Assessment and Refinement of Real-Time Travel Time Algorithms for Use in Practice September 6, 2007
Agenda Accomplisments TRB paper submitted Effects of Additional Detection Correlations Data Quality
Additional Detection Simulation of additional of detectors on I-5 and OR 217 Additional detectors simulated using probe vehicle speeds Terms: Instantaneous speeds – speeds measured by detectors at time when vehicle enters segment Real-time speeds – speeds measured by detectors at time probe passed the detector
I-5 NB SoD - Additional Detector Locations MP Macadam (299.7) Terwilliger (297.33) Capital (295.18) Spring Garden (296.26) MP – 295.5
I-5 NB SoD Appears there is some improvement when adding detector in the curves Note improvement when using real-time speeds instead of instantaneous speeds Suspect we perhaps did not catch transition periods on this segment. Or else, surprisingly, detection is ok. Midpoint Error (%) (Instantaneous speeds) Real Time Error (%) (Real-time speeds) Real Time Error – Additional detector between Terwilliger and Macadam (%) Real Time Error – Additional detector between Capital and Spring Garden (%) MAPESDPESEMAPESDPESEMAPESDPESEMAPESDPESE
I-5 SB SoD Additional Detector Locations MP 292 MP Haines (293.1) Upper Boones (291.25) MP 292
Errors due to Large Detector Spacing I-5 SB (South of Downtown) Date: 4/16/2007 Start Time: 16:19 Error: -32.2% 2.11 mi.
I-5 SB SoD Appears there is some improvement when adding detector in the curves Note improvement when using real-time speeds instead of instantaneous speeds Suspect we perhaps did not catch transition periods on this segment. Or else, surprisingly, detection is ok. Midpoint Error (%) (Instantaneous speeds)1 Real Time Error (%) (Real-time speeds) Real Time Error (%) – Additional detector south of I-5/217 junction Real Time Error (%) – Additional detector north of I-5/217 junction MAPESDPESEMAPESDPESEMAPESDPESEMAPESDPESE
I-5 SB SoD Additional detection appears to improve more than influence area adjustments Midpoint Error (%) (Instantaneous speeds) Real Time Error (%) (Real-time speeds) Real Time Error (%) – Additional detector south of I-5/217 junction Midpoint Error (%) – influence area adjustments MAPESDPESEMAPESDPESEMAPESDPESEMAPESDPESE
I-5 NB NoD - Additional Detector Locations Morrison (301.09) MP – Macadam (299.7) MP – 300.8
I-5 NB NoD Significant improvements with both locations Errors are still borderline high Midpoint Error (%) (Instantaneous speeds) Real Time Error (%) (Real-time speeds) Real Time Error – Additional detector between Morrison and Macadam (S) (%) Real Time Error – Additional detector between Morrison and Macadam (N) (%) MAPESDPESEMAPESDPESEMAPESDPESEMAPESDPESE
I-5 SB NoD - Additional Detector Locations MP – Swift/Marine (307.35) Columbia (305.97) Hood (299.25) Wheeler (302.17) MP –
I-5 SB NoD Detector between Marine/Columbia appears useful Detector at I-84 junction less useful Need to investigate additional detector locations? Marquam Bridge? Error level is still borderline high Midpoint Error (%) (Instantaneous speeds) Real Time Error (%) (Real-time speeds) Real Time Error – Additional detector between Marine and Columbia (%) Real Time Error – Additional detector near I-84 junction (%) MAPESDPESEMAPESDPESEMAPESDPESEMAPESDPESE
Effects of Additional Detection Between Macadam/ Terwilliger South of I-5/217 junction
Locations of Additional Detection Segment Description Length of Segment (mi.) Average Detector Spacing (mi.) Additional Detector Location 1 (Milepost) Additional Detector Location 2 (Milepost) I-5 NB (SoD) Terwilliger/Macadam (298.0) Capital/Spring Garden (295.5) I-5 SB (SoD) South of I-5/217 junction (292.0) North of I-5/217 junction (292.48) I-5 NB (NoD) Morrison/Macadam (300.5)Morrison/Macadam (300.8) I-5 SB (NoD) Marine/Columbia (306.5)I-84 Junction (301.25) OR-217 NB Just N of junction with I-5 (6.9) Greenburg/Scholls (4.3) OR-217 SB Walker/B-H Hwy (1.00)Allen/Denny (3.00)
Effects on Errors SegmentMidpoint Error (%)Real Time Error (%)Real Time Error with Detector at Location 1 (%) Real Time Error with Detector at Location 2 (%) MAPESDPESEMAPESDPESEMAPESDPESEMAPESDPESE OR 217 NB OR 217 SB I-5 NB SoD I-5 SB SoD I-5 NB NoD I-5 SB NoD
Correlation Investigations All Runs I-5 NB SoD I-5 SB SoD I-5 NB NoD I-5 SB NoD 217 NB217 SB Average loop speed Average probe speed Stddev probe speed Estimated travel time Minimum loop speed
Minimum Speed vs. Error Correlation Coefficient
I-5 SB SoD Avg Probe Speed vs. Error
I-5 NB NoD Problem location was always between Morrison and Macadam
Going Forward Goals Identified Last Meeting Estimate benefit of additional detection Simulated additional detectors for I-5 and 217 Try to answer question “under what conditions does algorithm accuracy degrade” Some correlation found when broken down by segment To do: correlate with rate of change of speed at detector station Algorithm refinement based on those two questions Where do we go from here?
