200th Seminar! Travel Time Estimation for Traffic Management and Traveler Information

1Travel Time Estimation for Traffic Management and Traveler InformationTravel Time Estimation for Traffic Management and Traveler Information

Robert L. BertiniRobert L. BertiniDepartment of Civil and Environmental EngineeringDepartment of Civil and Environmental EngineeringNohad A. Toulan School of Urban Studies and PlanningNohad A. Toulan School of Urban Studies and PlanningPortland State UniversityPortland State UniversityOregon Transportation Research and Education ConsortiumOregon Transportation Research and Education ConsortiumTransportation Seminar No. 200 Transportation Seminar No. 200 •• April 4, 2008April 4, 2008

Travel Time Estimation for Travel Time Estimation for Traffic Management and Traffic Management and Traveler InformationTraveler Information


ObjectivesObjectives

Travel time visualizationsTravel time fundamentalsPrevious researchFramework for sensor spacingAnalytical tool for sensor spacingFuture researchSeminar perspectives

Travel time visualizationsTravel time fundamentalsPrevious researchFramework for sensor spacingAnalytical tool for sensor spacingFuture researchSeminar perspectives


What Can I Do With Travel Time?What Can I Do With Travel Time?

Travel time is fundamentalCan be used to generate other thingsTravel time is multimodal

Travel time is fundamentalCan be used to generate other thingsTravel time is multimodal


What Can I Do With Travel Time?What Can I Do With Travel Time?

Measure itGuess itReport itPredict itForecast it

Measure itGuess itReport itPredict itForecast it


If I Get It Wrong?If I Get It Wrong?

Annoy travelersDestroy confidence in systemIncrease congestionWorsen safetyDamage air qualityIncrease fuel consumption

Annoy travelersDestroy confidence in systemIncrease congestionWorsen safetyDamage air qualityIncrease fuel consumption


How Can I Affect Travel Time?How Can I Affect Travel Time?

Increase/decrease mean valueIncrease/decrease variabilityAffect comparison between modes

Increase/decrease mean valueIncrease/decrease variabilityAffect comparison between modes


ContextContextContext

“Prediction is very difficult, especially about the future.”

—Niels Bohr


Portland Travel TimesPortland Travel TimesPortland Travel Times


London UndergroundLondon UndergroundLondon Underground


London UndergroundLondon UndergroundLondon Underground


2004 Tsunami2004 Tsunami2004 Tsunami


Runoff Travel TimeRunoff Travel TimeRunoff Travel Time


Fire Station Response TimeFire Station Response TimeFire Station Response Time


New York Commuter RailNew York Commuter RailNew York Commuter Rail


2004 Tsunami2004 Tsunami2004 Tsunami

Train journey times from Cambridge


Travel time difference train vs. car from Cambridge


FundamentalsFundamentalsFundamentals

Time

Dis

tan

ce

Slope = Speed

Travel Time


FundamentalsFundamentalsFundamentals

Time

Dis

tan

ce

Actual Travel Time

Free Flow Travel Time Actual – Free Flow= Delay


Fundamentals – One DetectorFundamentals Fundamentals –– One DetectorOne Detector

Time

Dis

tan

ce

Travel Time


Midpoint MethodMidpoint MethodMidpoint Method

Influence

Area 4Travel Time 4

(at t = 0)

Travel Time 1Influence

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 + TT4)


Fundamentals – Two DetectorsFundamentals Fundamentals –– Two DetectorsTwo Detectors

Time

Dis

tan

ce

Travel Time


Measurement ParametersMeasurement Parameters

Fixed LocationsStop Watch Method

Detectors (any kind)RF Toll TagsRF “Sign Posts”Video Image (license plate)Volume Based

Fixed TimesGPS + Wireless CommunicationCellular Phone

Fixed LocationsStop Watch Method

Detectors (any kind)RF Toll TagsRF “Sign Posts”Video Image (license plate)Volume Based

Fixed TimesGPS + Wireless CommunicationCellular Phone


Measurement: Fixed LocationsMeasurement: Fixed LocationsMeasurement: Fixed Locations

