Advances in Large-scale Microsimulation-based Dynamic Traffic Assignment



Advances in Large-scale Microsimulation-based Dynamic Traffic Assignment

Authors

Daniel Morgan, Caliper Corporation, Andres Rabinowicz, Caliper Corporation, Howard Slavin, Caliper Corporation

Description

This abstract describes a microsimulation-based dynamic traffic assignment (DTA) model that retains the fidelity of microsimulation while achieving running times that are practicable at the regional scale. A case study is presented.

Abstract

The North Florida Transportation Planning Organization (TPO), the Metropolitan Planning Organization (MPO) for the Jacksonville, FL metropolitan area, is among a growing body of MPOs in the United States and around the world moving toward activity-based model (ABM) deployment in the United States. The TPO leveraged the recently developed and validated Northeast Regional Planning Model (NERPM) ABM as the modeling platform for its 2040 Long Range Transportation Plan (LRTP).

Whereas the ABM supports exploration of a wider spectrum of policy considerations through a disaggregate model of traveler behavior and decision-making, the model continues to rely on a static traffic assignment to determine the transportation implications of those considerations. Thus, the rich level of detail achieved in the space domain, through use of parcels to represent trip origins and destinations, and in the time domain, through a finer-grained representation of departure time choice, is abandoned in a traditional zone-based, peak period-based traffic assignment.

Further, analysis of the impacts of transportation infrastructure projects, such as the construction of proposed high occupancy/toll lanes, whose performance would be directly influenced by the temporal and geospatial demand patterns that the ABM is capable of supplying is instead reliant on a conventional traffic assignment model that is insensitive to the effects of those patterns.

We present a microscopic traffic simulation-based dynamic traffic assignment (DTA) model developed as a substitute for the static traffic assignment model in applications requiring more accurate treatment of traffic flow phenomena, system performance, and/or level of service. The DTA model serves instead as a complementary tool extending the ABM into higher-fidelity, operationally-sensitive analyses.

The paper describes the DTA model and its various inputs, including traffic signal timings, and the tools that make it scalable for development and deployment in a great many metropolitan areas. Short of a rigorous calibration, we demonstrate the model’s goodness-of-fit with traffic counts as compared to the static traffic assignment. The paper also discusses the convergence and running time properties of the DTA.

To demonstrate the model and its design, we present an application of DTA in integration with the NFTPO’s ABM. The treatments of demand in the ABM and in the DTA model are made compatible in order to achieve a productive integration. The DTA model accepts lists of individual trips rather than aggregate trip matrices and operates on a network representation sufficiently dense so that the ABM's activity locations, based on parcels, have proximate access to local streets in the transportation network. However, the reliance of the NFTPO on a trip-based model for truck and external trip matrices presents challenges to integration. The design of the DTA’s interface with the ABM accommodates both the disaggregate and aggregate natures of the ABM and trip-based models, respectively.

Provided the unified list of trips from the ABM and trip-based model forecasts, the DTA model simulates critical operational and behavioral phenomena such as route choice, response to traveler information, and capacity reductions due to supply disruptions (e.g., incidents or work zones), making it possible for the NFTPO to connect their ABM directly to a tool for real-world scenario analyses.

The DTA is run in an iterative framework, in which drivers adjust their route choices each iteration of the simulation based on the experiences of prior simulations. Hence, route choices are appropriately sensitive to the network effects of signalized delay on coordinated or uncoordinated arterials, weaving and merging on freeway facilities, and bottlenecks that arise from the specific geometries of access and egress areas to and from managed or HOV lanes – phenomena that can only be accurately captured through the explicit vehicle-vehicle dynamics and vehicle-ITS interactions that are the strong suit of microscopic simulation.

To demonstrate the DTA model, we explore the convergence of the DTA and running times by iteration. Further, we present preliminary calibration results comparing the DTA model’s goodness of fit with traffic counts across the metropolitan area as compared to that of the static traffic assignment on which the ABM still relies.

Publisher

Association for European Transport