order to estimate the orientation of the excavator a scatter plot for these data points was drawn for each second Figure 3 36 shows the scatter plots for the first 3 seconds and the last 3 seconds of the whole three minute duration Based on the visual comparison with the video see Figure 3 34 it was observed that the orientation estimated by the wireless UWB system is almost the same as the expected orientation It can be further noted that the data of Tag 2 p2 are scattered over a larger area which is also clear from its standard deviation discussed in Table 3 18 b Day 2 d Day 4 Figure 3 3 Site Conditions for Each Day 75 a Upper View b Side View Figure 3 32 Excavator Tags Positions for Full Scale Outdoor Test Excavator image is taken from Google 2014 76 UWB Covered Area x Tag2 Tag3 Tag4 x Tag5 0 Tag7 T T T T X 37 47 57 67 106 116 126 y Figure 3 33 Raw Data Analysis of Five Tags for Full Scale Outdoor Test T Figure 3 34 Excavator Position at 12 53 PM on Day 4
104. n point in time The OC is applied so that the difference between two consecutive tag data entries does not violate the maximum operational speed limit of the equipment whereas the GC is applied based upon a fixed geometric relation between any given two tags attached to a rigid body and in step 5 the data can be further enhanced by representing several tags by an intermediate point by averaging several tags data which are attached to the same rigid body The flowchart for this iterative process is shown in Figure 2 14 19 B27 Corrected Points Initial Points Averaged Points th Initial Points Y B AB die P in Averaged Points Corrected Points A d E D2 i i e 23 Initial Points Initial Points x X c Corrections based on the GCs d Averaging of several tags data Figure 2 13 Illustration of Correction Process Vahdatikhaki amp Hammad 2014 2 4 2 Optimization based Method Vahdatikhaki et al 2014 proposed a correction method which is committed to determining the minimum amount of correction applicable to each tag that will result in a pose of construction equipment with a minimum amount of violation from all GCs and OCs The assumption of this method is that the equipment is equipped with a set of UWB tags and that every rigid part of the equipment is represented by at least two tags Furthermore for the compensation of the missing or erroneous data this method performs a multi step pr
105. nd also take into account the probability distribution Yi Oi Xactual Yi Oi Vactual DRMS 2 2 1 MRSE Zea i Factual He Or vacat 4 Bear Fastua x Along with the 2D and 3D accuracies Maalek amp Sadeghpour 2013 also calculated the precision of data Equations 2 3 amp 2 4 which is the standard deviation and the offset Equations 2 5 amp 2 6 which is the distance between the average estimated locations and the actual location The relationship between offset precision and accuracy is demonstrated in Figure 2 5 n ME 2 n 2 2D Precision Jaz o vin OG Xmean 4 tier Vi Ymean 2 3 n n Lie i Xmean Yi Oi Ymean i Zi Zmean 3D Precision pw GE LEGE TE 2 4 2D Offset v actual Xmean Vactuat Ymean 2 5 3D Offset Xactual Xmean Yactual x Ymean Zactuat Zmean 2 6 Maalek amp Sadeghpour 2013 also found that the phenomenon called Dilution of Precision Langley 1999 Mahfouz et al 2008 which is related to the geometry of the cell also has a strong impact on the accuracy of the UWB system Furthermore by removing the timing cables and comparing the accuracy of the UWB system with and without the timing cables Maalek amp Sadeghpour 2013 found that the overall accuracy using only AOA measurements is less than 53 cm in 2D Figure 2 6 a and less than 63 cm in 3D Figure 2 6 b They also noted that the relative error sh
106. nd then Compute Region See Figure A4 Site Manager File View Walls Area Help a Types amp E Objects Representations Areas Cells Object Locations Geometry D add 4 MN E DEL I GK Figure A4 ii After computing the region remove the drawn line iv Save this area by clicking the Area tab and then Save Area As Give this area a name v Goto Cells tab See Figure A5 and select the saved area from the drop down list 129 Site Manager File View Help EE Types amp E Objects ga Representations Areas Object Locations Geometry vl Vil viii ix Figure A5 Make a new cell by right clicking on the Site and then clicking on the New Geometry Cell If some unnecessary Geometry Cells are already present delete them first See Figure A6 Cells column below the Area Click on the new Geometry Cell and then click on the Add Extent button at the right bottom of the window See Figure A7 A dialog box will open click on the Save button Right click on the Geometry Cell and add a Location Cell If there are other Location Cell s present delete them Check if there are any errors at the bottom of window If there are any remove all of them by selecting them one by one from the drop down list and then pressing the Remove Object button See Figure A6 bottom 130 dA Sie Eig Geometry Cell 00001 Eig Location Cell 00001 d 9 415 7 Automatically deploy services for cells Cel
107. nitially the occurrence of the fusion cases was analyzed as shown in Table 4 9 It can be observed that mostly the second case occurred 59 in which only one EoI was detected in the VM However the occurrence of the full fusion case where both equipment should be detected by both of the sensory systems is very low i e 1196 This 1s because of the limitation of the image processing technique as the accuracy for the excavator was only 1796 see Table 4 5 Table 4 9 Data Fusion Cases Occurrence Case Occurrence 1 11 2 59 3 30 As the MSDF processes were performed twice therefore the results of the DF are analyzed separately in the following sections 4 4 3 1 DF with First Transformation Method The fusion results with the first transformation method of all the 11 frames of the first case were analyzed visually as shown in Figure 4 12 and Figure 4 13 It can be observed that in all 11 frames the final position for the excavator from the UWB and the video are very near to each other Likewise for the truck in 10 frames both data points are close to each other whereas in just one frame i e Frame 45 Figure 4 12 b the distance between the data points for the truck is large This large distance is because of incorrect transformation of the VM data from the VCS to GCS as in this frame the truck is relatively far from the camera and the first transformation method resulted in a large error Furthe
108. o location and facing angle of sensors as there is no significant accuracy difference between the wood framed site and the steel framed site For the static test in fully furnished office area Figure 2 10 they obtained an average accuracy of about 41 cm at the floor level whereas when the tag was elevated by 104 cm they obtained an accuracy of 50 cm In this test layout the tag was also carried by a human which significantly affected the accuracy based on human s orientation Moreover Cho et al 2010 conducted dynamic error tests in open space and closed space indoor construction sites The primary objective of the dynamic error test was to provide a framework to minimize the positioning errors of the UWB data in a specific area in real time They compared the differences between the tag s positions estimated by the UWB system and the probable known positions As in the dynamic tests they expected more random errors as the tag was carried by a human moving with various speeds and orientations therefore to improve the accuracy of the estimated positions in real time they applied the Kalman filter algorithm They also used Kalman smoother algorithm for smoothing the data Furthermore they detected 14 multiple outliers using the Rosner s test Rosner 1975 Rosner 1983 For the open space test they moved the tag in a pre determined S shape path The tag s location was updated every 50 ms They corrected the UWB data using the Kalman filt
109. ocessing on the raw data gathered from the tags before they can be used for the pose analysis Vahdatikhaki et al 2014 explained that the process which consists of several steps to increase the accuracy of the pose estimation Figure 2 15 begins with the averaging of data over a period of time and applying interpolation for filling the missing data Then the optimization based correction is applied which has further two phases 1 identification of center of rotation and 2 identification of the required corrections The first phase of the correction ensures that the series of captured data respect the relationship with the center of rotation whereas the second 20 phase minimizes the tag s data errors in such a wat that a number of GCs and OCs of the equipment are respected Finally once the errors are corrected the pose of the equipment is identified using the corrected data Determine DC and DC DCs involved in GC DC DCic ua ji Ar D DC DC Correct the location of the DC based on E Equally distribute E between DC and DC Identify the geometric constraints between DCs Yes Jetermine total number of GC Any Standar deviation nacceptable Determine Standard Deviation of all GCs Figure 2 14 Flowchart of Iterative Correction Process Vahdatikhaki amp Hammad 2014 21 Read raw data for At Average over df
