NATMEC June 5, 2006 Comparative Analysis Of Various Travel Time Estimation Algorithms To Ground...

Post on 31-Dec-2015

218 views 4 download

Transcript of NATMEC June 5, 2006 Comparative Analysis Of Various Travel Time Estimation Algorithms To Ground...

NATMEC June 5, 2006

Comparative Analysis Of Various Travel Time Estimation Algorithms To Ground Truth Data Using Archived Data

Christopher M. MonsereResearch Assistant ProfessorDept of Civil and Environmental EngineeringPortland State University

Sirisha Kothuri, Kristin Tufte, Robert L. Bertini,

School of Urban & Public AffairsPortland State University

Comparative Analysis Of Various Travel Time Estimation Algorithms To Ground Truth Data Using Archived Data

Christopher M. MonsereSirisha Kothuri, Kristin Tufte, Robert L.

Bertini, Aaron BreakstoneDepartment of Civil & Environmental Engineering

Intelligent Transportation Systems LaboratoryPortland State University

NATMECJune 5, 2006

Minneapolis, Minnesota

NATMEC June 5, 2006

Outline

• Introduction

• Study Area

• Data Sources

• Data Analysis

• Preliminary Conclusions

• Next Steps

NATMEC June 5, 2006

• Variety of technologies– Inductive loop detectors– Microwave radar– Automatic vehicle tag matching– Video detection– License plate matching– Cell phone matching

• Past research– General accuracy in free-flow conditions– Recurring congestion & incidents more challenging

• FHWA policy on DMS

Real-time Travel Time Estimates

NATMEC June 5, 2006

Portland ATMS• Freeway surveillance

– 485 inductive loop detectors• Dual loop• Mainline lanes• Upstream of on-ramps

– 135 ramp meters– 98 CCTV

• Traveler information• www.TripCheck.com• Real-time speed map• Static CCTV images

– 18 Dynamic Message Signs (DMS)

• 3 display travel times

NATMEC June 5, 2006

Objectives

• Build on recent evaluation of Oregon Department of Transportation (ODOT)’s travel time estimating and reporting capabilities

• Test other algorithms on select links using historical data

NATMEC June 5, 2006

Study Area

• 9 directional freeway links– I-5 (3)– I-205 (3) – I-84 (1)– OR-217 (2)

NATMEC June 5, 2006

Probe Vehicle• Data collected

– Standard probe vehicle instructions (FHWA)

– 5-10 runs – Transitional periods

targeted– Groups with 5-7 minute

headways

• Measures– 87 probe vehicle runs– 15 hours – 516 miles – 7 days (Wed – Fri)

NATMEC June 5, 2006

Probe Vehicle• Hardware

– Palm handheld computers

– Magellan GPS devices

• Software– ITS-GPS

Available at www.its.pdx.edu

• Individual runs and groups of probe vehicles

• Variety of traffic conditions– 45 percent congested

– 2 notable incidents

NATMEC June 5, 2006

Probe Vehicle Data• Individual runs

downloaded – “run” = several links

+ extraneous data

• Unique ID for each GPS record

• Runs plotted on freeway network

– Links color-coded

• Pertinent data segments extracted

last point on Link 9

last point on Link 2

first point on Link 3

NATMEC June 5, 2006

PORTAL• National ITS

Architecture ADUS• Funded by NSF• Direct fiber-optic

connection between ODOT and PSU

• 20-second data– Occupancy– Volume– Speed

• Customized travel time area – Conforms to TMOC

(Portland Regional Transportation Archive Listing)

www.portal.its.pdx.edu

NATMEC June 5, 2006

Travel Time – Midpoint Algorithm

Influence

Area iTravel Time i

(at t = 0)

Travel Time 1

(at t = 0)

Influence

Area 1

Travel Time 3

(at t = 0)

Influence

Area 3

Travel Time 2

(at t = 0)

Influence

Area 2

Link Travel Time

(TT1 + TT2 + TT3 + TTi)

NATMEC June 5, 2006

Travel Time - Coifman Algorithm

Time

Dis

tanc

e

downstream detector

upstream detector

NATMEC June 5, 2006

Methodology – Implementation • Coifman u/s • Coifman d/s • Coifman -

midpt • Coifman -

distwt • Midpoint• Midpoint

Average

NATMEC June 5, 2006

Analysis – Link 3 Trajectories (Uncongested)

292.00

293.00

294.00

295.00

296.00

297.00

298.00

299.00

300.00

17:02:00 17:04:00 17:06:00 17:08:00 17:10:00 17:12:00 17:14:00

Time

Dis

tan

ce

Probe Coifman_u/s Midpt Det_Loc_1

Det_Loc_2 Det_Loc_3 Det_Loc_4 Det_Loc_5

Det_Loc_6 Det_Loc_7 Det_Loc_8

NATMEC June 5, 2006

Analysis – Link 6 Trajectories (Large Spacing)

293.00

294.00

295.00

296.00

297.00

298.00

299.00

300.00

8:10:00 8:15:00 8:20:00 8:25:00

Time

Dis

tan

ce

Probe Coifman_d/s Midpt Det_Loc_1

Det_Loc_2 Det_Loc_3 Det_Loc_4

NATMEC June 5, 2006

Analysis – Link 8 Trajectories (Incident)

0.00

1.00

2.00

3.00

4.00

5.00

6.00

7.00

8:25:00 8:30:00 8:35:00 8:40:00 8:45:00 8:50:00 8:55:00 9:00:00

Time

Dis

tan

ce

Probe Coifman_u/s Midpt Det_Loc_1

Det_Loc_2 Det_Loc_3 Det_Loc_4 Det_Loc_5

Det_Loc_6 Det_Loc_7 Det_Loc_8

NATMEC June 5, 2006

Analysis – Probe Vehicle, Coifman, Midpoint

Link 3

Travel Time Observations (95% CI)

7.00

8.00

9.00

10.00

11.00

12.00

Probe Coifmand/s

Coifmanu/s

Coifman-midpt

Coifman-distw t

Midpt Midpt_Avg

Tra

vel T

ime

(min

s)

Link 4

Travel Time Observations (95% CI)

5.00

10.00

15.00

20.00

25.00

Probe Coifmand/s

Coifmanu/s

Coifman-midpt

Coifman-distw t

Midpt Midpt_Avg

Tra

vel T

ime

(min

s)

NATMEC June 5, 2006

Results - Incident Travel Times

Link 8 RMSE

0

5

10

15

1Coifman d/s Coifman u/s Coifman-midpt Coifman-distw t Midpt Midpt_Avg

Lowest Error

NATMEC June 5, 2006

Analysis - Coifman & Midpoint Errors

0.00

1.00

2.00

3.00

4.00

5.00

6.00

7.00

3 5 6 8 9 10 11 12 13

Link No

RM

SE

coifman u/s midpt

NATMEC June 5, 2006

Conclusions

• Coifman estimated travel times are more accurate than midpoint travel times

– Both algorithms estimate travel times fairly accurately during free flow conditions

• The accuracy is related to– Location and density of detectors– Location, formation and dissipation of queue

• Incidents & large spacing between detectors result in larger error in midpoint travel time estimates

NATMEC June 5, 2006

Next Steps

• Further testing of Coifman and Midpoint under varying traffic conditions

• Expand Coifman algorithm from current historical analysis to real time estimates

• Comparison to ODOT range of travel times • Data quality??

NATMEC June 5, 2006

Acknowledgements• Castle Rock Consultants

– Dean Deeter

• ODOT– Galen McGill– Stacy Shetler

• PORTAL Team• Volunteer Drivers