Time

Dis

tan

ce

x1

x2

Location Time

x1

x2


Measurement: Fixed LocationsMeasurement: Fixed LocationsMeasurement: Fixed LocationsD

ista

nce

x1

x2

Location Time

x1 t1

x2

Time



ista

nce

x1

x2

Location Time

x1 t1

x2 t2

Time

t1 t2



ista

nce

x1

x2

Location Time

x1 t1

x2 t2

Time

t1 t2


Measurement: Fixed TimesMeasurement: Fixed TimesMeasurement: Fixed Times

Dis

tan

ce

x1

Time

t1 t2 t3 t4 t5

Time Location

t1 x1

t2

t3

t4

t5



Dis

tan

ce

x1

Time

t1 t2 t3 t4 t5

x2

Time Location

t1 x1

t2 x2

t3

t4

t5



Dis

tan

ce

x1

Time

t1 t2 t3 t4 t5

x3

x2

Time Location

t1 x1

t2 x2

t3 x3

t4

t5



Dis

tan

ce

x1

Time

t1 t2 t3 t4 t5

x4

x3

x2

Time Location

t1 x1

t2 x2

t3 x3

t4 x4

t5



Dis

tan

ce

x1

x5

Time

t1 t2 t3 t4 t5

x4

x3

x2

Time Location

t1 x1

t2 x2

t3 x3

t4 x4

t5 x5


Use Counts (TTI Method)Use Counts (TTI Method)Use Counts (TTI Method)


Use Counts (TTI Method)Use Counts (TTI Method)Use Counts (TTI Method)


Flow (vph)

Sp

eed

(m

ph

)

Similar Idea Using Granular DataSimilar Idea Using Granular DataSimilar Idea Using Granular Data


WSDOT Travel TimesWSDOT Travel TimesWSDOT Travel Times


Data Collection Was DifficultData Collection Was DifficultData Collection Was Difficult

Data CollectionGreenshields, et al., 1947

SpeedometerGreenshields, et al., 1957


Previous Research: LAFSPPrevious Research: LAFSPPrevious Research: LAFSP


Previous: Highway 18Previous: Highway 18Previous: Highway 18


Previous Research: FrontierPrevious Research: FrontierPrevious Research: Frontier


Previous Research: ODOT VMS IPrevious Research: ODOT VMS IPrevious Research: ODOT VMS I

40

15 directional freeway 15 directional freeway linkslinks87 probe runs87 probe runs516 miles/12 drivers516 miles/12 drivers15 hours of data collected15 hours of data collected



0

2

4

6

8

10

12

14

16

18

20

0 2 4 6 8 10 12 14 16 18 20

Probe Travel Time (min)

Estim

ated

Tra

vel T

ime

(min

)Coifman (u/s)Coifman (d/s)MidpointCoifman - MidpointCoifman - DistwtMidpoint - Average





00

2

4

6

8

10

12

14

16

0 2 4 6 8 10 12 14

Bus Travel Time (min)

Estim

ated

Tra

vel T

ime

(min

)Coifman (u/s)Coifman (d/s)MidpointCoifman - MidpointCoifman - Distwt


ODOT Phase II: TufteODOT Phase II: Tufte

Data quality

Congestion

300 ground truth runs on I-5 and OR 217

Other algorithms

Data quality

Congestion

300 ground truth runs on I-5 and OR 217

Other algorithms


ODOT Phase IIODOT Phase II

Average error ~5-7%Need for more detection, but where?Average error ~5-7%Need for more detection, but where?

0

5

1015

20

25

30

3540

45

50

<-30 -30-20 -20-10 -10-0 0-10 10-20 20-30 >30% Error

% o

f Run

s

# of Runs - 67

II--5 N5 N

0

10

20

30

40

50

<-30 -30-20 -20-10 -10-0 0-10 10-20 20-30 >30% Error

% o

f Run

s

# of Runs - 60

II--5 S5 S


MotivationMotivation

Ongoing efforts to improve Ongoing efforts to improve freeway travel time estimatesfreeway travel time estimatesDisplay travel time ranges for key Display travel time ranges for key corridorscorridorsDesire to provide additional Desire to provide additional detectiondetectionNeed for Need for ““optimaloptimal”” decisiondecision--making aidmaking aid


