SHRP 2 Project L04 Incorporating Reliability Performance Measures in Operations and Planning Modeling Tools Reliability Technical Coordinating Committee Briefing

1
SHRP 2 Project L04Incorporating Reliability
Performance Measures in Operations and Planning
Modeling ToolsReliability Technical
Coordinating CommitteeBriefing
in partnership with

• National Academy of Sciences
• Irvine - April 8, 2010

2
Agenda

• Project Overview Methodology
• Data and Candidate Networks
• Anticipated products of the research
• Work Program Discussion

3
Methodology Framework Three Components to
Incorporate Reliability in Network Simulation
Models
Exogenous sources Input Scenario manager Scenario manager
Exogenous sources Input Demand - Special events - Day-to-day variation - Visitors - Closure of alternative modes Supply - Incidents - Work zones - Adverse weather
Endogenous sources Reliability-integrated Simulation model (meso, micro) Improvements to existing simulation tools Improvements to existing simulation tools
Endogenous sources Reliability-integrated Simulation model (meso, micro) Demand - Heterogeneity in Route Choice and User Responses to Information and Control Measures - Heterogeneity in vehicle type Supply - Flow breakdown and incidents - Heterogeneity in driver behavior (car following, lane changing) - Traffic control - Dynamic pricing
Performance measures Output Vehicle trajectory processor Vehicle trajectory processor
Performance measures Output - Travel time distribution - Reliability performance indicators - User-centric reliability measures - Travel time distribution - Reliability performance indicators - User-centric reliability measures
4
Integration in Planning Models

• Reliability-sensitive network equilibrium models
• Reliability affects travelers mode, departure
  time and route choice.
• Reliability measures are produced from the
  simulation models and fed back to the demand
  models.
• Iterate between demand models and network
  simulation until convergence to UE (or SUE).
• Output performance measures for policy evaluation
  and network planning/design.

5
Model Exogenous Sources Scenario Manager

• Scenario-based approach
• Construct discrete scenarios
• Conduct single-point estimation to produce
  results for each what if scenario
• Monte Carlo sampling
• Randomize demand and/or supply side parameters
  and establish the corresponding probability
  distribution functions.
• Conduct Monte Carlo simulation with regard to
  these random parameters
• Scenarios involving equilibrium traffic
  assignment
• Perform iterative equilibrium assignment for
  scenarios involving medium to long term changes
  in demand or capacity

6
Model Endogenous SourcesRoute Choice Behavior

• Route choice behavior and travel time reliability
  interact
• Reliability is a result of travel decisions
• Reliability affects route choice behavior
• Reliability in generalized cost function
• Heterogeneity in route choice behavior

7
Model Endogenous SourcesHeterogeneity in
Driving Behavior

• Microscopic simulation models
• Vehicle-related parameters, e.g. length, maximum
  acceleration/ deceleration, reaction time, safety
  distance, desired speed, desired
  acceleration/deceleration, Maximum give-way time
• Link-related characteristics, e.g. speed limits,
  visibility distance at junctions, maximum turning
  speed, slope (grade), reaction time variation
• Heterogeneity in car-following and lane-changing
  behavior, especially in the presence of heavy
  vehicles
• Mesoscopic simulation models
• Heterogeneity in vehicle types
• Varying and context-dependent impact on traffic
  performance

8
Model Endogenous SourcesFlow Breakdown and
Incidents

• Characterize flow breakdown as a collective
  phenomenon
• Probability of breakdown
• Breakdown duration
• Characterize flow breakdown and incident through
  individual decisions
• Describe driver behavior under extreme and
  incident conditions

9
Model Endogenous SourcesState-Dependent Traffic
Control

• State-dependent traffic controls - dynamically
  adjust the control variables based on the
  prevailing (or predicted) traffic conditions, for
  more effective management.
• State-dependent controls may introduce another
  source of unreliability/unpredictability to the
  system.
• Actuated signal control
• Ramp metering
• Variable message signs
• Dynamic pricing

10
Vehicle Trajectories Unifying Framework for
Micro and Meso Simulation

• Vehicle (particle) trajectories in the output of
  a simulation model enable
• construction of the path and O-D level travel
  time distributions of interest
• extraction of link level distributions
• Vehicle trajectories could be obtained from both
  micro- and meso-level simulation models
• Trajectories also obtained from direct
  measurement in actual networks, enabling
  consistent theoretical development in connection
  with empirical validation.

11
Vehicle Trajectory Processor
12
Methodology Framework Summary
Exogenous sources Input Scenario manager Scenario manager
Exogenous sources Input Demand Special events Day-to-day variation Visitors Closure of alternative modes Supply Incidents Work zones Adverse weather
Endogenous sources Reliability-integrated Simulation model (meso, micro) Improvements to existing simulation tools Improvements to existing simulation tools
Endogenous sources Reliability-integrated Simulation model (meso, micro) Demand Heterogeneity in Route Choice and User Responses to Information and Control Measures Heterogeneity in vehicle type Supply Flow breakdown and incidents Heterogeneity in car following behavior Traffic control Dynamic pricing
Performance measures Output Vehicle trajectory processor Vehicle trajectory processor
Performance measures Output Travel time distribution Reliability performance indicators User-centric reliability measures Travel time distribution Reliability performance indicators User-centric reliability measures
13
Modeling Platform Requirements Model Types
Roles
Type of Model Role in Framework
Planning(demand forecasting models) Provide traffic demand input to simulation models Demonstrate the use of reliability measures for route / mode choices (and potentially departure time choice) in an integrated demand-supply framework
Operations(meso/micro-simulation models) Incorporate parameters affecting travel time variability at operations level (supply side) Interface with Scenario Manager to obtain input based on exogenous sources / parameters Generate (trajectory-based) travel time output for reliability assessment Interface with Trajectory Processor to provide output for development of travel time distributions, reliability performance indicators user-centric measures
14
Planning Model Requirements

