Integrated ABM-DTA Model: First Application Experience and Lessons Learned
Corresponding Author: Peter Vovsha, WSP
Presented By: Peter Vovsha, WSP
The paper describes fist experience of development and application of a model system that integrates an Activity-Based travel demand Model (ABM) and Dynamic Traffic Assignment (DTA), taking a maximum advantage of the disaggregate nature of both models. With this approach, all interaction between ABM and DTA including generating a list of vehicle trips by ABM for DTA and providing Level-of-Service (LOS) variables by DTA for ABM is implemented at the individual level without an aggregation bias. The paper is based on the model applications for the Atlanta Regional Commission (ARC) and Ohio State DOT (ODOT) sponsored by the FHWA C10 grants.
The major innovative features of suggested approach can be summarized as follows:
1. List of trips generated by ABM for DTA with the necessary level of details, continuous temporal resolution for departure time, and tour organization ready for the DTA simulation without any additional post-processing.
2. Level-of-Service (LOS) variables generated by DTA for ABM from the individual trajectories for already simulated trips that are used instead of traditional aggregate Origin-Destination skims; in particular; this is achieved by using innovative methods for processing, storage, and LOS search in the databank of individual trajectories.
3. Individual schedule consistency between travel times and activity durations that is fully preserved as an inherent feature; it is ensured by the innovative schedule adjustment procedures included in the model system equilibration.
The paper discusses details of the model structure, software set-up, equilibration strategies and convergence, as well as associated run times. Application experience includes an extensive model validation for the base year and sensitivity tests for future scenarios. The advantages of fully individual microsimulation structure are portrayed in the context of highway pricing scenarios. It si shown that when both ABM and DTA incorporate an individual distributed Value Of Time (VOT) in a consistent way the results become more reasonable compared to the simulation where either ABM or DTA operate with aggregate segment-specific VOT. The paper discusses further necessary steps to make an integrated ABM-DTA model a practice-ready tool.