The first agent steps in agent-based transport planning

Abstract

Paper presented at the 28th Annual Southern African Transport Conference 6 - 9 July 2009 "Sustainable Transport", CSIR International Convention Centre, Pretoria, South Africa.

South Africa's socio-political history and rapidly changing urban dynamics, coupled with an eclectic mix of first and third world transport infrastructure presents a moving target to even the best traditional transport planning methodologies. A pressing need has emerged for a planning technology that can capture these dynamics and explore its evolution over time. In this regard, agent-based microsimulation has become a viable tool for analysing and predicting transport system performance, and promises to address the current and future needs of transport planners in a variety of scenarios. Our initial implementation of Multi-Agent Transport Simulation Toolkit (MATSim), a large-scale agent-based microsimulation framework proposed by Balmer et. al. (2006), models the passenger vehicle traffic of Gauteng for 2001. This paper describes the basic principles of operation of MATSim, and identifies data sources relevant to current and future states of the South African implementation.