110. of the true positive cases over the 101 summation of true positive and false negative cases 2 Precision which is the percentage of correctly detected cases that are real positive and 3 Accuracy which is the percentage of correct results These metrics were calculated using Equations 4 2 4 3 and 4 4 The values for these metrics for both Eol are presented in Table 4 5 It can be noted that the precision for both Eol is very close however the accuracy is very different As discussed earlier this is because of the similarity in the color of several parts of the excavator and the background of the image Recall Tp 4 2 ecall Tp F Precision fp 4 3 recision TE 4 3 T T Accuracy P 2 4 4 T Fy Ta Fn Table 4 5 Performance Metrics for Detector Performance Metric Excavator 90 Truck Recall 17 53 75 29 Precision 85 81 Accuracy 17 64 One reason for low accuracy can be that although the video frame rate was 30 fps only one frame from each second was used for image processing To further analyze the low accuracy issue image processing was applied to 10 frames from a 1 second period in which the excavator was detected and 10 frames from a 1 second period in which the excavator was not detected The later analysis was repeated twice for two periods one second each After this analysis it was observed that for the period in which the excavator was detected the a
111. on Case 1 with Second Transformation Method 114 a Truck UWB Truck Video 8 Truck UWB Truck Video a Truck UWB Truck Video x Exc UWB Exc Video xExc UWB 5Exc Video xExc UWB 5s Exc Video a Frame 57 b Frame 58 c Frame 59 2 Truck UWB Truck Video 2 Truck UWB Truck Video x Exc UWB s5Exc Video x Exc UWB Exc Video 7 7 _ 5 5 4 4 3 3 u MG MEI i py Lo 1 1 0 0 0 1 2 3 4 0 1 2 3 4 d Frame 60 e Frame 61 Figure 4 16 Last 5 Frames for Data Fusion Case 1 with Second Transformation Method 115 a Truck UWB 4 Truck Video Truck UWB a Truck Video x Exc UWB s5Exc Video x Exc UWB Exc Video FT mE a Frame 22 b Frame 24 a Truck UWB 4 Truck Video Truck UWB a Truck Video x Exc UWB m Exc Video x Exc UWB s Exc Video 7 5 4 3 2 1 0 0 1 2 3 4 c Frame 88 d Frame 89 Figure 4 17 Correct Association during Fusion Process with Second Transformation Method 116 4 5 Summary and Conclusions In this chapter our proposed MSDF based approach for the monitoring of construction equipment is presented This chapter discussed the high level system architecture and the hardware and the software components of the proposed approach The limitations of each technology were identified along with the complimentary aspects that will justify the need for fusion The steps for the proposed MSDF method including the UWB and image pre processing data alignment data associati
112. on and position estimation have been discussed Furthermore the implementation of the proposed approach is also discussed using a case study which described the features and implementation limitations of the current prototype and demonstrated the applicability of the proposed method In the case study the UWB positioning data were better than the positioning of image processing The reasons for this are 1 the UWB data were very good with an MDR ofless than 1 and they were also processed by averaging over time and averaging over tags 2 although the video frame rate was 30 fps only one frame from each second was used for image processing and 3 the accuracy of the image processing was low because of the similarity in the background color and the color of equipment The following conclusions are drawn from this research 1 The results of the MSDF approach can be affected by the error propagation phenomenon as the two system components 1 e the UWB positioning and the image processing have some intrinsic inaccuracies Applying image processing to more frames per second and then averaging image data over specific duration of time can improve the positioning accuracy of the video data 2 The results of the MSDF approach are also dependent on the synchronization and alignment of data Our fusion success rate improved from 96 when applying the image location expressing method for the image alignment to 100 when applying the spatial t
113. on estimated by the wired and the wireless UWB systems the data collected from these tests are enhanced by firstly averaging each tag s data over a period of 1 sec and then the data of both tags were averaged Figure 3 13 shows the tracked paths of the person after the aforementioned processing By comparing Figure 3 13 a and Figure 3 13 b it can be observed that after the processing the paths of the person tracked by the wired system is more accurate than the paths tracked by the wireless system Furthermore the data from the wireless system are still noisy 52 d Wireless Tag 2 Figure 3 12 Wired and Wireless UWB System II Raw Data 33 The performance of the systems is compared for the second set of tests related to a tag sliding on the sloped rope Initially the consistency of the logged data is analyzed and presented in Table 3 11 It can be observed that for both systems i e wired and wireless the AUR is almost the same and is slightly less than the EUR One reason for the low AUR and high MDR of the wired system can be that the start point of the movement of the tag was actually out of the UWB 4 a Wired covered area see Figure 3 11 b Wi Table 3 11 AUR and MDR Analysis for Slope Test reless Figure 3 13 Wired and Wireless UWB System II Averaged Data Tag Wired Wireless AUR Hz MDR AUR Hz MDR 3 77 8 98 3 86 7
114. oolbox within MATLAB This application works in three stages which are 1 annotating the images 2 training the system and 3 detecting the Equipment of Interest EoI in the images Image Annotation For annotation images were sampled according to the video frame rate The sampling criterion was selected to be 1 fps i e 1 image out of 30 images corresponds to 1 frame from each second s data The 100 images corresponding to the first 100 seconds were annotated as a sample The sampled images were annotated by drawing a rectangular box around each Eol As in this case study there were two Eols a truck and an excavator so for each EoI the sampled images were annotated separately System Training After the image annotation the system was trained using the 100 annotated images and 750 negative images The training was also performed separately for each Eol Eol Detection For the detection of the equipment in the images the sampling criterion was selected to be 1 fps This sampling criterion was selected in order to simplify the synchronization of the UWB and video data for fusion as the UM would pass 1 location per second for each equipment to the FM as discussed in Section 4 4 2 1 Therefore out of the total test duration the last 100 seconds data were sampled for detection i e 100 images As the total test duration was 180 seconds data of 20 seconds were overlapped between annotation and detection However in order to
115. oth or not Figure 3 7 shows the UWB area settings for these tests The first set of tests was conducted with the wired UWB system whereas the second set was conducted with the wireless system Within each set the first test was conducted with the RC crane placed at the center of the UWB covered area see Position A in Figure 3 7 as described in Table 3 6 while in the second test the RC crane was placed near the edge of the UWB covered area see Position B in Figure 3 7 to evaluate the impact of the phenomenon of DoP as discussed in Section 3 2 and the effect of the surrounding objects at the two locations Furthermore in each test the RC crane was elevated from the ground by almost 1 meter in order to improve the location accuracy as discussed in Section 3 2 Table 3 6 Description of Indoor Dynamic Tests I Test Name UWB System Position 1A Wired A 1B Wired B 2A Wireless A 2B Wireless B The same tags were used in both set of tests and their EUR was set to the maximum i e 34 Hz as only three tags were present in the UWB covered area For filtering SIF was used in which the value of MRM was set to 5 for the wired tests whereas for the wireless test its value was set to 3 These tests were conducted using the following steps 1 Attach three tags to the tip of the boom of RC crane 2 Place the RC crane on a cart of height 0 9 m and position it so that the 2D coordinates of the center of rotat
