Bus Arrival Time Reliability Analyses and Dynamic Prediction Model Based on Multi-source Data by Ling Shi A thesis submitted in partial fulfillment of the requirements for the degree of Master of Science In Transportation Engineering Department of Civil and Environmental Engineering University of Alberta ©Ling Shi, 2016 Abstract In recent years, with increased urbanization and vehicle ownership, congestion levels have been increasing in urban areas although new infrastructure is being developed to meet the increasing demand. An understanding of bus service reliability is necessary to develop strategies that help transit agencies provide better services. In addition, a reliable and accurate vehicle arrival prediction system can help in making public transportation more attractive. This study first aims to quantify and determine temporal and weather related bus arrival time variability in Edmonton. A multinomial logit model is developed and estimated, which relates early, late and on-time bus arrivals to weather, temporal and operating characteristics. The model results show that the probability of on-time failures increases during PM peak periods, as buses progress further along their routes and under adverse weather conditions. Secondly, a proposed bus tracing algorithm and bus arrival time prediction algorithm are then applied to predict bus travel time using Global Positioning System (GPS) data and Vehicle Detect System (VDS) data, also, a regression model based on the factors found from the multinomial logit model is applied as a comparison. A case study is conducted on one selected bus routes in Edmonton, to evaluate the performance of the proposed algorithm in terms of prediction accuracy. The results indicate that the proposed algorithm is capable of achieving satisfactory accuracy in predicting bus arrival time. ii Acknowledgement This thesis could not have been completed without the help and support of many people, only a few of whom are listed below. I would like to thank my committee members for their guidance. I am especially grateful to my supervisor, Dr. Qiu, who has always encouraged me to do my best during my MSC program at the University of Alberta. Thanks to my team members in the Centre for Smart Transportation which has a wonderful research and collaboration atmosphere. I would like to thank Dr. Liu, Dr. Zhang and Jiangcheng Li. Their instruction provides significant support not only on the research ideas, but also on the research techniques. I also would like to express thanks to the City of Edmonton for providing data for this study. The contents of this paper reflect the views of the authors and not necessarily the view of the City of Edmonton. Finally, I wish to express appreciation to my parents and girlfriend. Only with their support and understanding can I finally finish my graduation study. Thanks for their loving consideration and great confidence in me through all these years. iii TABLE OF CONTENTS CHAPTER 1 INTRODUCTION ............................................................................................................................... 1 1.1 Background ........................................................................................................................................................ 1 1.2 Problem Statement ............................................................................................................................................. 2 1.3 Research Scope ................................................................................................................................................... 4 1.4 Structure of Thesis ............................................................................................................................................. 4 CHAPTER 2 LITERATURE REVIEW ................................................................................................................... 6 2.1 Transit Service Reliability ................................................................................................................................. 6 2.2 Transit Schedule Adherence Analysis .............................................................................................................. 7 2.2.1 Weather Factors .............................................................................................................................................. 7 2.2.2 Operational Factors......................................................................................................................................... 8 2.2.3 Temporal Factors .......................................................................................................................................... 10 2.3 Bus Arrival Time Prediction ........................................................................................................................... 11 2.3.1 Historical Data Based Models ....................................................................................................................... 11 2.3.2 Regression Models ......................................................................................................................................... 12 2.3.3 Kalman Filtering Models ............................................................................................................................... 12 2.3.4 Artificial Neural Network Models .................................................................................................................. 13 2.4 Summary of Literature Review .......................................................................................................................... 13 CHAPTER 3 DATA PREPROCESSING ............................................................................................................... 16 iv 3.1 Test Site Description ........................................................................................................................................ 16 3.2 Data Description ............................................................................................................................................... 17 3.2.1 APC Data ....................................................................................................................................................... 18 3.2.2 Weather Data ................................................................................................................................................. 19 3.2.3 GTFS Data ..................................................................................................................................................... 20 3.2.4 VDS Data ....................................................................................................................................................... 22 3.2.5 Road Network Data........................................................................................................................................ 24 3.3 Data Integration ............................................................................................................................................... 25 3.3.1 Spatial Data Representation .......................................................................................................................... 25 3.3.2 Uniform LRS Establishment ........................................................................................................................... 26 3.3.3 Topology Relationship Establishment ............................................................................................................ 26 CHAPTER 4 EVALUATION OF FACTORS AFFECTING BUS ON-TIME PERFORMANCE.................... 30 4.1 Introduction ...................................................................................................................................................... 30 4.2 Methodology ..................................................................................................................................................... 30 4.3 Results and Discussions ....................................................................................................................................... 32 4.3.1 Temporal Analysis ......................................................................................................................................... 33 4.3.2 Impact of Weather .......................................................................................................................................... 