Portland ATMSPortland ATMS

Freeway Surveillance 502 inductive loop detectors~175 stations

Dual loop (act as single loop)1.2 mile average spacingUpstream of on-ramps

135 ramp meters98 CCTV

ATISwww.TripCheck.com

Real-time speed mapStatic CCTV images

18 dynamic message signs (DMS) 3 display travel times

Freeway Surveillance Freeway Surveillance 502 inductive loop detectors502 inductive loop detectors~175 stations~175 stations

Dual loop (act as single loop)Dual loop (act as single loop)1.2 mile average spacing1.2 mile average spacingUpstream of onUpstream of on--rampsramps

135 ramp meters135 ramp meters98 CCTV 98 CCTV

ATISATISwww.TripCheck.comwww.TripCheck.com

RealReal--time speed maptime speed mapStatic CCTV images Static CCTV images

18 dynamic message signs (DMS) 18 dynamic message signs (DMS) 3 display travel times 3 display travel times


Freeway Detector LocationsFreeway Detector LocationsFreeway Detector Locations


Portland Speed/Travel Time InfoPortland Speed/Travel Time InfoPortland Speed/Travel Time Info


Bus/Arterial SpeedsBus/Arterial SpeedsBus/Arterial Speeds


Hypothetical x-t PlaneHypothetical Hypothetical xx--tt PlanePlane

x1MeasuredSpeed

i

Actual Travel TimeFree Flow Travel Time

FreeFlowSpeed

Extrapolated Travel Time

l=

Segm

ent L

ength

x

tExtrapolatedSpeed

Over-prediction

t1 = Time Interval

Delay


Freeway Corridor ExampleFreeway Corridor Example

305.67 Columbia

305

300

295

290

285

4:00

8:00

12:0

0

16:0

0

20:0

0

285.26 Wilsonville

291.00 Carman

298.48 Iowa

302.22 Rose Qtr.

80

60

40

20

0

vfvAC

VMSMP

Time 02/08/07 Northbound I-5

DOWNTOWN

1 mile

A

B

DA C

AC CD

vCD DA


Traffic Flow Relation and DynamicsTraffic Flow Relation and Dynamics

q

k

qC

vf vc

vACqA

A

BC

D

A

B

C

D

A

x

t

vCD

bn

tdeact

vCDvAC

vfDAAC CD

Aassumptions:Aassumptions:

1 mile segment1 mile segmentss ~ 0.1~ 0.1--1.0 mile1.0 mileqqA A = 2000 vph= 2000 vphqqC C = 1800 vph= 1800 vphvvf f = 60 mph= 60 mphvvc c = 30 mph= 30 mphvvCD CD = = --17.1 mph17.1 mphvvAC AC = = --7.5 mph7.5 mph


0

500

1000

1500

2000

2500

0 50 100 150 200Density (veh/mi)

Flo

w (

veh

/h

r)

qA =2000

qC =1800

vf =60

vc =30vCD = -17.1

D

CB

AvAC = -7.5


0

500

1000

1500

2000

2500

0 50 100 150 200Density (veh/mi)

Flo

w (

veh

/h

r)