116. ous Real time Location Systems RTLSs for improving the safety and productivity of construction projects When integrated with real time data analysis systems RTLS can contribute to make the construction environment smarter and safer by identifying safety hazards and inefficient resource configurations Previous research shows that the Ultra Wideband UWB technology an emerging type of RTLS is suitable for the identification and tracking of construction resources However a thorough study to evaluate the impact of the factors that affect the performance of the UWB RTLS in construction projects is still required This research investigates the performance of UWB RTLSs in indoor and outdoor environments along with the evaluation of the factors which affect their performance Moreover the harsh environment of construction sites and the complex nature of construction projects provide numerous challenges for an individual technology to deliver accurate information in a timely manner Therefore this research also proposes a Multi Sensor Data Fusion MSDF approach which leverages the benefits of video recording and image processing as a complimentary data source It was found that the UWB RTLS is an effective tool to monitor construction resources however some of the UWB data can be missing or erroneous and the quality of the data can be improved by applying a suitable data enhancement method to accurately localize construction equipment Furthermore it w
117. ove the ground and the other end was at the ground 50 7 7m r 7 p Performance Comparison At first the first set of tests is analyzed in which the movement of a person is tracked For this set primarily the consistency of logged data 1s analyzed Table 3 10 presents the AUR and MDR analysis for this set It can be noted that for the wired system the AUR is almost the same as EUR for both tags and the MDR is also 0 Whereas in the case of the wireless system the AUR is slightly less than the EUR and some data are missing Table 3 10 AUR and MDR Analysis for Indoor Dynamic Tests II Wired Wireless Tag AUR Hz MDR AUR Hz MDR 1 4 19 0 3 85 9 37 2 4 18 0 3 88 8 40 To further analyze each tag s performance with respect to location accuracy the paths based on raw data are drawn for each tag as shown in Figure 3 12 In this figure the black line shows the locations of tag estimated by the UWB system whereas the red line shows the actual path followed by the person By comparing Figure 3 12 a and Figure 3 12 c it can be observed that for tag 1 the wired system has estimated fairly good locations while the locations estimated by the wireless system are too noisy Similar observation can be made about tag 2 by comparing Figure 3 12 b and Figure 3 12 d To further improve the location of the pers
118. ows an average decrease of 114 2 in 2D accuracy and 58 09 in 3D however despite this decrease the average accuracy using only AOA measurement is still less than 1 m with 27 cm of accuracy in 2D and 37 cm in 3D Through this analysis they concluded that removing the timing cables will decrease the accuracy but the UWB system can still provide a location accuracy of less than one meter To simulate the signal blockage scenario Maalek amp Sadeghpour 2013 turned off the sensor with best Line of sight LoS In this case the location would be estimated without the best signal However this will not simulate the exact signal blockage scenario because multipath effects would not be considered which are present in the real signal blockage situation Also this simulation would represent a special case of another experiment which they conducted by reducing the number of sensors So this experiment with a turned off sensor can be considered as an experiment with seven sensors X Actual point of interest Sample location estimations from UWB Sample Mean a Offset approaches zero b Precision approaches zero c Legend Figure 2 5 Relationship between Offset Precision and Accuracy Maalek amp Sadeghpour 2013 100 100 A Ez TDOA amp AOA Bl a 80 s 2 80 4 vo i 3 2 9 3 E 40 amp 40 4 Q 204 8 20 amp Accuracy Accuracy 0 1 0 T T T m 0 0 2 0 4 o 0 0 2 0 4 0 6 m a 2
119. p right on each slave But for daisy chaining the master is connected into the input timing cable socket top right of slave 1 Any other sockets from slave 1 are then connected to the input ports top right of slave 2 and slave 3 as shown in Figure A11 Master Slave 1 Slave 2 Slave 3 Figure A11 Daisy chain for TDoA 141 Troubleshooting Connectivity Issue In order to make sure that the bridges are connected and are able to talk to each other 1 Make sure the DHCP server is running as administrator Attach a laptop to a bridge Assign following IP address to the network adapter local area connection 192 168 0 77 mask 255 255 255 0 Type following command on run box cmd Command ping 192 168 0 50 to test if the laptop has connection to the access point Command ping 192 168 0 51 to test if the laptop has connection to bridge78 Command ping 192 168 0 53 to test if the laptop has connection to bridgeD3 Command ping 192 168 0 54 to test if the laptop has connection to bridge00 on ad Qo SVIN ON oe If all ping commands return reply it means that bridges are interconnected Next step is to see if the Ubisense server laptop is connected to sensors Steps e Go to location engine configuration e Sensor status tab and check the IP address Data Logging Issue If the tags are visible in the Location Engine Software Application but are not recorded using any logger then the issue 1s in the
120. plinary knowledge in systems control systems integration signal processing artificial intelligence probability statistics and specific application area Luo et al 2002 In recent years prospective research has been conducted to explore the applicability of various techniques for multi sensor data fusion systems and its applications Gustafsson et al 2002 explored potential applications of sequential Monte Carlo methods in positioning navigation and tracking problems Smith amp Singh 2006 reviewed various approaches for target tracking Massimiliano et al 2011 applied sensor fusion for people tracking and Nazar 2009 compared two statistical estimation and noise filtering techniques based upon Kalman Filtering for multi sensor data fusion 25 Environment m i EMMI EMMI EMMI EMMI EMMI EMMI EMMI EMMI Data Fusion Process I Decision Makers Systems Humans Figure 2 17 High Level Architecture of Multi Sensor Data Fusion System 2 6 1 Techniques used for Multi Sensor Data Fusion Integration of data systems face various challenges out of which two are most common Firstly data from disparate sources are often heterogeneous Secondly different sources can provide conflicting data Conflicts can arise because of incomplete data erroneous data and out of date data Reporting incorrect data might be misleading and even harmful as the system may int
121. propriate Candidate Then at the Section 2 in Figure A9 make a new group named Sensors Then at the Section 3 in Figure A9 name this sensor as the last 2 digits of its MAC address Click on the button Add survey point at the bottom of the dialog box Repeat Step 11 for each sensor Place a tag in the middle of the area measure its distances from P1 and P2 dit amp dog and save in the calibration sheet Repeat step 10 and 11 and make a new group named Calibration Tag and name the tag Tag 1 Go to properties of a sensor and click on the button Position at surveyed and select the appropriate group and the point Repeat step 15 for each sensor For calibration follow the step 15 from the previous section At the sub step v select the tag s survey group and point rather than entering the tag s coordinates manually 136 r Survey Point Finder Filter the reference points shown Select a cell to a Select the reference points from T A which to calculate the new point Select from all sensors and survey points v Select from all sensors and survey points v Select reference point 2 gt Enter the distances from the new survey point to each of the reference points along with the height of the new point The new survey point is at one of the following 2 candidate points Check the map to confirm that one of them is correct and select it using the radio button Candidate 1 x y
