INNOVATIONS IN TRAVEL TIME RELIABILITY MONITORING EMPOWERMENT FROM EMERGING TECHNOLOGIES

Transcription

1 INNOVATIONS IN TRAVEL TIME RELIABILITY MONITORING EMPOWERMENT FROM EMERGING TECHNOLOGIES Presentation to the: Research and Innovative Technology Administration (RITA) October 19, 2011 George List, PhD, PE North Carolina State University

2 Main Topics The metrics of interest Understanding the data Building the Textbook Creating TT-PDFs Causes and counteractions Predicting travel times Future advances

3 Desired Time of Arrival (DTA) Window Metrics of Interest Destination Location Reliability On-time or not Disutility, risk, and regret Desired Time of Arrival (DTA) Window Origin Time Actual Time of Arrival (ATA) P(late), P(on-time), P(early) Consistency Variance in travel time Disutility Disutility Function On-time independent Actual Time of Arrival Individuals and averages Causal Factors Most important ones Separate or together Differences by percentile

4 4 Understanding the Data Driver behavior Operating plan impacts Trends in time Trends across time Trends by condition Spot rates Space rates Individual vehicles Averages Segments Routes

5 5 Travel Times on Freeways 400 Travel Times Monday Jan Travel Times Thursday Jan 20, :00:00 2:24:00 4:48:00 7:12:00 9:36:00 12:00:00 14:24:00 16:48:00 19:12:00 21:36:00 0:00:00 0 0:00:00 2:24:00 4:48:00 7:12:00 9:36:00 12:00:00 14:24:00 16:48:00 19:12:00 21:36:00 0:00:00

6 Cumulative Probability Cumulative Probability Cumulative Probability TT-CDFs for Freeways 120% CDFs for Transition-to-Peak Travel Rates 100% 80% 60% 40% 20% All Days 13-Jan 20-Jan 22-Jan 24-Jan 0% Travel Rate (sec/mi) CDFs for Off-Peak Travel Rates CDFs for Peak Travel Rates 120% 120% 100% 100% 80% All Days 80% All Days 60% 13-Jan 60% 13-Jan 40% 20-Jan 22-Jan 40% 20-Jan 22-Jan 20% 24-Jan 20% 24-Jan 0% Travel Rate (sec/mi) 0% Travel Rate (sec/mi)

7 Trip Time (min) Travel Times on Arterials Trip Times (6 to 3) / Week of North Multi-day trips NY Shorter trips with stops :00 4:48 9:36 14:24 19:12 0:00 4:48 Time of Observation at 3 Actual travel times

8 CDFs for Arterials (Histograms) 8

9 9 Developing the TT-PDFs State-of-the-Practice State-of-the-Art Processing Steps Average Rate Estimator Average Segment Times TT 1b TT-PDF Inference Segment TT-PDFs TT 2b Incidence Matrix Estimator Segment TT-PDFS and Incidence Matrices TT 3b Route Route TT Simulator TT- PDFs Data Types Volumes/ Occupancies TT 1a Measured Rates TT 2a Measured Segment Travel Times TT 3a Measured Segment and Route Travel Times TT 4a Traffic Data Sources Infrastructure-based Single Loop Dual Loops AVI Some AVL Vehicle-based AVL

10 10 Developing TT-PDFs and TR-PDFs Generating PDFs and Measures of Interest Data Type Enhancement or Type #1 Type #2 Type #3 Type #4 Metric Single Loops Double Loops AVI AVL Passage Times Not applicable Not applicable Use signal strength or Use passage times for bounce-back time Lat/Lon locations Average Spot Rates Spot Rates for Individual Vehicles Average Times or Rates for Segments Segment IV-PDFs Incidence Matrices AVG-PDFs for Segments or Routes IV-PDFs for Routes Use occupancy, flow, and assumed vehicle length Cannot be obtained Combine adjacent sensor spot rates Use average times or rates and IV-PDF typical of the traffic conditoins Base on field studies or similar segment-tosegment flow conditions Add estimated segment or route times or rates Simulation based on IV- PDFs and Coincidence Matrices Directly computed by the sensor Could be obtained Combine adjacent sensor spot rates Use average times or rates and IV-PDF typical of the traffic conditoins Base on field studies or similar segment-tosegment flow conditions Add estimated segment or route times or rates Simulation based on IV- PDFs and Coincidence Matrices Not needed Use signal strength or bounce-back times Determine from adjusted IV-PDFs Adjust the observed IV- PDFs to account for unequipped vehicles Use equipped vehicles on adjacent segments Compute from segment or route IV-PDFs Use equipped vehicles or simulation based on IV- PDFs and Coincidence Matrices Not needed Use GPS speeds at Lat/Lon locations Determine from adjusted IV-PDFs Adjust the observed IV- PDFs to account for unequipped vehicles Use equipped vehicles on adjacent segments Compute from segment or route IV-PDFs Use equipped vehicles or simulation based on IV- PDFs and Coincidence Matrices

11 11 Developing Passage Times Especially for Bluetooth data (Last-First)/2? Max signal strength?

