VII Data Characteristics for Traffic Management: Task Overview and Update 21 June 2006 Karl Wunderlich Fellow, Transportation Analysis.

VII Data Characteristics for Traffic Management:

Task Overview and Update

21 June 2006

Karl Wunderlich

Fellow, Transportation Analysis

2ControlNumber

Scope

Examine capability of VII probe data to support (specifically):– Signal control– Ramp metering– Traveler information

This capability must be examined with respect to key variables:– Facility type (arterial/freeway/rural) and geometry– Congestion levels and road/weather conditions– Market penetration– VII probe message management

• In-vehicle• At the roadside and in backhaul communication

Near-term analytical emphasis is on the support of Day 1 applications– For example, off-line periodic signal retiming versus “real-time”

adaptive signal control

3ControlNumber

Objectives Identify likely content of collected VII probe messages

passed to traffic managers or traveler information service providers under realistic conditions

Develop (where possible) algorithms that will estimate key measures from the collected probe data, for example:– Vehicle volumes by lane and turning movements– Travel times and intersection delays

Estimate the accuracy of these algorithms with respect to the key variables from previous slide (e.g., market penetration)

Provide USDOT with an understanding of key tradeoffs along a spectrum of issues/conditions (e.g., privacy)

4ControlNumber

Staffing and Coordination

Mitretek Systems Team

Michael McGurrinKarl Wunderlich

Meenakshy VasudevanEmily Parkany

Phil Tarnoff, U-Md.

USDOT Task Manager

Brian Cronin

Use Case Development(BAH)

VII Data Elements(PB)

Key VII-Related Activities

5ControlNumber

Approach

Data needs assessment– Define the data required by traffic management and traveler

information applications– Qualitative assessment of data produced by VII to meet these

identified needs

Analytical assessment of VII probe data– Develop an analytical tool that takes…

• Vehicle trajectory data• Specific probe message management strategy• Assumed RSE deployment

… and produces the associated VII probe data content– Trajectory data will come from a variety of sources:

• Observed (e.g., NGSIM or floating car data)• Simulated (e.g., from a traffic simulation)

– Develop algorithms to process this probe data into measures of interest (e.g., link travel time)

6ControlNumber

VII Data CharacteristicsTask

Data NeedsAssessment

1/1

Data Needs White Paper

3/1 4/1 5/1 6/1 7/1 8/1 9/1 10/1 11/1 12/1

PrelimMatrix

Analytical ToolDevelopment and

Evaluation

Kickoff Briefing 1 Briefing 2

DownselectStrategies (I)

Draft WP

Day 1 Final Report(draft)

Acquire/Prep Trajectory Data

Build TrajectoryConverter

AcquireTraffic Simulationand Test Networks

Enhance Converter

Briefing 3

write-up

Multi-RSE Strategy Evaluation

TradeoffAnalyses

write-up

revisionsData

Characteristics WP (final)

Coordination/Progress Briefings

DownselectStrategies (II)

Initial StrategyDevelopment

Initial Strategies

Assess Needs

Observed Data Track

SimulationTrack

ExpendedFunding

RemainingFunding

Width indicates relative Mitretek LOE Deliverables

Co

mp

lete

d

Pla

nn

ed(+

in

tern

al d

raft

)

1FTE

Der. Algs. (II)

12 June 2006

Planned

Coordination Meetings

Completed

Bi-Weekly Status Updates(Scheduled)

Brian/Karl

Full Team

Completed

Validate Sim Trajectories

Preliminary Strategy Evaluation

Derivation Algs. (I)

7ControlNumber

Key Deliverables Data Needs White Paper (completed)

– Broad, qualitative assessment of Day 1 and later needs

Applications Preliminary Requirements Matrix (completed)

– High-level assessment of the capability of VII data to meet the identified short- and long-term needs

Data Characteristics White Paper (1 September 2006)

– Summary of findings, primarily from observed data analysis

– Initial assessment of capability of VII probe data to support Day 1 applications

Draft Day 1 Final Report (1 January 2007)

– Update and expansion of the September white paper

– Results from the analysis of simulated trajectories

– More comprehensive assessment of key tradeoffs

8ControlNumber

From Trajectories to Measures

Time

Po

siti

on

VehicleTrajectories

Extract SampleDepending on Market Penetration

12 3

4

PopulateWith SnapshotsAccording to Message HandlingStrategy

ProcessSnapshotsTo EstimateMeasures

TravelTime

QueueLength

Other

9ControlNumber

Observed Data Sets, Floating Car:Strengths and Weaknesses Floating car trajectory data

– Strengths:

• Trajectories are long (30+ miles in some cases)

• Arterial, freeway, rural road facilities

• Light to heavy congestion conditions

• Some “other data” collected that looks like VII data elements (e.g., weather or turn signal disposition)

– Weaknesses:

• Only one vehicle tracked

• Ground truth measures can’t be directly observed for aggregate traffic flow – just one vehicle