122. r Status tab and check if all the sensors are working While sensors are rebooting you will not see their IP Address Once they are booted then their IP Address should be visible Check if Master Sensors light is steady green If it is not steady green then consult the Section 5 4 1 of Location Engine Configuration Manual at Page 57 to figure out the problem Note In this configuration all sensors light should be steady green when you turn on the RF Power If you are using Ubisense Software v2 1 9 go into Location Engine and press CTRL SHIFT A to get into advanced mode On the master sensor go into the properties and then to the Control tab Uncheck the Distance Error Mode and reboot all sensors Wait until the Master Sensors light becomes steady green 15 Sensors should be calibrated If you are facing calibration issues which are not covered in this manual then consult the Section 5 5 of Location Engine Configuration Manual at Page 61 The calibration steps are as follows these steps are critical i Set the radio channel and standard error for the cell Right click on the Cell containing the sensors you wish to calibrate and select Properties In the cell properties dialog set RF power to 255 this is the maximum Now select the Geometry tab and enter a standard error limit for the cell Set the appropriate values for Ceiling Floor and Max Standard Error For Lab 9 415 the values are 132 ii iii iv
123. ransformation from control point pairs method In future research an advanced image processing technique with intensive training can be applied to the video frames to improve the positioning accuracy of the video data Furthermore a more robust estimation technique can be used for position estimation stage of the MSDF 117 paradigm Finally to overcome the error propagation issue robust filtering techniques like Kalman Filter or Particle Filter can be applied to the output of each sensory data source 118 CHAPTER5 CONCLUSIONS AND FUTURE WORK 5 1 Summary of Research This research investigated the applicability of UWB RTLS for localizing construction equipment Through the literature review it was noted that the UWB RTLS is suitable for the identification and tracking of construction resources However the factors that affect the performance of the UWB system in construction environment were not specifically defined and evaluated and this issue was the motivation of this research The focus of this research was to evaluate the factors that affect the performance of the UWB system to analyze and compare the performance of the wired and the wireless UWB systems for indoor environments in a dynamic mode and to assess the performance of the wireless UWB system for outdoor construction environment under dynamic conditions Another focus of this research was to use several UWB tags to track a particular construction equipment and then
124. re 108122 are lase six digits of tag s serial number and click OK Define Tag Range and their Update Rate In Location Engine go to sub tab Tags gt go to tab Range gt click New gt specify tag IDs gt specify update rate Slower QoS and Faster QoS should be the same Note Below are the formulas for calculating update rate of tags Update Rate msec QoS 7 453 1000 Update Rate Hz QoS 7 453 1000 Of c ic Qo Update Rate Hz 7 453 For example if you want that each tag s location should be updated 10 times per second 10 Hz then gan ULL QoS i04 7453 7 gt So you have to select a value LESS THAN 13 417 Assign tags to objects In Location Engine go to sub tab Owners gt go to tab Ownership gt click New gt specify type e g Compact Tag gt specify Object e g C108122 gt enter Tag ID e g 020 000 108 122 17 For logging tags go to Desktop gt Ubisense gt UWB Logger 2 1 9 gt UWB Logger sln It will open a C software application in Visual Studio Press Ctrl F5 It will open a dialog See Figure A8 Select Object Types then select Object names and then press start Select a location to save the log file and specify the name of log file then press Save When you want to stop logging press the Stop button 18 Go to the saved log file and check the logged tags 134 EV 9 415 w Location Visualization Objects Types Objects Names E Peson CompactTag Example Vehicle
125. rence TL EUR 3000 2500 2000 1500 1000 500 Time Difference msec NLM 300 500 500 Sequence of readings c S3 3000 2500 2000 1500 1000 500 Time Difference msec 500 3000 N Nn o Time Difference msec Time Difference TL EUR FJELLA ut LJ ht 100 200 300 400 Sequence of readings b S2 Time Difference TL EUR Figure 3 20 Control Charts for Update Rate Analysis 200 300 Sequence of readings d S4 400 500 500 2 c S amp 4 amp o MN T S3 E 3 E S2 XM S o SI Update Rate Hz Figure 3 21 Cumulative Probability vs Update Rate Analysis for Outdoor Dynamic Test Figure 3 22 shows the raw data of all four tags for Period 1 From this figure it can be observed that there is very low consistency between the locations provided by each tag In conformity with the results presented in Table 3 14 and Table 3 16 it can be observed that the performance of S4 and S3 are much better than S1 and S2 Furthermore to simplify the analysis one location is calculated based on the four tags by averaging each tag over a period of 3 sec which is selected based on trial and error and then averaging all four tags As th
126. rmore the results of the data association were analyzed and presented in Table 4 10 It can be noted that the association results are fairly good with the correct association of 96 The results of the incorrect association were further analyzed visually as shown in Figure 4 14 It can be observed that all of these four frames are from Case 2 in which only the truck is detected by 109 the VM Additionally in each of these four frames the video data point of the truck was nearer to the UWB data point of the excavator due to which the k NN algorithm associated these two data points The cause of this incorrect association can be the error propagation from one component of the system to the other which eventually affected the performance of the whole system The error might have originated either from the positioning of the UWB system or from the image processing application or it can be the combined effect of the small errors from both system components Table 4 10 Data Association Results Association Occurrence Correct 96 Incorrect 4 4 4 3 2 DF with Second Transformation Method Visual analysis was performed for the results of fusion with the second transformation method results of all the 11 frames of the first case as shown in Figure 4 15 and Figure 4 16 It can be observed that in all 11 frames the data points for both pieces of equipment the excavator and the truck from the UWB and the video are clo
127. ronment and you have skipped Point 4 then follow instructions under section Outdoor Testing Configuration at Page 08 of this manual to get the coordinates of the tag Perform orientation calibration on each sensor Right click on the sensor and select Orientation Calibration Write the tag s serial number and coordinates in the corresponding fields Note If the system is unable to get tag s data then most probably either the tag is in sleep mode shake the tag to wake it up or the Activity Threshold is NOT set properly Monitor the cell to check if it is able to see the tags Right click on the cell and check the Monitor menu item If you have many tags in the cell you might want to enter a tag id into the Tag field in the monitor controls so you can watch the events for a single tag Note Check whether tags are working or not If LED on a tag is blinking then it s working 133 16 Tags should be registered in the system Concept Every TAG is associated with an OBJECT and each object is of a specific TYPE i ii iii iv Create Type In Site Manager go to sub tab Types gt go to tab Type gt click New gt specify a name e g Person Vehicle Compact Tag etc and click OK Note Types cannot be deleted so do NOT make unnecessary types Create Object In the Site Manager go to sub tab Objects go to tab Object click New gt specify a Type and enter name e g For type Compact Tag specify C108122 whe