12 Estimating Average Spot Rates Enter Exit 80 Speed Field Sensor Field Sensor 10 0 Bluetooth 0:00:00 2:24:00 4:48:00 7:12:00 9:36:00 12:00:00 14:24:00 16:48:00 19:12:00 21:36:00 0:00:00 Speed Speed Exit Speed Enter

13 Percentage of Values in 50sec Range ( PDF) Percentage of Values in 50sec Range ( PDF) 13 Estimating Spot Rate Distributions Off-Peak Travel Rate PDFs Approximate Peak Travel Rate PDFs 30% 25% 25% 20% 20% All 15% All 15% 13-Jan 13-Jan 10% 20-Jan 22-Jan 24-Jan 10% 20-Jan 22-Jan 24-Jan 5% 5% 0% Travel Rate (sec/mi) 0% Travel Rate (sec/mi)

14 Travel Time (min) Segment Time and Rate PDFs Travel Times Weekdays :00:00 2:24:00 4:48:00 7:12:00 9:36:00 12:00:00 14:24:00 16:48:00 19:12:00 21:36:00 0:00:00 Time of Day 11

15 Percentile (CDF) Percentile (CDF) Percentile (CDF) PDFs for TTs and TRs 120% 100% CDFs by Regime - Weekdays 39 80% TranFrPk 60% Mar15Pk Peak 40% TranToPk Off-Peak 120% 100% TR-CDFs by Regime - Weekdays 20% 0% Trip Time (min) 80% TranFrPk 60% Mar15Pk Peak 9 40% 20% TranToPk Off-Peak 120% CDFs by Regime - Weekdays 0% Trip Rate (min/mi) 100% 80% 10 60% 40% 20% TranFrPk Mar15Pk Peak TranToPk Off-Peak 11 0% Trip Time (min)

16 Trip Time (min) 16 Finding Travel Times in Trip Times 6 5 NY North Trip Times (6 to 3) / Week of Especially for arterials and long freeway segments :00 4:48 9:36 14:24 19:12 0:00 4:48 Time of Observation at 3 This is arterial data

17 17 Finding the Operational Impacts 6 5 North NY This is arterial data Signal control impacts

18 18 Average Times or Rates for Segments Upstream rate Downstream rate Segment Upstream rate Downstream rate Bluetooth data for vehicles Spot rates from field sensors Regression analysis Works for all => works for average Analyze each regime Segment rate Regime adjustment a b ab 2 ab a b (1 ) ab a b

19 Cumulative Probability 19 Fitting the Density Function Segment Transition-to-Peak CDF Four-Parameter Burr Distribution 120% 100% 80% k: a continuous shape parameter (k > 0); α: a second continuous shape parameter (α > 0); β: a continuous scale parameter (β > 0); and γ: a continuous location parameter (unbounded), less than the minimum observed value 60% 40% 20% 0% Time (min) Observed Estimated

20 Average Time and Rate PDFs for Routes Pick starting times Pick a network condition Generate the average rates and times Walk the time-space matrix

21 Incidence (Correlation) Matrices 21

22 Incidence (Correlation) Matrices In this case, simulated data Segment ab Segment bc τ bc (sec/mi) % 0% 0% 0% 0% 0% 0% 80 0% 0% 0% 0% 0% 0% 0% 100 0% 0% 8% 24% 5% 0% 0% 120 0% 0% 6% 21% 7% 1% 0% 140 0% 0% 1% 7% 2% 0% 0% 160 0% 0% 1% 3% 2% 0% 0% τ ab (sec/mi) 180 0% 0% 1% 2% 1% 0% 0% 200 0% 0% 0% 2% 1% 0% 0% 220 0% 0% 0% 1% 1% 0% 0% 240 0% 0% 0% 1% 1% 0% 0% 260 0% 0% 0% 0% 0% 0% 0% 280 0% 0% 0% 0% 0% 0% 0% 300 0% 0% 0% 0% 0% 0% 0% >300 0% 0% 0% 0% 0% 0% 0%

23 23 Vehicle PDFs for Routes TT-PDFs for segments Incidence matrix for pairwise regimes Monte Carlo simulation

24 24 Finding the Regimes Regimes Peak Off-peak Transition Weather Incidents Special Events Other Akin to signal timing plans

25 Cumulative Probability Travel Time (minutes) Travel Time (minutes) Finding the Regimes (2) Variations in Travel Time by Time of Day during Workdays Travel Time versus VMT for Various Situations Demand Incident Normal Event Weather :00 4:48 9:36 14:24 19:12 0:00 4:48 Time of Day 120% 100% Travel Time CDFs VMT/hr 80% 60% 40% 20% 0% Travel Time (min) Normal Incident Weather All

26 Regimes for a Single Load Condition 26

27 Differences among Load Conditions 27

28 Building the Textbook Fundamental outputs Travel and trip time probability distribution functions (TT-PDFs, TR-PDFs) Individual vehicles or averages Vocabulary Data streams: SPS (Single Point Sensor), AVI, AVL Travel rates: time per distance - minutes/mile - spot and segment Monuments: mid-link locations Segments: monument to monument links Routes: sequences of segments Regimes: different operating conditions different PDFs Time periods: spans of time, e.g., 5 minute time slice, AM peak User groups: travelers of a similar type Use cases: a specific question about travel time reliability TPC segments: contiguous fragmentation of the highway network

29 29 More Thoughts about PDFs AVI: location time stamps ( v, x, y, t E) AVL: bread crumbs / virtual AVI ( v, x, y, t, s, h E) SPS: flow and travel rates for time slices ( x, y, f, r, n, k E) Ways to think about the PDFs f ( t i, c, d, E) or g( t, ) where h( i, c, d, E) t t t t t v vi c E all random variables, or a a a a a v vi c E,

30 30 Considering Signal Timing Impacts t = t v + t c t v = t min 1000 simulated vehicles 4-signal arterial t min = 220 sec Random arrivals at first signal

31 Questions / Thank you