Download presentation
Presentation is loading. Please wait.
Published byHolly Wheeler Modified over 9 years ago
1
Assessment and Refinement of Real-Time Travel Time Algorithms for Use in Practice Nov 8, 2006
2
Outline Ramp Meter Data Fidelity Assessment Inrix Data Update Data Collection Plan Travel Time Best Practices Results Schedule update
3
Ramp Meter Data Fidelity Assessment Impacts of Various Factors on Travel Time Estimation Accuracy Algorithms Detector Spacing Data Quality
4
Algorithm Comparison: Uncongested Runs I-5 N (217-405) I-5 S (Bridge-84) I-5 S (405-217) 217 SI-205 S (84-O.City) I-84 E (5-205)
5
Algorithm Comparison: Congested Runs I-5 SB (405-217) 217 N217 SI-205 N (5-O.City) I-5 NB (84-Bridge) I-5 S (Bridge-84) I-84 E (5-205) Large detector spacing Some probe runs encountered an incident Significant recurring congestion
6
Algorithms: Trajectory Comparison
7
Conclusions - Algorithms FHWA says 90% accuracy is ideal, accuracy must be no less than 80% (Agrees with what we discussed last time) No algorithm is consistently better and consistently < 10% Many runs have error > 10% Appears to be associated with large detector spacing and incidents Need more data to verify impacts of algorithms, spacing, etc. Moderate impact from algorithm, but probably not enough to overcome infrastructure issues (more when we examine other states practices)
8
Detector Spacing Impacts-Analytical More detector stations => more data samples Lower error due to more samples If one detector has issues, others will mitigate that problem Shockwave Propagation When an incident/bottleneck occurs far from a detector, it takes time for the congestion to reach the detector Shockwave propagation 12-16 mph, 15 mph = 4 minutes/mile 2 miles 1.5 miles = approx. 6 minutes Detector 1 Detector 2 Bottleneck
9
Detector Spacing Impacts: Congested Conditions
10
I-205 NB – Stafford – MP 3.55 – Lane 2
11
I-5 NB – Delta Park – MP 306.51 – Lane 2
12
Conclusions – Detector Spacing and Data Quality Detector Spacing Expect and think we see association with detector spacing Need more data to verify Are also creating an analytical model for detector spacing impacts Data quality Suspect there is an impact Need more data to verify Would like to clarify speed calculation procedure
13
Inrix Data Provide flow and travel time data XML data stream Data Sources Current data is a processed version of the ODOT Region 1 loop detector data As of mid-November, probe data will be included (transponder detectors from instrumented fleets (taxis, etc.))
14
Data Validation Inrix has validated accuracy of their data for three east-coast cities The networks in these cities use probe data only Validation not valid for Portland (different city, probe + detector data) Potential good source of data, but do not believe we can use it as ground truth without more validation We are getting sample data (NDA in process) Meeting with Inrix Technical Staff?
15
Ground Truth Data Collection – Phase 1 Initial Study to confirm methodology and process Issues: Collection Process Corridor selection - focus on two corridors for this phase Number of Runs Timing of Runs
16
Quality Counts Recommended by ODOT personnel $45/hour + mileage (currently ~$0.49/mile) We provide list of highways, hours, and a list of locations on the highways that we want timed They use a stopwatch and record the time when they pass each specified location
17
Corridor Selection Corridor Selection Criteria Moderate-severe recurrent congestion Variable loop detector spacing (some low some high) to allow evaluation of spacing effects Some situations with high data quality Construction Schedule – avoid times/areas when there is construction Propose: OR 217 (‘good’ conditions) I-205 or I-5 (more variable detector spacing) Credit to Sue Ahn for her ideas for corridor selection in SWARM
18
217 N, Weekdays - April, 2006 traffic flow
19
217 S, Weekdays - April, 2006 traffic flow
20
217 Notes Congestion: moderate congestion both NB and SB Congestion NB and SB in both AM and PM Peaks PM congestion generally worse than AM SB congestion generally worse than NB Detector Spacing: good NB: 9 stations SB: 11 stations Length: ~7 miles Data Quality 217 N - ~1% disqualified data 217 S - ~2% disqualfied data
21
Timing & Cost Specifics – 217 PM Peak 217 S Peak: 3:00-6:00 Min/Max/Avg TT: 14.3/25.0/20.5 min 217 N: Peak: 4PM – 6PM Min/Max/Avg TT:10.2/14.2/12 min Average Round trip ~32 minutes Need ~50 runs for 5% error at 95% confidence Start with 20 runs/corridor 2 runs/hr, 10 hrs = 20 runs = $450 ($45/hr) Gas cost: 20 runs * 7 miles * 0.5/mile = $70 Data from weekdays – April, 2006