Will be most valuable for looking at travel time derivation issues over longer links, potentially widely dispersed RSEs

Road Weather Management

10ControlNumber

Observed Data Sets: NGSIM

NGSIM data are high-resolution vehicle trajectory data– Processed video images from multiple high-angle cameras– Near 100% of all vehicle positions traced at 0.1 sec intervals– Detailed lane position and disposition to other vehicles– Two freeway data sets, one arterial data set

Strengths: 100% vehicle coverage Weaknesses: Short coverage areas (under 1 km)

11ControlNumber

Simulated Vehicle Trajectories:Strengths and Weaknesses Simulated trajectory data

– Strengths:

• Most facilities of interest can be modeled

• 100% tracking of vehicles

• Ground truth measures can be directly obtained

• Congestion levels and other elements can be systematically adjusted

– Weaknesses:

• Validity of detailed trajectories under congestion is poorly understood

• Time and effort to build and calibrate realistic networks

Will be most valuable when attempting to deal with incremental tradeoffs for key issues like market penetration and buffer size

Sample Trajectory Conversion:Columbus, Ohio: Route 33 and I-270

• Run Type : GPS (Floating Car)• Distance: 62.0 Miles• Travel Time: 93.8 Minutes• Average Speed: 39.6 mph• RSE Spacing : 2.3 miles between RSEs (on average)• Snapshots per Mile: 10.0• Vehicle IDs (Transmit/Produced): 32 / 42• Snapshots per ID (Transmit/Produced): 9.4/13.7• Total number of Snapshots: 618

– Stop Snapshots: 23– Start Snapshots: 13– Periodic Snapshots: 582

Columbus, OhioExpected RSE Location, GPS Trace

Walk-Through of Default VII Probe Message Process

• Location:– A congested segment on I-270

• What we will examine:– 50 Snapshots taken right after vehicle RSE interaction

• Time– 3133 to 3448 seconds (5.25 minutes)

• Distance:– 1.9 Miles

I-270 Route

Time 3133-3244 (1.85 Min)0.69 Miles

43 secs (7 SS)

Spd 20-28

12 secs (4 SS)

Spd 0-9T 3196

(1 Stop)

48 Secs(1 Start)Spd 10.5

Periodic 11 Stop 1 Start 1 Capacity 13/30Periodic 0 Stop 0 Start 0Deleted

Time 3244 – 3356 (1.87 mins)0.69 Miles

12 secs 4 SS

Spd 12-19T 3356

(1 Stop)

Periodic 27 Stop 2 Start 1 Capacity 30/30

29 secs 6 SS

Spd 22-32

14 secs 2 SS

Spd 43

1 secs 1 SS

Spd 19

20 secs 7 SS

Spd 4-12

Buffer is full 3.25 mins after the last vehicle RSE Interaction

Periodic 4 Stop 0 Start 0Deleted

Deleted from SS from Time 3133- 3151 (0.3 mins)

Time 3356- 3448 (1.9 mins) 0.54 Miles

4 Secs(1 Start)Spd 11.0

20 Secs(5 SS)

Spd 13-19

41 Secs(10 SS)

Spd 13-19

Periodic 26 Stop 2 Start 2 Capacity 30/30Periodic 20 Stop 0 Start 0Deleted

Deleted from SS from Time 3133- 3276 (2.4 mins)

Does not report to a RSE for

another 4.7 Mins

Deleted SS Time 3133-3244 (1.85 Min) 0.69 Miles

Deleted SS Time 3244 – 3356 (1.87 mins) 0.69 Miles

95 additional snapshots are deleted beforeThe vehicle interacts with another RSE

Deleted Snapshots by Location

First RSEInteraction

Last RSEInteraction

Estimating Travel Time from Snapshots

OVERALLActual = 94 minutes

Calculated = 67 minutesError = 29%

A =321C = 159E = 50%

A = 574C = 242E = 58%

A =100C = 120E = 20%

A =279C = 220E = 21%

A = 160C = 195E = 22%

A =363C = 292E = 20%

A = 151C = 174E = 15%

A = 120C = 100E = 17%

A =200C = 179E = 11%

Actual = 234 secCalculated = 236 sec

Error = 1%

A = 260C = 262E = 1%

A = 460C = 483E = 5%

A = 154C = 168E = 9%

Preliminary Observations

• For uncongested conditions:– the default strategy provides fairly good geographic

coverage and accuracy

• For congested conditions even with relatively closely spaced RSEs:– The default plan results in significant buffer overflow – The deleted snapshots leave significant geographic

gaps– Gaps have impact on accuracy of travel time

estimation

Analysis: Next Steps

• Evaluate more Data Sources– Columbus, Ohio GPS – Salt Lake City, Utah I-15 GPS runs – Dulles Toll Road GPS runs– I-66/Route 50 GPS runs– NGSIM validation data

• Evaluate VISSIM simulated runs• Test alternative thresholds and strategies for VII

probe message process • Test sensitivity to a range of RSE locations and

densities