128. s tracking technologies to improve their accuracy as well Cho et al 2010 concluded that the accuracy of the UWB system is low in dynamic and closed space situation than in static and open space situation Furthermore through this study they validated that although the accuracy of the wireless UWB system is lower than the wired one the wireless UWB system is still capable of tracking mobile assets in indoor construction sites with 17 an accuracy of about 50 cm in static condition and 65 cm in dynamic condition for a highly congested closed space The wireless UWB system was used in this research however this analysis does not take into account the conditions of outdoor construction environment as the tests were conducted in indoor environments Zhang et al 2012a proposed a post processing method to improve data quality and transform the location data into useful information that can be used for near real time decision support systems Moreover they tested the UWB system using the proposed method to estimate the pose of a crane and concluded that the pose of the crane boom can be estimated in near real time using the UWB system Although they performed a thorough analysis using the UWB system they only used the wired UWB system Vahdatikhaki amp Hammad 2014 proposed a framework based upon the integration of UWB RTLS with a simulation model of construction operations in order to enhance the simulation model continuously by capt
129. s Candidate 1 at the Section 1 highlighted in Figure A9 The x and y should be zero and z should be its height Now make a group named Survey Points at the Section 2 highlighted in Figure A9 Then name it as P1 at the Section 3 highlighted in Figure A9 Then click on the button Add survey point at the bottom of the dialog box Enter the coordinates of P2 as Candidate 1 at the Section 1 highlighted in Figure A9 Its y value would be zero and the x value would be its distance from P1 and the z would be its height Then select the group Survey Points from the drop down list at Section 2 highlighted in Figure A9 Name it as P2 at the Section 3 highlighted in Figure A9 Then click on the button Add survey point at the bottom of the dialog box From the Section 4 highlighted in Figure A9 select the group Survey Points from the first drop down list then select P1 from the corresponding drop down list Then select the group Survey Points from the second drop down list and select P2 from the corresponding drop down list 11 Now enter the distance from the Sensor 1 to P1 dy in the Reference 1 distance box at 12 13 14 15 16 17 the Section 5 highlighted in Figure A9 Similarly enter the distance from the Sensor 1 to P2 d21 in the Reference 2 distance box and enter its height in the Height box Two different set of coordinates would appear as Candidate 1 and Candidate 2 at the Section 1 highlighted in Figure A9 select the most ap
130. s was drawn for each second Figure 3 42 shows the scatter plots for 3 seconds from the first minute and 3 seconds from the last minute As the excavator was not stationary during this three minute period so the 83 actual positions of the excavator during the first minute and the last minute are not the same These actual positions are shown in Figure 3 43 Based on the visual comparison with the video see Figure 3 43 it was observed that the orientation estimated by the wireless UWB system is almost the same for the first minute as the expected orientation however for the last minute the estimated orientation is not similar One reason for this error in the UWB data can be that during the last minute the excavator was at the edge of the UWB covered area P2 Pi P3 Figure 3 41 Schematic View of Orientation of Excavator Excavator image is taken from Google 2014 Moreover to assess the accuracy of the wireless UWB system further analysis was conducted based on the angle between the lines formed by joining Tags 1 and 2 and Tags 1 and 3 as shown in Figure 3 41 The expected angle between these two lines is 90 The actual angle 04 was calculated using the individual UWB tag s data which were averaged over a period of 1 sec The mean and the standard deviation of the error e between 0 and 0 0 0 were calculated which were found to be 8 09 and 34 8 respectively From these values it can be observed that
131. se to each other Moreover by comparing Figure 4 12 b and Figure 4 15 b it can be observed that for Frame 45 unlike for the fusion process performed with the first transformation method the data points for the truck are close to each other Finally the results of the data association were analyzed and it was noted that the association results are very good with the correct association of 100 Hence it is concluded that the second transformation method has produced much better results as the correct association rate improved from 96 to 100 Moreover the results of the four frames in which the data were incorrectly associated using the first transformation method were analyzed as shown in Figure 4 17 By comparing Figure 4 14 and Figure 4 17 it can be observed that in each of these frames the location of truck estimated by the VM is changed 110 Co Truck UWB Truck Video Truck UWB Truck Video a Truck UWB Truck Video x Exc UWB m Exc Video x Exc UWB m Exc Video x Exc UWB s5Exc Video 7 7 6 6 I 5 5 4 4 3 3 2 2 1 1 0 0 1 2 3 4 0 1 2 3 4 0 1 2 3 a Frame 23 b Frame 45 c Frame 46 Truck UWB Truck Video 2 Truck UWB Truck Video Truck UWB 4 Truck Video x Exc UWB a Exc Video x Exc UWB s Exc Video x Exc UWB s5Exc Video 7 7 6 6 __ __ __ 5 5 4 4 3 3 2 2 1 1 0 0 1 2 3 4 0 1 2 3 4 0 1 2 3 d Frame 47 e Frame 50 f Frame 56 Figure 4 12 First 6 Frames for Data Fusion Case 1 wit
132. st 2011 fused information from various sensors for estimating the motion of a vehicle and the characteristics of its surroundings He studied and compared various maps in particular road maps which make use of the fact that roads are highly structured and allows relatively simple and powerful models to be employed He showed how the information of the lane markings obtained by a front looking camera can be fused with inertial measurement of the vehicle motion and radar measurements of vehicles ahead to compute a more accurate and robust road geometry estimate Furthermore he showed how radar measurements of stationary targets can be used to estimate the road edges and applied a special filter to the radar data for constructing a representation of the map of the stationary objects around the vehicle For tracking moving targets he focused on the extended targets i e targets which potentially may give rise to more than one measurement per time step He also introduced a framework to track the size and shape of a target Ciftcioglu et al 2007 described sensor data fusion in autonomous perceptual robotics They represented a visual perception by a probabilistic model where the model receives and interprets visual data from the environment in real time The perception obtained in the form of 2D measurements is used for the robot navigation They processed the visual data in a multi resolution form via wavelet transform and optimally estimated
133. static tests were conducted in four different types of indoor building spaces open space wood framed building site steel framed building site and fully furnished office area For assessing the accuracy of the wireless UWB system they used the difference in the Euclidean distance between the tag s known position and the UWB estimated position For the open space test they elevated the tag by 35 cm to give it a better LoS and obtained an accuracy of 17 02 cm For the wood framed building site they obtained an accuracy of 46 cm with the tag elevated by 94 cm whereas when the tag was on the floor they obtained an accuracy of 63 cm They also collected the data with the same test layout where a human was carrying the tag with an elevation of 130 cm For this data the accuracy of the wireless UWB system was 59 cm In this case they expected to obtain better accuracy as the tag was more elevated but the accuracy dropped down Therefore they concluded that the human body has negative affect on the quality of communication between a tag and the sensors They also found that this conclusion is aligned with the literature Welch et al 2002 For the steel framed building site when the tag was on the floor they obtained an accuracy of 56 cm whereas when the tag was elevated by 104 cm better accuracy was obtained i e 38 6 cm From the results of the wood framed site and the steel framed site tests they concluded that accuracy seems to be more sensitive t
134. t sees 57 Figure 3 17 Tag positions on eXCdVAtOL id eade piene o gre IS EREY RE Bo Nee hae o RA RAM ERE NS uid 57 Figure 3 18 Area Settings for Outdoor Dynamic Test seen 60 Figure 3 19 Tracked Roller for Outdoor Dynamic Test essere 60 Figure 3 20 Control Charts for Update Rate Analysis sese 64 Figure 3 21 Cumulative Probability vs Update Rate Analysis for Outdoor Dynamic Test 65 Figure 3 22 Raw Data of All Tags Tor Period Lus iiis GRANSKES 66 Figure 3 23 All Tags Averaged with At 3 sec se 67 Figure 3 24 Geometric Constraints for Outdoor Dynamic Test esse 67 Figure 3 25 Results of Simplified Correction Method All Tags Averaged for At 3 sec 68 Figure 3 26 Results of Simplified Correction Method S3 amp S4 Average for At I sec 68 Figure 3 27 Results of Optimization based Method All Tags Averaged for At 3 sec 69 viii Figure 3 28 Site View on May 22 2014 before Visit essere 70 Figure 320 UVB Sensor TAL SSN 72 Figure 3 30 UWB Covered Area for Full Scale Outdoor Test sees 73 Figure 3 31 Site Conditions for Bach Day ssxissscssssescensavseianss nett Ante dann RE HE RISE ESM AR INE EAMAR e SR HEAR URA 75 Figure 3 32 Excavator Tags Positions for Full Scale Outdoor Test Excavator image is taken from Google DONA asa cos ete ve
