Multi-agent dynamic traffic assignment incorporating GPS diary and personalized supernetwork approach



Last Updated: 2-2016


This thesis aims at incorporating the integrated space-time accessibility concept from supernetwork structure into the framework of simulation-based dynamic traffic assignment (DTA), trying to explore its effectiveness.
A methodology is proposed first, using multi-modal simulation to reproduce both motorized and non-motorized trips of real-life individual, together with a one-year GPS diary data as source for travel demand input, enhancing the resolution of detail this demand can capture. The simulation is then conducted using a multi-agent approach (MATsim), maximizing the benefits of representing the heterogeneous choice preferences of individuals in multidimentional activity and travel programs. Based on the estimates of a mixed logit model, the proposed approach is operationalized via modifying the utility functions for evaluation of individual’s plan in response to the integrated accessibility concept. The supernetwork structure is also partially reproduced with a multi-layer network consisting of private vehicle network and public transit network.
Following the preparations above, the approach is eventually implemented by taking Eindhoven region as a study area. Results suggest that this integrated approach is a promising solution for transportation planning, and possess high applicability in other scenarios as no major adjustment is required. Future works include parking activity simulation and household joint-decision simulation, and incorporation with population synthesis for large agent population.