Large-scale multimodal transport modelling. Part 2: Implementation and validation

Abstract

Recent developments in agent-based transport simulation provide promising results. However, the agent-based approach is frequently criticized for its apparent dependence on vast amounts of, mostly unattainable, data. In this second paper of a two-part series, we demonstrate the use of the Multi-Agent Transport Simulation (MATSim) with the multi-modal synthetic population generated in the first part of the series. The paper addresses the co-evolutionary approach with which agents attempt to maximize their individual utilities using both time and cost elements. The agent-based model is run for the Nelson Mandela Bay Metropolitan area using multiple modes that include commercial vehicles. We validate the simulation results against regional traffic counts, as well as other spatio-temporal anchors. The co-evolutionary nature of agents’ decision-making is shown as the simulation progresses, and we analyse how agents’ choices evolve over time as the simulation moves towards a steady-state. The paper ends with a few examples of how the disaggregate modeling can be used to answer policy and infrastructure decisions.