135. t the cause of the high MDR This observation along with the results of Section 3 3 1 2 where the wired and wireless systems were compared under exactly the same condition can suggest that the high MDR is an inherent limitation of the wireless UWB system Table 3 16 MDR Analysis sec for Outdoor Dynamic Test Des ion Missing Data Rate sec S1 S2 S3 S4 1 5 37 33 31 00 10 33 6 33 2 5 44 33 34 67 13 33 3 00 3 5 23 67 23 00 17 67 2 00 4 5 46 67 34 67 7 67 3 00 5 5 43 67 41 00 11 33 11 33 6 4 5 27 41 35 56 16 67 2 22 Total 29 5 37 34 33 28 12 77 4 69 Figure 3 21 shows the cumulative probability of occurrence of the various update rates with which the data was logged for each tag This figure should be read from right to left For instance it can be observed that for tag S1 90 of the data was logged with an update rate of 62 more than 2 Hz while 8096 of the data was logged with an update rate of more than 4 3 Hz However for tag S4 9096 of the data was logged with an update rate of more than 7 7 Hz while 80 of the data was logged with an update rate of more than 7 9 Hz Similar analysis can be observed for tag S2 and S3 63 Time Difference TL EUR 3000 N Nn 2000 1500 1000 Time Difference msec 500 LI L MULL 0 100 200 300 400 500 Sequence of readings a S1 Time Diffe
136. tems for indoor environments in a dynamic mode and 3 evaluate the performance of the wireless UWB system for outdoor construction environment under dynamic conditions This chapter is organized as follows The factors that affect the performance of the UWB system are discussed in Section 3 2 Section 3 3 demonstrates the indoor and outdoor dynamic experiments which were conducted to analyze the performance of the UWB system for construction projects and the conclusions and recommendations are presented in Section 3 4 3 2 Factors affecting the UWB System s Performance Setting up a UWB system requires several crucial steps including the placement of sensors measuring the coordinates of sensors configuration of network connection and configuration of various software components These steps are detailed in Appendix A Furthermore the settings of the wired and the wireless UWB systems are not similar as shown in Figure 2 4 For the wired system Figure 2 4 a all sensors are connected with each other through Ethernet cables for the estimation of TDoA and all sensors are also connected to a network switch for the communication of sensors with the server Whereas for the wireless system Figure 2 4 b each sensor is connected to the wireless bridge in order to communicate with the computer server Several factors affect the performance of the UWB system which are listed in Table 3 1 In terms of system settings the wireless system is mor
137. ters that control the behavior of the filter out of which 12 parameters are common to all types of IF One of the 12 common parameters is Minimum Reset Measurements MRM which represents the minimum number of supporting measurements required A single measurement can be either an azimuth an elevation or a TDOA between two sensors Ubisense 2013 In the wired setting if two sensors see a tag there will be five measurements azimuth and elevation from each sensor plus a TDOA whereas in the wireless setting if two sensors see a tag there will be four measurements just the azimuth and elevation from each sensor as there is no TDOA in the wireless setting Another important factor that affects the overall performance of the UWB system is the number of UWB sensors used to monitor the area Table 3 2 summarizes the effect of number of sensors and location method on the UWB system s performance It can be observed that for the wireless UWB system AOA only as the number of sensors increases 2 or more the system would be able to estimate the location of tags more accurately and with only one sensor the system can only provide 2D position without extra information However for the wired system TDOA AOA two sensors are enough for a good 3D location accuracy Moreover the size and geometry of the sensor cell are very critical It is preferable that the sensor cell would be in a square like geometry If a sensor cell has a poor geometry th
138. the 5th floor of the Engineering and Visual Arts Complex EV building of the Concordia University s Downtown campus The objectives of this test were to 1 evaluate the performance of the wireless UWB system for indoor security applications and 2 evaluate the performance of the two types of the UWB tags which are the compact tags and the slim tags Figure 2 1 Four persons were involved in this test each having two UWB tags one compact and one slim Table 3 4 lists the IDs of the tags used by each person and Figure 3 1 a shows the position of the tags carried by each person It can be observed from Figure 3 1 a that the slim tag was placed near the body of the person whereas the compact tag was kept slightly away from the body also the compact tag was slightly more elevated than the slim tag The Expected Update Rate EUR of the tags was set to 16 Hz because only 8 tags were present in the monitored area The data were collected for 3 minutes Static Information Filtering SIF was used with the value of MRM set to 5 The covered area by the UWB sensors was 8 48 m x 8 72 m as shown in Figure 3 2 As this test was conducted using the wireless UWB system so the UWB sensors were connected to the wireless bridges as shown in Figure 3 1 b Figure 3 1 c and Figure 3 1 d show the positions of UWB sensors in the monitored area Video of the test was also recorded using a Sony IP PTZ Camera as an extra source of information to validate the
139. the assets 2 sensors which reads the tags data 3 location engine which is a software used to localize the tags 4 middleware which connects the location engine data with a software application and 5 end user software application UWB is a special type of RTLS which transmits and receives short duration pulse of Radio Frequency RF energy Lee et al 2009 Malik 2009 explains that UWB is a carrier less radio technology that uses wide bandwidth i e exceeding 500 MHz or 20 percent of the arithmetic center frequency whichever is lower and is normally used in short range wireless applications Malik 2009 also explained that UWB based positioning has several advantages over other RTLS technologies which includes high accuracy better performance in challenging RF environments no interference from other RF systems and relative immunity to multipath fading The immunity to multipath fading is because UWB pulses are narrow and occupy the entire UWB bandwidth The early applications of UWB technology were primarily related to radar The UWB system used in this research is developed by Ubisense Group PLC Ubisense 20132 This UWB system comprises of the following parts 1 tags for monitoring assets 2 sensors for reading tags 3 timing cables or wireless bridges for the connectivity of sensors with each other and with the host computer 4 location engine for calculating tag s position using various techniques and
140. the utilization of the collected data was useful for safety decision making However they suggest that the image based safety assessment method has some limitations which can be overcome by integrating tracking devices such as UWB or GPS with the image based safety assessment method Shahi et al 2012 incorporated a UWB RTLS system to track activities in a construction project in order to automate the estimation of the construction projects progress Although the scope of their research was limited to ductwork HVAC and piping activities on the project but their proposed model is scalable to a complete construction project Also they showed a comparison of concrete steel and piping projects and noted that the number of changes occurring during construction may be significantly higher for piping and industrial projects in comparison to steel or concrete building construction They also found that although automated object recognition 29 and material tracking techniques that use the 3D Computer Aided Design CAD or Building Information Modeling BIM model as a priori information may be accurate for concrete or steel structures they may be ineffective for tracking the progress of piping and many other mechanical and electrical services carried throughout most of the projects El Omari amp Moselhi 2011 integrated various automated data acquisition technologies to collect data from construction sites required for progress measure