34 4.3.3 Multinomial Logit Analysis ............................................................................................................................ 38 CHAPTER 5 BUS ARRIVAL TIME PREDICTION ............................................................................................ 43 5.1 Introduction ...................................................................................................................................................... 43 5.2 Methodology ..................................................................................................................................................... 44 v 5.2.1 Real-Time Bus Tracing .................................................................................................................................. 44 5.2.2 Real-Time Bus Trajectory Reconstruction ..................................................................................................... 46 5.2.3 Regression Model .......................................................................................................................................... 48 5.3 Results and Discussions ....................................................................................................................................... 49 5.3.1 Studied Route ................................................................................................................................................. 49 5.3.2 Travel Time Analysis...................................................................................................................................... 51 5.3.3 Bus Arrival Time Prediction .......................................................................................................................... 55 5.3.3.1 Results based on GPS Data .................................................................................................................... 55 5.3.3.2 Results based on Regression Model ....................................................................................................... 59 5.3.3.3 Results Comparison ............................................................................................................................... 60 CHAPTER 6 CONLUSIONS ................................................................................................................................... 61 6.1 Research Summary and Limitation ................................................................................................................ 61 6.2 Future Work ..................................................................................................................................................... 62 REFERENCES .......................................................................................................................................................... 64 vi LIST OF TABLES Table 1 Multinomial Logit Model for Bus On-time Performance (Asymptotic t values in parentheses)................................................................................................................................... 41 Table 2 Arrival Time Result on Dec 2nd ....................................................................................... 55 Table 3 Arrival Time Result on Dec 3rd........................................................................................ 56 Table 4 Arrival Time Result on Dec 4th ........................................................................................ 57 Table 5 Goodness of Fit Statistics ................................................................................................ 59 Table 6 Statistics of Bus Travel Time Estimation Models ........................................................... 60 Table 7 Average MAPE of the Prediction Models ....................................................................... 60 vii LIST OF FIGURES Figure 1 Service Day Map of ETS (City of Edmonton 2015d) .................................................... 17 Figure 2 APC System.................................................................................................................... 18 Figure 3 Samples of APC Data ..................................................................................................... 19 Figure 4 Samples of Hourly Weather Data ................................................................................... 19 Figure 5 Samples of Daily Weather Data ..................................................................................... 20 Figure 6 Samples of Bus Stop GTFS Feed (City of Edmonton 2015b)........................................ 21 Figure 7 Samples of Bus Route GTFS Feed (City of Edmonton 2015a)...................................... 21 Figure 8 Samples of Bus Trip GTFS Feed (City of Edmonton 2015c) ........................................ 21 Figure 9 The Real Time Bus Data of Edmonton (City of Edmonton 2015e) ............................... 22 Figure 10 Samples of VDS Data ................................................................................................... 23 Figure 11 Loop Detector Locations in Edmonton ........................................................................ 23 Figure 12 Samples of Road Network GIS Data ............................................................................ 24 Figure 13 Entity-Relationship of Data Integration ....................................................................... 29 Figure 14 Visualization of POA under Clear Weather ................................................................. 34 Figure 15 Visualization of SDV under Clear Weather ................................................................. 34 Figure 16 Visualization of POA on Weekday .............................................................................. 35 Figure 17 Visualization of SDV on Weekday .............................................................................. 36 Figure 18 Visualization of POA on Saturday ............................................................................... 36 Figure 19 Visualization of SDV on Saturday ............................................................................... 37 Figure 20 Visualization of POA on Sunday ................................................................................. 37 Figure 21 Visualization of SDV on Sunday ................................................................................. 38 Figure 22 The Workflow of Real-Time Bus Tracing ................................................................... 45 viii Figure 23 Result Sample of Real-Time Bus Tracing .................................................................... 46 Figure 24 The Workflow of Real-Time Bus Trajectory Reconstruction ...................................... 47 Figure 25 The Map of the Route 33 (City of Edmonton 2015f) ................................................... 50 Figure 26 The Study Segment of the Route 33 ............................................................................. 51 Figure 27 Loop Detector Locations along the Study Segment ..................................................... 51 Figure 28 Box Plot of Travel Time from Dec 2nd to Dec 4th ........................................................ 54 Figure 29 Box Plot of AM/PM Travel Time from Dec 2nd to Dec 4th .......................................... 54 Figure 30 Prediction Trajectories vs Reference Trajectories on Dec 2nd ...................................... 56 Figure 31 Prediction Trajectories vs Reference Trajectories on Dec 3rd ...................................... 57 Figure 32 Prediction Trajectories vs Reference Trajectories on Dec 4th ...................................... 58 Figure 33 Estimated Versus Actual Travel Time ......................................................................... 60 ix List OF ABBREVIATION AFC Automatic Fare Collection ANN Artificial Neural Networks APC Automatic Passenger Counters APTS Advanced Public Transportation Systems AVL Automatic Vehicle Location AVT Automatic Vehicle Track DATS Disabled Adult Transit Service ETS Edmonton Transit System GIS Geographic Information System GPS Global Position System GTFS General Transit Feed Specification ITS Intelligent Transportation Systems LRS Linear Referencing System LRT Light Rail Train MAPE Mean Absolute Error PA Percentage of Arrivals within On-Time Threshold RMSE Root Mean Squared Error SD Standard Deviation of Arrival Time Variance TSR Transit Service Reliability VDS Vehicle Detect System x