Large-scale multimodal transport modelling. Part 1: Demand generation

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 paper we show that the required data is mostly available, even for a country like South Africa. This paper addresses demand generation, the first step in a two-part series, and focuses on three components. Firstly, we demonstrate how a synthetic population of private individual agents is generated using existing and available data. Using iterative proportional fitting (IPF), the population is generated for the Nelson Mandela Bay Metropolitan, and includes 24-hour activity chains for both primary activities such as home, work and education, and also secondary activities: shopping, leisure and other. Secondly, a commercial vehicle population is generated, a novel contribution in an agent-based setting. The commercial vehicles include both intra- and inter-provincial vehicles with different activity chain characteristics. Lastly, the paper demonstrates how an accurate road network is extracted for the Multi Agent Transport Simulation (MATSim) using open data.