• Ability of planning model to use quantitative
  measures of travel time variability in demand
  forecasting processes (i.e., beyond the common
  practice of using average travel time and cost)
• expected travel time
• schedule delay
• travel time standard deviation (inferred vs
  experienced)
• Ability to achieve at least some consistency
  between simulation-generated reliability measures
  and those used in mode / route / departure time
  choice models
• Preference for activity-based planning models in
  order to incorporate schedule delay and other
  micro-level, reliability-related measures

15
Operations (simulation) Model Requirements

• Ability to address most typical urban/suburban
  type of traffic conditions
• vehicle/particle-based computational approach
  fidelity
• uninterrupted and interrupted flow with various
  types of facilities (incl. managed lanes) and
  control (signalized, stop/yield, etc.)
• multi-vehicle classes (auto, truck, bus),
  preferably with varying characteristics
• multi-simulation periods
• Ability of underlying submodels (route choice,
  lane choice, etc.) to endogenize certain
  variability sources
• route choice and driver behavior heterogeneity
• incident and flow breakdown characteristics
• state-dependent traffic control
• Ability to generate vehicle/particle-based
  trajectories
• may require open-source models or access to
  code

16
Software Code Access / Modification Requirements

• Ability to access / tweak programming code for
  endogenizing time variability sources / factors
• some software developers would be keen to assist
  (depending on level of effort involved)
• Open source sub-models (e.g., NGSim-developed
  lane change and other models)
• already available in some software packages
  (Dynasmart, Aimsun, Vissim)
• Various forms of intervention through programming
  tools (API)
• available for most commonly used simulation
  platforms in North America(Paramics, Vissim,
  Aimsun, Transmodeler, Dynasmart, Dynameq, Vista,
  etc.)

17
Data Requirements

• Traffic data for model adaptation / re-validation
• Ancillary data for parameterization of time
  variability sources (endogenous
  exogenous)e.g., special events, incidents,
  weather
• Travel time data for
• reliability analysis / concept confirmation
• model output verification / checking

18
Travel Time Data

• Trajectory-based by vehicle trip(X, Y
  coordinates and time stamp)
• Capturing both recurring and non-recurring
  congestion on a range of road facilities (from
  freeways to arterial roads and possibly managed
  lanes)
• Sufficient sampling and time-series to allow
  statistically meaningful analysis
• Ability to tie travel time data to ancillary
  data for time variability sources (to allow
  parameterization for simulation testing purposes)

19
Potential Data Sources / Inquiries made to date

• GPS- and Cell-probe data provide most promising
  prospects for large scale spatial and temporal
  coverage
• INRIX (national)
• NAVTEQ (national)
• MyGistics (Chicago region)
• Google (national) -no response
• ITIS and FCD for validation (Missouri)
• Calmar truck data (California, New York, etc.)
• Intellione (Toronto) -prelim. tests undertaken
• major navigation services provider -prelim.
  tests undertaken

20
Preliminary Data Tests to date
21
Demo Site Selection Considerations

• large urban/suburban area
• typical congestion-related travel time
  variability characteristics
• existing models that meet L04 technical approach
  / simulation functional requirements
• network size /configuration for meaningful
  measurement of time variability
• vehicle trajectories / time distributions
• data availability
• primarily trajectory travel times
• other considerations
• willingness of jurisdictional authority to
  participate in the project and/or provide data
  and base model
• familiarity of research team staff with candidate
  network, data and model

22
Potential Sites - (best candidates so far noted
with )

• Atlanta(trajectory data availability concerns)
• Baltimore - Washington DC area
• California (San Francisco / Bay Area)
• Chicago(cost considerations may be prohibitive)
• New York City / Metro Area (most model
  requirements already met, wide-area GPS data from
  various sources)
• Toronto (most models already in place or close
  to completion, wide-area GPS cell probe data)
• Montreal(models in place, GPS data can be
  arranged, institutional/jurisdictional concerns)
• other areas (Seattle, Phoenix, Detroit, Austin)

23
Project Products

• Reports
• Phase I reviews in detail fundamental approach,
  includes supporting data, candidate networks,
  reliability measures
• Phase II reports the results of model calibration
  and validation, includes guidelines and materials
  for full replication of phase II
• Phase III report incorporates reliability into
  travel models
• Outreach
• Pilot demonstrations of the simulation model
• Brochure, website, how to CD
• Information sessions and demonstrations
• Visualization tools

24
Product Audience for SHRP 2 L04

• Practitioners and researchers
• Software vendors and developers
• Operations managers, planners in transportation
  agencies interested in practical implications