141. tion for Standardization ISO and the International Electrotechnical Commission IEC IEEE EIA 1998 Some other standards include AQAP 160 standard which is used in NATO projects and is a modification of the ISO IEC 12207 standard and the DoD STD 2167A is widely used for military projects DoD 1988 According to ISO IEC 12207 the development process is divided into the following steps Opitz et al 2004 IEEE EIA 1998 1 System requirements analysis and design 2 Software requirements analysis 3 Software architectural design 4 Software detailed design 5 Software coding and testing 6 Software integration and qualification testing and 7 System integration and qualification testing Opitz et al 2004 developed a simplified development model as shown in Figure 2 16 based on the standard and suggested that at each step traceability of 23 requirements consistency test coverage appropriateness conformance and feasibility should be ensured FEY j 975 z 23 S 9 S oO eg ao Qos a ae an 2 Oss E EE O vo ann 5 3 2 4 H P o 5 d o 2 T oc rh 52 7 z e e o 1 9 S 3 2 2 E D 3 oO YS S 77 og oS 5 ox o z8 D 23 SB 5 o c o S g 2 Y E 2 Lad 3 2 5 E ov 5 lt o oS eg 2 ro 09 7 o6 amp c 5 D 8 HW Figure 2 16 The development model Opitz et al 2004 2 6 Multi Sensor Data Fusion Multi Sensor Data Fusion MSDF also kno
142. to combine data from all tags to estimate the pose of the tracked equipment This approach enhanced the data and smoothens the tracking of movement of the equipment Furthermore efficient data enhancement methods are also applied in several tests to minimize the errors in the UWB data Through the analysis of the performance of the UWB system mainly the wireless version in the uncertain conditions of construction environment it was noted that some limitations are imposed by the harsh nature of construction environment on the performance of the UWB system In an effort to overcome the limitations of the UWB RTLS this research also proposed an MSDF based approach which leverages the benefits of the video recording and the image processing as a complimentary data source It was observed that the limitations of the UWB system are complimented by the image processing and also vice versa Therefore the proposed MSDF approach is designed to ensure that the required information is available for accurately localizing construction equipment by fusing data from two sensory data sources which are the UWB RTLS and image processing based on video recording 119 5 2 Research Contributions and Conclusions Our main contribution in this research was the evaluation of impact of factors that affect the performance of the UWB RTLS in construction environment under dynamic conditions and to leverage the features of a complimentary data source i e vid
143. towards 1 using more UWB sensors in a cellular architecture to monitor large construction sites 2 applying a more efficient image processing technique with extensive training in the MSDF model to maximize the effectiveness of fusion and 3 applying robust filtering techniques in the MSDF model to filter out noisy readings from the UWB and the image processing systems 121 REFERENCES Chi S amp Caldas C H 2012 Image Based Safety Assessment Automated Spatial Safety Risk Identification of Earthmoving and Surface Mining Activities Journal of Construction Engineering and Management 138 3 341 351 Cho Y K Youn J H amp Martinez D 2010 Error modeling for an untethered ultra wideband system for construction indoor asset tracking Automation in Construction 19 43 54 Ciftcioglu O Bittermann M amp Sariyildiz I 2007 Sensor Data Fusion in Autonomous Robotics In Proceedings of the Second International Conference on Innovative Computing Information and Control Dalal N amp Triggs B 2005 Histograms of oriented gradients for human detection n Proceedings of the 2005 IEEE Computer Society Conference on Computer Vision and Pattern Recognition CVPR 05 pp 886 893 Dibitonto M Buonaiuto A Marcialis G Muntoni D Medaglia C amp Roli F 2011 Fusion of radio and video localization for people tracking n Ambient Intelligence 258 263 DoD 1988 DOD STD 2167 Defens
144. tributes of the Field of View FoV of the camera which are 1 the upper and lower bounds along the X axis in GCS and 2 the upper and lower bounds along the Y axis in GCS The measurement process of these attributes is shown in Figure 4 8 In Figure 4 8 the xWorldLimits shows the upper and lower bounds of the X axis whereas the yWorldLimits shows the upper and lower bounds of the 103 Y axis The values of these attributes for this case study are shown in Table 4 6 As the origin of the image axis is at the top left corner of the image and the origin of the GCS is at the bottom left corner of the room therefore to solve this issue the values of yWorldLimits were inverted by selecting the values of upper and lower bounds of Y axis as negative maximum i e 5 7 and 0 respectively and later on the absolute value of the transformed y coordinate yacs was used for data association Furthermore this method scales the image according to the provided attributes of the FoV as shown in Figure 4 9 Through visual validation of results it was noted that the GCS coordinates provided by this method were satisfactory when the truck was at position T and the excavator was at position E However when the truck is at position T or moving from positions T to T then the transformed coordinates have larger error Table 4 6 Values of Attributes of Field of View Axis Upper Bound Lower Bound X 1 3 5 Y 5 7 0 ImageE
145. ure coordinates of sensors w r t origin point As an example the coordinates of sensors of Lab 8 415 are stated below i Sensor 1 Master 00 11 CE 00 3A EF X 0 76 Y 0 226 Z 1 71 ii Sensor 2 00 11 CE 00 3A F9 X 3 26 Y 0 22 Z 1 68 iii Sensor 3 00 11 CE 00 3A D3 X 3 91 Y 5 62 Z 1 9 iv Sensor 4 00 11 CE 00 3B 00 X 0 18 Y 6 27 Z 2 33 5 Login to Laptop For ThinkPad there is no username or password and for HP Laptop use the following login info 127 10 11 Username umroot Password umroot 7 Disable Wi Fi of laptop Connect the bridge with laptop via straight Ethernet cable Check the IPv4 address of laptop Control Panel gt Network and Sharing Center gt Local area connection OR Ethernet gt Properties gt IP Version 4 It should be as below IP 10 133 0 1 Subnet Mask 255 255 0 0 Locate the DHCP server Desktop gt Ubisense 2 PC DHCP Server Right click on the dhepsrv exe file and run as administrator Ping the following IP Addresses Press Window Key R and then in the dialog box write ping 10 133 0 237 t to check connectivity of sensors with computer i 10 133 0 237 MAC D3 ii 10 133 0 240 MAC EF iii 10 133 0 241 MAC F9 iv 10 133 0 242 MAC 00 Connectivity diagram is shown in Figure A2 If there is any problem in connectivity then consult the Troubleshooting section at the end of this manual Ubisense Platform Control DHCP server
146. uring motion information about the truck and excavator Their proposed framework provides a method for capturing processing analysing filtering and visualizing the equipment states along with enhancing the accuracy of the equipment state identification The data processing is done by considering the equipment specific geometric and operational constraints Although their proposed framework is tracking technology independent and can work with various types of RTLS technologies they used wired UWB system in an indoor environment to demonstrate the feasibility of their proposed framework Rodriguez 2010 investigated the utilization of UWB system in improving productivity and safety of construction projects by collecting data from a construction site and organizing them into useful information needed for management They found that UWB is an effective tool to monitor construction resources because it provides accurate information in real time 2 4 Data Enhancement Methods Two data enhancement methods are reviewed which are used for enhancing the data from the UWB system These methods are 1 Simplified Correction Method SCM and 2 Optimization based Method OM Both of these methods are based on Operational Constraints OCs and Geometric Constraints GCs where OCs limits the tags movement e g moving too 18 fast and GCs relates different tags with respect to the geometric consistency of the equipment e g fixed distance
