University Transportation Centers UTC Conference for the Southeastern

  • Slides: 32
Download presentation
University Transportation Centers (UTC) Conference for the Southeastern Region November 16 -17, 2017 Reitz

University Transportation Centers (UTC) Conference for the Southeastern Region November 16 -17, 2017 Reitz Union, University of Florida Campus A Methodology to Assess the Quality of Travel Time Estimation Based on Connected Vehicle Data Md Shahadat Iqbal Ph. D Fellow Department of Civil and Environmental Engineering, Florida International University, Miami, FL 33174

Introduction Connected Vehicle (CV) • Can communicate with each other • Multidimensional benefit •

Introduction Connected Vehicle (CV) • Can communicate with each other • Multidimensional benefit • Safety • Mobility • Environment • Provide information to the traffic management center FIGURE 1: Connected Vehicle technology demonstration (Courtesy: Qatar Mobility Innovations Center) 2

Introduction Objective Data Preparation Achievement 1 Achievement 2 Achievement 3 Conclusion Travel Time q

Introduction Objective Data Preparation Achievement 1 Achievement 2 Achievement 3 Conclusion Travel Time q Travel time is an important performance measurement for the transportation system q It could be calculated utilizing different technologies such as vehicle reidentification technologies including Bluetooth, and Wi-Fi, GPS data, and Probe vehicle technology q The accuracy and reliability of the estimated travel time is an important indicator in order to use it for different purposes q Objectives: Develop a methodology to assess the quality of travel time estimation based on CV data 3

Introduction Objective Data Preparation Scenario 1 Scenario 2 Conclusion 4

Introduction Objective Data Preparation Scenario 1 Scenario 2 Conclusion 4

Introduction Goal & Objectives Data Preparation Scenario 1 Scenario 2 Conclusion CV Data Emulation

Introduction Goal & Objectives Data Preparation Scenario 1 Scenario 2 Conclusion CV Data Emulation Vehicle Trajectory Data ü Real Word data ü Simulation Conversation Tool TCA User Input Connected Vehicle Data 5

Introduction Goal & Objectives Data Preparation Scenario 1 Scenario 2 Conclusion Time (T 1),

Introduction Goal & Objectives Data Preparation Scenario 1 Scenario 2 Conclusion Time (T 1), speed (S 1), position(P 1) 6

Introduction Goal & Objectives Data Preparation Scenario 1 Scenario 2 Conclusion Time (T 2),

Introduction Goal & Objectives Data Preparation Scenario 1 Scenario 2 Conclusion Time (T 2), speed (S 2), position(P 2) 7

Introduction Goal & Objectives Data Preparation Scenario 1 Scenario 2 Conclusion Time (T 3),

Introduction Goal & Objectives Data Preparation Scenario 1 Scenario 2 Conclusion Time (T 3), speed (S 3), position(P 3) 8

Introduction Goal & Objectives Data Preparation Scenario 1 Scenario 2 Conclusion Vehicle Trajectory Data

Introduction Goal & Objectives Data Preparation Scenario 1 Scenario 2 Conclusion Vehicle Trajectory Data Vehicle ID Time Speed Position … V 1 T 1 S 1 P 1 . . . V 1 T 2 S 2 P 2 … V 1 T 3 S 3 P 3 … V 1 T 4 S 4 P 4 … . . . … Trajectory Conversion ü ü Message type (BSM) CV proportion Data loss rate … CV Data 9

Introduction Goal & Objectives Data Preparation Scenario 1 Scenario 2 Conclusion CV Data Emulation

Introduction Goal & Objectives Data Preparation Scenario 1 Scenario 2 Conclusion CV Data Emulation • CV data • Available from few testbeds and pilot projects • CV proportions are very low Vehicle Trajectory Data ü Real Word data ü Simulation • Emulation of CV data • Trajectory Conversion Algorithm (TCA) • J 2735 Standard by Society of Automotive Engineers (SAE) International Conversation Tool TCA User Input Connected Vehicle Data 10