Real data I-5 Macadam2/8/07


Types of TransitionsTypes of TransitionsTypes of Transitions

Time

Dis

tan

ce Frontal stationary

Backwardrecovery

Forwardrecovery

Forward forming

Backwardforming

Rear stationary



Time

Dis

tan


Backwardrecovery

Forwardrecovery

Forward forming

Backwardforming

Rear stationary


d =

l=

Segm

ent L

ength

x

t

A, D, DA: uncongested C: congested

vf vc

ttf = l/vf ttc = l/vc

Estimation When HomogeneousEstimation When Homogeneous

VHT Actual = VHT EstimatedVHT Actual = VHT Estimated



Time

Dis

tan


Backwardrecovery

Forwardrecovery

Forward forming

Backwardforming

Rear stationary



Time

Dis

tan


Backwardrecovery

Forwardrecovery

Forward forming

Backwardforming

Rear stationary



Time

Dis

tan


Backwardrecovery

Forwardrecovery

Forward forming

Backwardforming

Rear stationary



A

B

C

D

A

x

t

bn

tdeact

vCDvAC

vfDAAC CD


t

x

svc

vf

tc

vAC

Congestion Signalα = Lag Time

j1

Shockttf

l

z j3

j2

umaxttf

Transition UncongestedTransition Uncongested CongestedCongested


t

x

svc

vf

tc

vAC


j1

Shockttf

l

z j3

j2

umaxttf



t

x

svc

vf

tc

vAC


j1

Shockttf

l

z j3

j2

umaxttf



t

x

svc

vf

tc

vAC


j1

Shockttf

l

z j3

j2

umaxttf



t

x

svc

vf

tc

vAC


j1

Shockttf

l

z j3

j2

umaxttf


UNDERUNDER OVEROVER


t

x

svc

vf

tc

vAC


j1

Shockttf

l

z j3

j2

umaxttf



x

l

s=0.50

x

s=0.25

x

s=0.10

t

l

t

x

s=0.33

Sensor Density Affects Lag TimeSensor Density Affects Lag Time



0.0 0.2 0.4 0.6 0.8 1.0Sensor Spacing (miles)

7.0

7.5

8.0

8.5

9.0

9.5

VHT

Predicted VHT

Actual VHT



-25%

-20%

-15%

-10%

-5%

0%

5%

10%

15%

20%

25%


VHT

Erro

r Error (Penalty)

Error (Absolute Value)

OVER

UNDER

Error (Additive)



-25%

-20%

-15%

-10%

-5%

0%

5%

10%

15%

20%

25%


VHT

Erro

r

7.0

7.5

8.0

8.5

9.0

9.5

VHT

Actual VHT

Error (Additive)

Error (Penalty)


Predicted VHT



A

B

C

D

A

x

t

bn

tdeact

vCDvAC

vfDAAC CD


t

x

s

tr

vCD

Recovery Signalα′ = Lag Time

vc

vf

j1

Wavettc

l

zj3j2

umax

ttf

Transition CongestedTransition Congested UncongestedUncongested


t

x

s

tr

vCD


vc

vf

j1

Wavettc

l

zj3j2

umax

ttf


UNDERUNDEROVOV ERER


t

x

s

tr

vCD


vc

vf

j1

Wavettc

l

zj3j2

umax

ttf


UNDERUNDEROVEROVER


t

x

s

tr

vCD


vc

vf

j1

Wavettc

l

zj3j2

umax

ttf


UNDERUNDEROVOV ERER




3.8

3.9

4.0

4.1

4.2

4.3

4.4

4.5

4.6

4.7

VHT

Predicted VHT

Actual VHT



-15%

-10%

-5%

0%

5%

10%

15%

20%

25%

30%

35%


VHT

Erro

r

Error (Additive)

Error (Penalty)


OVER

UNDER



-15%

-10%

-5%

0%

5%

10%

15%

20%

25%

30%

35%


VHT

Erro

r

3.8

3.9

4.0

4.1

4.2

4.3

4.4

4.5

4.6

4.7

VHT

Error (Additive)

Error (Penalty)


Predicted VHT

Actual VHT



2.0

2.5

3.0

3.5

4.0

4.5

5.0

VHT

Actual VHT

Underprediction OnlyUnderprediction Only

Predicted VHT


15%

16%

17%

18%

19%

20%

21%


VHT

Erro

r Error



15%

16%

17%

18%

19%

20%

21%


VHT

Erro

r

2.0

2.5

3.0

3.5

4.0

4.5

5.0

VHT

Error

Predicted VHT

Actual VHT



Considering Both TransitionsConsidering Both Transitions


11.1

11.3

11.5

11.7

11.9

12.1

12.3

12.5

12.7

12.9

13.1

VHT

Predicted VHT

Actual VHT



-15%

-10%

-5%

0%

5%

10%

15%

20%

25%


VHT

Erro

r

Error (Additive)

Error (Penalty)




-15%

-10%

-5%

0%

5%

10%

15%

20%

25%


VHT

Erro

r

11.1

11.3

11.5

11.7

11.9

12.1

12.3

12.5

12.7

12.9

13.1

VHT

Error (Additive)