147. uted based on the previous state estimate whereas in the measurement update process this prior estimate is combined with direct measurements of the state coming from other sensors thus obtaining the new updated state estimate Nazar 2009 27 Table 2 1 Fusion Stages amp Techniques Smith amp Singh 2006 Hall 1992 Fusion Stages Techniques 1 Data Alignment Coordinate Conversion Nearest Neighbor Joint Probabilistic Data Association 2 Data Association Lagrangian Relaxation Artificial Neural Networks Fuzzy Logic Kalman Filter Particle Filter 3 Position Estimation Multiple Model Algorithms Multiple Resolutional Filtering Artificial Intelligence Approaches Bayesian Inference Dempster Shafer D S Rule Artificial Neural Networks 4 Identity Estimation Expert Systems Voting and Summing Approaches Distributed Classification Grewal amp Andrews 2008 discussed that within the domain of MSDF KF is exclusively used for two purposes 1 estimation and 2 performance analysis of sensors For estimation KF allows to estimate the state of dynamic systems with certain types of random behavior by using information from sensory sources while for performance analysis of sensors KF helps to determine which type of sensors would perform better for a given set of design criteria These criteria are typically related to estimating accuracy and system cost
148. vae tesi im uc suene d Ed Lec T Mia canal her 76 Figure 3 33 Raw Data Analysis of Five Tags for Full Scale Outdoor Test 77 Figure 3 34 Excavator Position at 12 53 PM on Day 4 ssssssssseeeeeeeeeenen ene 78 Figure 3 35 Schematic View of Orientation of Excavator Excavator image is taken from Google NI ne 78 Figure 3 36 Scatter Plots for Orientation of Excavator Period 1 79 Figure 3 37 Angle Calculation for Accuracy Assessment Excavator image is taken from G ogle 2014 sut enteros tee tpe btt aene AEG edet tdt due epa Ted de ERA dS teu 80 Figure 3 38 Actual Angle 0 Calculation for Accuracy Assessment sssessseee 81 Figure 3 39 Error Distribution for Accuracy Assessment Period I sess 81 Figure 3 40 Tracked Movement of Excavator for Period 2 seen 83 Figure 3 41 Schematic View of Orientation of Excavator Excavator image is taken from Goose 2014 esse cose cosa ie due de ir ub ce RD QN de diese gie t iut diae pce ase neue toe vut ul Que idus 84 Figure 3 42 Scatter Plots for Orientation of Excavator Period 2 sess 85 Figure 2 43 Excavator Position wy isoisiectea tete iain P SERE IGEROR S Mr usb qud ten baa deside stein 86 Figure 3 44 Error Distribution for Accuracy Assessment Period 2 esses 86 Figure 4 1 Proposed Approach Overview
149. view ecosistema ees 1 12 Research Objectives oet eo tuc Eb peo eia ut NE 2 1 3 Thesis Organizatia ass 2 CHAPTER 2 VITERA TURE REY IEW utarte eerte rect ATOKE ENTO EDA EOK A Pec ia dev ed 4 2 NNN 4 2 2 Ultra Wideband Real Time Location System sse 4 2 3 Applications of UWB RTLS in Construction Management seen 7 24 JBDapbnhlancemenelVlethoda sso d o EQ euis use east E ca 18 24 1 Simplified Correction Method oio oet rodeo epa a red don ar eed od 19 2 4 2 Optimization based MOUHOQ Lauder 20 25 DAA 22 2 6 Multi Sensor Data Fusion uskadd nete nice eve pete cede in a oec Se eng 24 2 6 1 Techniques used for Multi Sensor Data Fusion 26 2 7 Applications of Multi Sensor Data Fusion srrrsrvrrnnvrrnrnrrrnrrnnrrnnrernrrrerrernsesnnresnrsrrrnernn 29 2 7 1 Applications in Construction Management essen 29 2 7 2 Other MSDF Positioning Applications sssesssseseeeeeneeneeeneeneee 31 20 Summary aces fata vias a Ana dace d deat pu ree aset e PUR a Rig NU UN Ua x ae AG Rao E Ra EUR 32 CHAPTER 3 EXPERIMENTAL PERFORMANCE ANALYSIS OF UWB RTLS 33 31 roe 33 3 2 Factors affecting the UWB System s Performance sese 33 3 Experimental WOEK se esce dat e Ne e SN 37 550 B r Dynamic Ter 38 3 3 2 Outdoor Dynamic Wet massenes seedet P tu ag ed EEE iih 55 3 4 Summary Conclusions and Recommendations sss 87 CHAPTER 4 FUSING UWB AND
150. wired UWB system is comparable However in contrast the performance of the wireless system varies with respect to the position of the tracked object as shown in Figure 3 10 b Similar conclusion can be drawn by comparing the AUR and MDR of these tests presented in Table 3 7 The AUR and MDR of tests 1A and 1B are somehow similar whereas for tests 2A and 2B the performance of the wireless UWB system in terms of AUR and MDR is better when the RC crane was at Position B From this analysis it is concluded that the phenomenon of DoP does not strongly affect the performance of the wired UWB system in indoor environments however the performance of the wireless UWB system is affected by DoP One reason for this difference in performances is that the wireless UWB system only uses AoA technique to estimate the position of the tagged object and when the tagged object is not in the middle of the UWB covered area then the estimation based on angles is not accurate whereas the wired UWB system estimates using TDoA technique in addition to the AoA technique so the angle calculation is affected by DoP but the calculation based on time difference can produce accurate results Table 3 9 Accuracy Analysis for Tests 1A and 2A Mean radius difference cm Standard ve ie radius difference Tag 1A 1A 2A 2A 1A 1A 2A 2A without with without with without with without with SCM SCM SCM SCM SCM SCM SCM SCM Tag I 18 12
151. wn as Sensor fusion refers to the integration of sensory data from multiple sensors to provide more reliable and accurate information This fusion of information reduces overall uncertainty and offers potential advantages including but 24 not limited to redundancy correctness reliability and thus increases the accuracy with which the environment is observed by the system Multiple sensors increase reliability of the system by providing redundant and timely information about the environment This redundant information from multiple sensors allows efficient environment perception that is hard to achieve using single sensor Multiple sensors also provide more timely information as a result of either the actual speed of operation of each sensor or the parallel processing that may be achieved as part of the integration process Also in this scenario even when a sensor is deprived the system is still capable of compensating lacking information by reusing data obtained from other sensors Luo et al 2002 Figure 2 17 shows a high level architecture of MSDF It can be observed that sensors perceive the environment through the transfer of Energy Material Wealth Mass and Information EMMI Langford 2012 Then through EMMI sensors transfer data to the fusion process which then converts the data into meaningful information which is then available to the decision makers MSDF is a rapidly evolving research area and requires multidisci
152. xtentinWorldX 8 xWorldLimits 2 10 2 3 E 5 6 7 8 9 10 EIET ImageSize 2 2 0 0 ImageExtentinWorldY 4 WorlkdLimits 1 5 Figure 4 8 Measurement of Attributes of Field of View adapted from MATLAB 2014a 104 Figure 4 9 Coordinate Transformation Using First Method Therefore in order to minimize the location errors during the coordinate transformation process another transformation method was used which was based on MATLAB s spatial transformation from control point pairs MATLAB 2014b This method requires two pairs of coordinates for 4 control points in order to calculate the transformation matrix and then to scale the image according to the transformation matrix The first pair for each control point is its VCS coordinates and the second pair is its GCS coordinates The 4 control points provided to the method are shown in Figure 4 10 their values in cm for their both pairs are shown in Table 4 7 and the scaled image is shown in Figure 4 11 Then the VCS coordinates of the center point of the rectangle of each Eol xy yp from Equation 4 5 were transformed according to the scaled image and saved as Xc Yc As the GCS coordinates of the control points were provided in cm 105 therefore the resulted transformed coordinates Xe y were also in cm After analysis of the transformed coordinates X Yc it was found that there is some systematic error Equations 4 6 and 4 7 were used

Download Pdf Manuals

Related Search

Related Contents

SUSE Linux Enterprise Desktop  Sistema PX Manuale Tecnico di Installazione e Programmazione  Travel Honey Watch  Mode d`emploi Comment mettre à jour vos disponibilités  SB425 User Manual - Legacy Power Conversion  intext:Bedienungsanleitung filetype:pdf  1241 WS3  RTR970B-PRO FTR 970B-PRO  User's Manual  Serbatoio prefabbricato di pompaggio tipo ABS Sanimat  

Copyright © All rights reserved. Failed to retrieve file