Overall Result Summary Segment Description Avg Detector Spacing (mi.) MAPESDPEMPESENum Runs Percent Estimates with Error < 20% Percent Estimates with Error < 30% I-5 NB SoD I-5 SB SoD I-5 NB NoD I-5 SB NoD OR 217 NB OR 217 SB I-84 EB I-84 WB I-205 NB I-205 SB I-405 NB I-405 SB US 26 EB US 26 WB
Data Quality Trying to understand and improve data quality Visit from UW-researchers in September to investigate potential calibration issues Hoping to create data quality reports useful to ODOT Following slides demonstrate types of detector issues currently observed
I-5 NB Denver Ave – Speed & Occupancy I-5 NB at Denver Ave Milepost: Number of Lanes: 3 Plots show all lanes; These speeds appear somewhat suspect; in previous observations, speeds appeared less suspect. Speed vs. TimeOccupancy vs. Time
I-5 NB Terwilliger Blvd – Speed & Occupancy I-5 NB at Terwilliger Blvd Milepost: Number of Lanes: 3 Plots show all lanes. Speed vs. TimeOccupancy vs. Time
I-205 NB at Johnson Creek Low nighttime speeds are a serious problem at this detector. Daytime speeds look reasonable. Lane 3 shows similar issues, lane 1 shows fewer issues with low nighttime speeds. Plots for one day only; however, issues are consistent across multiple days. I-205 NB at Johnson Creek Milepost: 16.2 Number of Lanes: 3 Note: station does show congestion on Aug 31, second speeds for Lane 2; midnight to 5am, Aug 15, second speeds for Lane 2; entire day of Aug 15, 2007.
Lane 3, Occupancy vs. TimeLane 3, Speed vs. Time I-84 WB – Sandy – Lane 3 Lane 3 exhibits high number of occupancy readings > 95%. I-84 WB at Sandy Blvd Milepost: 2.4 Number of Lanes: 3
I-84 WB – Sandy – Lanes 1 and 2 Lane 1, Speed vs. Time Lane 1 and Lane 2 exhibit a 20mph difference in free flow speeds. Lane 1 and Lane 2 also exhibit - low maximum occupancies. Lane 2, Speed vs. Time I-84 WB at Sandy Blvd Milepost: 2.4 Number of Lanes: 3
I-5 NB – Spring Garden St Lane 2 – Flow vs. Occupancy Lane 3 Flow vs. Occupancy Lane 3 exhibits an unusually flow/occupancy ratio. I-5 NB at Spring Garden Milepost : Number of Lanes: 3 Plots from August 3, 2007
I-205 SB – ORE 224/82 nd Ave Spd > 100 Lane 1 exhibits a high number of readings with speed > 100. Lane 1 speeds differ significantly from lane 2 speeds. I-205 SB at ORE 224/82 nd Ave Milepost: Number of Lanes: 3 Lane 1, Speed vs. TimeLane 2, Speed vs. Time
I-84 EB – 16th Lane 3 exhibits high number of readings with occupancy > 95%; lane 1 exhibits high number of readings with volume > 17, Lanes 1 and 3 exhibit high number of readings with speed = 0 and volume > 0. This detector generally appears problematic – perhaps this detector is degrading? I-84 EB at 16th Milepost: 1.3 Number of Lanes: 3 Lane 1, Volume vs. TimeLane 2, Volume vs. Time
I-5 NB - Multnomah Lanes 1 and 3 exhibit a high number of readings with occupancy > 95%. Lanes 1 and 3 behave very differently from lane 2. I-5 NB at Multnomah Milepost: 1.3 Number of Lanes: 3 Lane 2, Speed vs. TimeLane 3, Speed vs. Time