Error (Penalty)


Predicted VHT

Actual VHT


t

x

svc

vf

tc

vAC

Congestion Signalttf=α = Lag Time

j1

Shock

l

z j3j2

umaxttf

Detector at End of SectionDetector at End of SectionDetector at End of Section


t

x

svc

vf

tc

vAC

Congestion Signalttf=α = Lag Time

j1

Shock

l

z j3j2

umaxttf

Detector at End of SectionDetector at End of SectionDetector at End of Section


Effects of Detector at EndEffects of Detector at EndEffects of Detector at End

Regime ACUnderprediction

DetectorLocation

Umax(Min)

Lag Time(Min)

VHT/MilePred

VHT/MileAct

Under %Error

Midpoint 0.56 4.00 2.78 3.55 22%

Downstream 0.11 0.00 0.56 0.59 5%

OverpredictionDetectorLocation

Umax(Min)

Lag Time(Min)

VHT/MilePred

VHT/MileAct

Under %Error

Midpoint 0.56 4.00 4.44 3.95 -13%

Downstream 0.11 0.00 8.89 6.91 -29%


The FutureThe Future

Better estimates and forecastsFusion of fixed and mobile sourcesUbiquity of integrated information

In-vehicleIn-deviceMultimodal routing/decision-makingCustomizableBetter management

Better estimates and forecastsFusion of fixed and mobile sourcesUbiquity of integrated information

In-vehicleIn-deviceMultimodal routing/decision-makingCustomizableBetter management


Seminar PerspectivesSeminar Perspectives

Inspired by Berkeley’s Transportation Science Seminar, originated by G.F. Newell, 1965

First seminar October 5, 2000, Benefits of Archived ITS Data: Measuring Capacity at a Freeway Bottleneck

200 seminars completed

Began streaming video October 2002: 165available for download and streaming

Began podcasts (mp3) in October 2007: 30podcasts now available

Venue for student/faculty interaction

Strong involvement of transportation community

Inspired by Berkeley’s Transportation Science Seminar, originated by G.F. Newell, 1965

First seminar October 5, 2000, Benefits of Archived ITS Data: Measuring Capacity at a Freeway Bottleneck

200 seminars completed

Began streaming video October 2002: 165available for download and streaming

Began podcasts (mp3) in October 2007: 30podcasts now available

Venue for student/faculty interaction

Strong involvement of transportation community


PodcastsPodcastsPodcasts


Seminar PerspectivesSeminar Perspectives

Organized by graduate students?More social interaction before/after?More point/counterpoint?We’re open to other ideas!First air transportation seminar on May 9Other topics we haven’t covered?

Organized by graduate students?More social interaction before/after?More point/counterpoint?We’re open to other ideas!First air transportation seminar on May 9Other topics we haven’t covered?


AcknowledgementsAcknowledgements

Travel time project teams, past and presentKristin Tufte, Sirisha Kothuri, David Lovell, Ben Zielke, Rafael Fernandez, Ed Anderson Sutti Tantiyanugulchai, Roger Lindgren, Monica LealGalen McGill, Jack Marchant, Dennis Mitchell, Oregon Department of TransportationPortland State University Distance Learning Center for making the streaming easyMy colleagues Jennifer Dill, Chris Monsere and John GliebeSeminar enthusiasts and participantsRyan Gratzer

Travel time project teams, past and presentKristin Tufte, Sirisha Kothuri, David Lovell, Ben Zielke, Rafael Fernandez, Ed Anderson Sutti Tantiyanugulchai, Roger Lindgren, Monica LealGalen McGill, Jack Marchant, Dennis Mitchell, Oregon Department of TransportationPortland State University Distance Learning Center for making the streaming easyMy colleagues Jennifer Dill, Chris Monsere and John GliebeSeminar enthusiasts and participantsRyan Gratzer


200th Seminar! Travel Time Estimation for Traffic Management and Traveler Information

Documents

traffic management

multimodal travel time

modes7travel time estimation

it5travel time estimation

traveler informationwhat

traveler informationrobert

traveler informationhow

traveler informationif