SAN FRANCISCO COUNTY TRANSPORTATION AUTHORITY San Francisco DTA Project: Model Integration Options Greg Erhardt DTA Peer Review Panel Meeting July 25 th,

SAN FRANCISCO COUNTY TRANSPORTATION AUTHORITY

San Francisco DTA Project: Model Integration Options

Greg Erhardt

DTA Peer Review Panel MeetingJuly 25th, 2012

SAN FRANCISCO COUNTY TRANSPORTATION AUTHORITY

Agenda

• 9:00 Background• 9:30 Technical Overview – Part 1

• Development Process and Code Base/Network Development

• 10:15 Break• 10:30 Technical Overview – Part 2

• Calibration and Integration Strategies

• 12:00 Working Lunch / Discussion• 2:00 Panel Caucus (closed)• 3:30 Panel report• 5:30 Adjourn

3

Outline

• Foundations• Extracting LOS from DTA• Feeding LOS to SF-CHAMP• Transit Integration• Feeding Additional Information to DTA• Key Questions and Next Steps

Foundations

5

Goals of Integration

• Better consistency between supply & demand models to avoid unreasonable gridlock

• Better LOS information for demand models• Temporal differences• Operational details• Reliability

• Better demand information to DTA• User heterogeneity

6

Temporal Resolution

Microsimulation ABM

Microsimulation DTA

List of individual

trips

Aggregate LOS skims

for all possible trips

Microsimulation ABM

Microsimulation DTA

List of individual

trips

Individual trajectories

for the current list of

trips

LOS for the other potential

trips?

Integration of ABM and DTA (Direct)

Integration of ABM and DTA (Aggregate Feedback)

Source: SHRP2 L04 Draft Report December 2011

7

Convergence & Stability

• Averaging Methods• MSA or similar• Applicable where continuous variables

are being averaged• Enforcement Methods

• Unique to microsimulation• Examples:

• Gradually freeze a portion of households or decisions

• Analytically discretizing mode choices

Extracting LOS from DTA

9

Evaluate Levels of Temporal Resolution

• At what level do travelers think about time and cost? 1 hour? 15 minutes? 1 minute?

• How much do travel times and cost change within 15 minutes? 1 hour? 3 hours?

10

Spatial & Temporal Expansion

• SF-CHAMP covers 9-county area, so need to merge DTA skims with static

• Currently only modeling PM peak—use 24-hour DTA for feedback of all periods

11

Measure Reliability

• Could attempt to measure reliability within time periods

• Could attempt to measure reliability across days

• Reference SHRP2-L04

Feeding LOS to SF-CHAMP

13

Integrate with Trip Time-of-Day Model

• Existing trip TOD model based on time shift from preferred departure time

• Uses static ½ hour skims

• Autos only• Could

substitute dynamic skims

Effect of Time Shift on Utility

14

Feedback Skims in Existing 5 Time Intervals

• SF-CHAMP currently accepts skims based on 5 periods: Early AM, AM Peak, Midday, PM Peak, Night

• Tour TOD model operates at this resolution

• DTA times could be averaged to these periods

• Alternately, logsums from a trip TOD model could be used and fed back to SF-CHAMP at this resolution

15

Feedback Skims with More Disaggregate Time Intervals

• Feedback dynamic skims in 1 hour, 30 minute, or 15 minute resolution

• Would require replacing SF-CHAMP’s tour time-of-day model

• Provides additional sensitivity to time-of-day differences in upstream models

16

Feedback Individual Vehicle Trajectories

Microsimulation ABM

Microsimulation DTA

List of individual

trips

Individual trajectories

for the current list of

trips

Consolidation of individual schedules (inner loop for departure / arrival time

corrections)

Sample of alternative origins, destinations, and departure times

Individual trajectories for potential

trips

Source: SHRP2 L04 Draft Report December 2011

Possible Scheme for Fully Disaggregate Integration

17

Fully Disaggregate ABM-DTA Integration

Fully disaggregate models on both ends could allow: • Fully consistent daily schedule for

each traveler, adapting to differences in planned versus actual travel times

• Moving unit of analysis in DTA from trip to a tour, allowing for the timing of stops to be accounted for in the DTA

• Possible representation of user heterogeneity in DTA

• Sampling of alternatives removes dependency on TAZs and allows any level of spatial disaggregation

18

Behavioral Considerations

Aggregate Feedback

• Assumes day is planned ahead based on average conditions

• All possible zones/time periods included in feedback

• Upstream model decisions based on averages or logsums

Disaggregate Feedback

• Allows adaptation based on conditions encountered during the day

• Sampling of alternatives assuming limited traveler information

• Upstream model decisions based on specific travel times

How do we think people make decisions?

19

Practical and Policy Considerations

• How to integrate with transit, and transit time-of-day choice?

• Would the result be stable across scenarios? Would additional disaggregation propagate simulation noise?

• How to achieve convergence—averaging versus enforcement?

• Disaggregation could allow flexibility in spatial definitions

• Level of temporal aggregation should be related to policies being considered

What do we think is useful?

Transit Integration

21

Use Existing Transit Pathbuilder

Several options: 1. Code transit trajectories in RUNTIME

field2. Attach average DTA link travel time

to static network used for building transit skims

3. Parse out transit trajectories to individual links

4. Modify SF-CHAMP to accept transit skims for shorter time periods

5. Incorporate a transit trip departure time model

22

Implement Dynamic Transit Assignment

• Could use FastTrips• Could incorporate:

• Dwell times based on boardings and alightings

• Bus bunching• Delays due to roadway congestion

Feeding Additional Information to DTA

24

User Heterogeneity in DTA

• In SF-CHAMP, each traveler has their own value of time

• This could be incorporated into DTA through additional user classes• Requires additional runtime

• Alternately, individual vehicles in the simulation could be assigned different values of time• Requires restructure of DTA software

Key Questions & Next Steps

26

Key Questions

• Should we expand temporally (to 24-hours)?

• Should we expand spatially (to 9-counties)?

• At what temporal resolution do people make:

• Routing decisions?• Path type choice (toll vs. no-toll) decisions?• Trip departure time decisions?• Mode choice decisions? • Destination choice decisions? • Tour and activity scheduling decisions? • Tour and activity participation decisions?

• How can transit be modeled? • How can reliability be measured?

27

Next Steps

• Evaluate what is feasible within this project

• Consider which approaches offer long-term promise

Questions?