Introduction Goal & Objectives Data Preparation Scenario 1 Scenario 2 Conclusion Vehicle Trajectory Data

Introduction Goal & Objectives Data Preparation Scenario 1 Scenario 2 Conclusion Vehicle Trajectory Data • Real world vehicle trajectory data • Collected under the Next Generation Simulation (NGSIM) program • Simulation • Microsimulation (VISSIM) 11

Introduction Goal & Objectives Data Preparation Scenario 1 Scenario 2 Conclusion NGSIM Data •

Introduction Goal & Objectives Data Preparation Scenario 1 Scenario 2 Conclusion NGSIM Data • Arterial location • Peachtree Street, Atlanta • PM peak traffic (15 Minute) • Freeway location • U. S. Highway 101 (Hollywood Freeway), Los Angeles, California • AM peak traffic(15 Minute) • Both 0. 5 mile in length 12

Introduction Goal & Objectives Data Preparation Scenario 1 Scenario 2 Conclusion Simulation Data •

Introduction Goal & Objectives Data Preparation Scenario 1 Scenario 2 Conclusion Simulation Data • Microscopic simulation • VISSIM • Arterial section • Glades Road, Boca Raton, FL • One mile section is considered • Four intersection • Freeway section • Similar to NGSIM freeway • Calibrated according to HCM 13

Introduction Goals and Objectives Data Preparation Scenario 1 Scenario 2 Conclusion 14

Introduction Goals and Objectives Data Preparation Scenario 1 Scenario 2 Conclusion 14

Introduction Goal & Objectives Data Preparation Scenario 1 Scenario 2 Conclusion Why travel time

Introduction Goal & Objectives Data Preparation Scenario 1 Scenario 2 Conclusion Why travel time estimation is challenging? q Vehicle ID change over time q Vehicle tracking is not possible Distance travel by ID 59 Distance travel by ID 77 ID: 59 ID: 77 Silence period 15

Introduction Goal & Objectives Data Preparation Scenario 1 Scenario 2 Conclusion Travel Time Calculation

Introduction Goal & Objectives Data Preparation Scenario 1 Scenario 2 Conclusion Travel Time Calculation • 16

Introduction Goal & Objectives Data Preparation Scenario 1 Scenario 2 Conclusion Sources of Stochasticity

Introduction Goal & Objectives Data Preparation Scenario 1 Scenario 2 Conclusion Sources of Stochasticity • Identify the specific vehicles on the link that are equipped with CV devices • Time when the temporary ID for each vehicle changes • Monte Carlo simulation ID: 59 ID: 77 17

Introduction Goal & Objectives Data Preparation Scenario 1 Scenario 2 Conclusion Assessment of the

Introduction Goal & Objectives Data Preparation Scenario 1 Scenario 2 Conclusion Assessment of the Accuracy and Reliability of Travel Time Estimation • Travel time accuracy and reliability measures • • • MAPE MAD RMSE SDPE 95% Absolute Percentage Error 85% Absolute Percentage Error • Each of the run represents a single day Name Mean Absolute Percent Error (MAPE) Description Equation Average absolute percentage difference between the estimate and ground truth Mean Absolute Deviation (MAD)/Mean Absolute Error Average of errors Root Mean Squared Error (RMSE) Square root of the average of the squared error The Standard Deviation of Percentage Error (SDPE) Square root of the average of the squared percentage errors 18

Introduction Goal & Objectives Data Preparation Scenario 1 Scenario 2 Conclusion Assessment of the

Introduction Goal & Objectives Data Preparation Scenario 1 Scenario 2 Conclusion Assessment of the Accuracy and Reliability of Travel Time Estimation • 19

Introduction Goal & Objectives Data Preparation Scenario 1 Scenario 2 Conclusion Regression Analysis Results