I-5 NB - Multnomah Lanes 1 and 3 exhibit a high number of readings with occupancy > 95%. Lanes 1 and 3 behave very differently from lane 2. I-5 NB at Multnomah Milepost: 1.3 Number of Lanes: 3 Lane 2, Occupancy vs. TimeLane 3, Occupancy vs. Time
I-5 NB - Multnomah 20-second occupancies, July 9, 2007 I-5 NB at Multnomah Milepost: 1.3 Number of Lanes: 3 Lane 3, Occupancy vs. Time, 7-8AMLane 3, Occupancy vs. Time
OR 217 SB – Walker Rd Lane 2 speeds from July 9, 2007 – Aug 3, 2007 Lane 2 exhibits an unusually high flow occupancy ratio. Lane 1 and Lane 3 also exhibit similar behavior. OR 217 SB at Walker Rd Milepost: 0.76 Number of Lanes: 3
I-205 NB at Glisan Lane 1 20-sec speeds Aug 15, 2007 Lanes 1 and 2 exhibit an unusually high flow occupancy ratio. I-205 NB at Glisan Milepost: Number of Lanes: 3
Selected Stations HighwayLocationMilepostLane Number* Reason SelectedStationid (Detectorids) I-5 NBDenver Ave306.51AllObserved anomalies1022 (1204, 1205, 1206) I-5 NBTerwilliger Blvd297.33AllObserved anomalies1013 (1129, 1130, 1131) I-84 WBSandy Blvd2.43 (also 1 & 2) High number of readings with occupancy > 95% 1061 (1474) I-5 NBSpring Garden Unusually high flow/occupancy ratio1011 (1115) I-205 SBORE 224/82 nd Ave High number of readings with speed > (1938) I-84 EB16 th EB1.331,3Several issues – see detailed slide1056 (1432, 1434) I-5 NBMultnomah296.61,3High number of readings with occupancy > 95% 1012 (1121, 1123) OR 217 SBWalker Rd0.762Unusually high flow/occupancy ratio1080 (1552) I-205 NBGlisan21.121,2Unusually high flow/occupancy ratio1142 (1949, 1950) I-205 NBJohnson Creek16.2Observed low nighttime speeds.1046 * I-5 NBORE 217/Kruse Way ,3Observed low nighttime speeds.1007 (1074, 1075) I-5 NBUpper Boones Overnight low speeds are less of a problem1006 (1058) I-5 NBCapitol Hwy Observed low nighttime speeds1010 (1106) *Lane 1 is the left-most lane
I-5 NB SoD MP MP – Macadam (299.7) Terwilliger (297.33) Capital (295.18)
I-5 SB SoD MP 292 MP Haines (293.1) Upper Boones (291.25)
OR 217 NB MP 6.9 MP 4.3 Scholls (3.85) Greenburg (4.65)
OR 217 SB B-H Hwy (1.92) MP 1.00 MP 3.00 Walker (0.76) Allen (2.55) Denny (3.12)
Overview - Recap
Potential Solutions Errors due to changes in conditions Short Term Trending Historical Data Errors due to non-functioning detectors Low cost portable detectors Historical Data Errors due to large detector spacing Additional Detection
Sources of Estimation Errors Detailed analysis of high error runs on I-5 ( Sections both north and south of downtown) OR – 217 (NB & SB) Three main sources of errors Errors due to changes in conditions Errors due to non-functioning detectors Errors due to large detector spacing
Errors due to Changes in Conditions I-5 NB North of Downtown Date: 4/24/2007 Start Time: 17:47 Error: 27.13%
Errors due to Non-Functioning Detectors OR-217 NB Date: 4/16/2007 Start Time: 16:16 Error: % Missing Detectors
Errors due to Large Detector Spacing I-5 SB (South of Downtown) Date: 4/16/2007 Start Time: 16:19 Error: -32.2% 2.11 mi.
Probe Travel Time vs. Error
Midpoint Travel Time vs. Error
Average Speed vs. Error
Speed Differential vs. Error (I-5 High Error Runs Only)
Speed Differential vs. Error (I-5 High Speed Differential Runs Only)