22
Timing – 217 AM Peak AM Peak 217 N Peak: 7:30-8:15 Min/Max/Avg TT: 8.2/10.6/9.3 min 217 S: Peak: 7:00-9:00 Min/Max/Avg TT:12.5/20.7/16.1 min Avg round trip time ~25 min Data from weekdays – April, 2006 Similar costs $500 for 20 runs
23
Detector Locations - 217
24
I-205 N, Weekdays – October, 2006 traffic flow Detector spacing poor before mp 8
25
I-205 S, Weekdays – October, 2006 traffic flow Detector spacing poor after mp 8
26
Detector Locations I-205 SB Stark/Washington mp 20.34 SB Clackamas Hwy mp 12.67 SB Johnson Creek mp 16.24
27
I-205 Notes Detector spacing is poor for mileposts 0-8 Do not collect data on that portion of 205 – means can not capture the congestion that occurs there Consider collecting mp 13 – mp 20 Congestion: Some congestion on northern end of I-205 NB and SB, AM and PM peaks NB AM congestion appears worst Detector Spacing: See Map Data Quality I-205 NB - ~1% disqualified data I-205 SB - ~2% disqualified data
28
I-5 N Weekdays – October, 2006 traffic flow
29
I-5 S, Weekdays – October, 2006 traffic flow
30
Detector Locations I-5 S of Downtown NB, Nyberg, mp 289.4 NB, Macadam, mp 299.7
31
Detector Locations I-5 N of Downtown NB, Macadam, mp 299.7 SB, Swift/Marine, mp 307.35
32
I-5 Notes Congestion: N of Portland: SB congestion in AM and PM peaks, NB congestion PM peak S of Portland: Minimal SB congestion, NB congestion through curves in AM peak NB PM congestion (going over the bridge) appears worst More severe congestion than 205 Detector Spacing: Variable - See Map Data Quality I-5 NB - ~2% disqualified data I-5 SB - ~4% disqualified data
33
Other States – Best Practices…
34
Milwaukee, WI Detector Spacing 0.25 miles in urban areas 2 miles in rural areas Data from detectors transmitted to TOC Center Freeway Traffic Management System (FTMS) Server Travel Time = Known Distance/Average Speed Website updated every 3 minutes DMS signs updated every 1 minute No probe vehicle data; all detector derived travel times
35
San Antonio, TX Travel Times calculated from/to major interchanges Detectors Loop Detectors Video Detectors Point travel speeds used to calculate travel times from detector to detector Segment travel speed is the lower of u/s and d/s speed Point to point travel times are summation of segment travel times Travel times posted on TransGuide website use the same algorithm
36
Chicago, IL DMS Travel Times From three sources (IPASS, RTMS, Loops) GCM Webpage Only IPASS travel times IPASS Data Travel times from toll plaza to toll plaza Based on toll transponder data collected by ETC system > 1.5 million users on tollways Significant number of probe vehicles provide time stamp and location Travel times calculated using location and time stamp information High quality of data RTMS Data IDOT Loop detector data
37
Houston, TX Vehicle Probes with transponder tags Readers collect data as vehicles pass 2-3 miles apart Time Location of probe Software Average Speeds Average Travel Times Transtar website DMS Updated every 10 minutes
38
Nashville, TN RTMS detectors 0.25 mile spacing Speeds Ensure data quality by regular calibration CCTV cameras Travel Time verification Data Collection & Processing Average speed from RTMS every 2 minutes Travel time calculation average speed and distances between sensors Travel Times automatically posted to the DMS by TMC software Travel Times are only reported for segments < 5 miles
39
Atlanta, GA VDS Cameras Monitoring and Video Detection Cameras Fixed black and white cameras Placed along all major freeways Provide volumes and speeds Travel Times between 6 a.m. – 9 p.m. Average speeds from Video Detection Cameras Software Automatic message generation for DMS
40
San Francisco, CA Existing Caltrans System Dual Loop Detectors Speeds New MTC System Antennas to read FasTrak Toll Tags Average Travel Times and speeds of all vehicles 511 System Combination of data from both sources to calculate travel times
41
Other Cities www.smartroute.com Real Time Traveler Information Boston Miami St. Louis North Carolina New Jersey
42
Summary Two main approaches for generating travel times In house Loop Detectors High Density (0.25 mile spacing) Video Detectors RTMS Toll Tags Private providers Smartroute Systems Inrix
43
Schedule Update…
Similar presentations
© 2024 SlidePlayer.com. Inc.
All rights reserved.