Introduction Goal & Objectives Data Preparation Scenario 1 Scenario 2 Conclusion Regression Analysis Results for Travel Time Estimation Error Shapiro-Wilk normality test Adjusted Rfor model squared value residual (p-value) Mean of the residual -18 (*10 ) β 0 β 1 -02 10 (* ) β 2 -03 10 (* ) β 3 -05 10 (* ) β 4 -07 10 (* ) R-squared value MAPE 4. 963 -15. 1 4. 52 -6. 18 2. 89 0. 991 0. 986 0. 998 1. 33 SDPE 6. 328 -15. 7 4. 78 -6. 53 3. 04 0. 993 0. 989 0. 767 -2. 67 RMSE 2. 611 -15. 5 4. 67 -6. 29 2. 87 0. 993 0. 989 0. 612 1. 34 MAD 2. 064 -15. 5 4. 75 -6. 60 3. 13 0. 991 0. 987 0. 100 -7. 47 95%** 10. 24 -9. 80 1. 61 -9. 99 - 0. 990 0. 987 0. 931 -1. 33 85%*** 8. 654 -14. 6 4. 35 -6. 02 2. 85 0. 988 0. 981 0. 626 -4. 54 MAPE 13. 957 -9. 70 1. 79 -1. 27 - 0. 995 0. 994 0. 153 0. 70 SDPE 18. 954 -10. 6 1. 97 -1. 38 - 0. 994 0. 993 0. 114 0. 09 RMSE 14. 984 -10. 4 1. 92 -1. 33 - 0. 994 0. 993 0. 194 1. 11 MAD 11. 012 -9. 62 1. 76 -1. 24 - 0. 995 0. 994 0. 198 0. 22 95%** 38. 590 -10. 5 1. 85 -1. 24 - 0. 989 0. 986 0. 401 3. 97 85%*** 26. 523 -11. 1 2. 14 -1. 51 0. 993 *All coefficients (β) are significant at 95% confidence interval of t-test ** 95% absolute percentage error curve represents the 5% highest errors among the travel time estimates *** 85% absolute percentage error curve represents the 15% highest errors among the travel time estimates 0. 990 0. 120 4. 21 Arterial Freeway Error 20

Introduction Goal & Objectives Data Preparation Scenario 1 Scenario 2 Conclusion Travel Time Accuracy

Introduction Goal & Objectives Data Preparation Scenario 1 Scenario 2 Conclusion Travel Time Accuracy Measures Freeway Arterial 21

Introduction Goal & Objectives Data Preparation Scenario 1 Scenario 2 Conclusion Travel Time Accuracy

Introduction Goal & Objectives Data Preparation Scenario 1 Scenario 2 Conclusion Travel Time Accuracy Measures MAPE (Freeway) MAPE (Arterial) 22

Introduction Goal & Objectives Data Preparation Scenario 1 Scenario 2 Conclusion Probable Travel Time

Introduction Goal & Objectives Data Preparation Scenario 1 Scenario 2 Conclusion Probable Travel Time Accuracy Measures by Year MAPE (Freeway) * Max = Maximum Min = Minimum MAPE (Arterial) 23

Introduction Goal & Objectives Data Preparation Scenario 1 Scenario 2 Conclusion Findings • For

Introduction Goal & Objectives Data Preparation Scenario 1 Scenario 2 Conclusion Findings • For freeway • CV data could be useful for the very first year of CV implementation • For Arterial • It would take three years • Considering MAPE <10% • It would take six years • Considering 95% absolute percentage error <10% • Different agencies can adopt different accuracy measurements 24

Introduction Goals and Objectives Data Preparation Scenario 1 Scenario 2 Conclusion 25

Introduction Goals and Objectives Data Preparation Scenario 1 Scenario 2 Conclusion 25

Introduction Goal & Objectives Data Preparation Scenario 1 Scenario 2 Conclusion Effect of Demand

Introduction Goal & Objectives Data Preparation Scenario 1 Scenario 2 Conclusion Effect of Demand Variation on the Accuracy of Estimated Travel Time • 26

Introduction Goal & Objectives Data Preparation Scenario 1 Scenario 2 Conclusion Estimated Travel Time

Introduction Goal & Objectives Data Preparation Scenario 1 Scenario 2 Conclusion Estimated Travel Time Error Equations for Simulation Data (Arterial) • Investigated degree of saturations are • 0. 3, 0. 6, 0. 7, 0. 8, and 0. 9. • Considered MP in percentage • 1, 2, 3, 4, 5, 10, 20, 30, 40, 50, 60, 70, 80, and 90 • 100 Monte Carlo run( In total 7, 000 simulation run) 27

Introduction Goal & Objectives Data Preparation Scenario 1 Scenario 2 Conclusion Estimated Travel Time

Introduction Goal & Objectives Data Preparation Scenario 1 Scenario 2 Conclusion Estimated Travel Time Error Equations for Simulation Data (Freeway) • investigated v/c ratios are • 0. 36, 0. 5, 0. 72, 0. 86, and 1. 01 • Considered MP in percentage • 1, 2, 3, 4, 5, 10, 20, 30, 40, 50, 60, 70, 80, and 90 • 100 Monte Carlo run( In total 7, 000 simulation run) 28

Introduction Goal & Objectives Data Preparation Scenario 1 Scenario 2 Conclusion Findings • Accuracy

Introduction Goal & Objectives Data Preparation Scenario 1 Scenario 2 Conclusion Findings • Accuracy at different year after the CV mandate • For freeway: year two • For arterial: year six • Accuracy increases with the increase of the sample size (v/c ratios) • At higher v/c ratios • Congestion • higher variations in travel time between vehicles Results in a lower accuracy v/c Year AMPE 1 2. 66 2 2. 07 3 1. 82 4 1. 46 5 1. 30 0. 1 10 0. 95 15 0. 91 20 0. 94 25 0. 95 1 1. 81 2 1. 23 3 0. 97 4 0. 61 5 0. 46 0. 5 10 0. 11 15 0. 07 20 0. 09 25 0. 11 1 2. 5 2 1. 92 3 1. 66 4 1. 3 5 1. 15 1 10 0. 8 15 0. 76 20 0. 78 25 0. 8 * Error more than the minimum acceptable threshold Freeway Travel Time Accuracy (%) SDPE 13. 2* 9. 64 8. 1 5. 86 4. 89 2. 47 1. 92 1. 91 1. 93 12. 84* 9. 28 7. 74 5. 5 4. 53 2. 12 1. 56 1. 55 1. 57 11. 82 8. 26 6. 72 4. 47 3. 50 1. 09 0. 53 0. 55 AMPE 13. 60* 12. 45* 11. 94* 11. 16* 10. 80* 9. 70 9. 14 8. 95 8. 89 7. 74 7. 23 6. 45 6. 10 4. 99 4. 43 4. 24 4. 18 13. 08* 11. 93* 11. 42* 10. 65* 10. 29* 9. 18 8. 62 8. 43 8. 37 Arterial SDPE 13. 2 9. 64 8. 09 5. 84 4. 87 2. 45 1. 88 1. 87 1. 89 12. 86* 9. 29 7. 75 5. 5 4. 53 2. 11 1. 54 1. 53 1. 55 11. 85 8. 29 6. 74 4. 5 3. 52 1. 1 0. 54 0. 52 0. 54 29

Introduction Goals and Objectives Data Preparation Scenario 1 Scenario 2 Conclusion 30

Introduction Goals and Objectives Data Preparation Scenario 1 Scenario 2 Conclusion 30

Conclusion CV Mandate to all new vehicles 1/1/2020 2021 2022 2023 2024 2025 2026

Conclusion CV Mandate to all new vehicles 1/1/2020 2021 2022 2023 2024 2025 2026 2027 2028 1/1/2022 1/1/2026 Travel time estimation in freeways Travel time estimation in Arterial 2029 2030 Timeline 31

Thank you! Question? 32

Thank you! Question? 32