This paper explained the most important aspects of a multi-agent simulation approach to transportation and mobility planning. These aspects include a simulation of the physical system, modules which compute strategic/tactical decisions of the agents, and a learning framework which models how agents adapt to the system and to each other. It was then shown how far we are with a real-world implementation of such a system. Since such real-world systems typically are very large (several millions of agents), computing aspects were discussed, indicating that it is possible to simulate such systems in acceptable amounts of time.
As pointed out in the introduction, it is believed that such a microscopic approach to mobility simulation offers important methodological advantages over the more traditional aggregated approaches. In particular, it is possible to model each traveler's decision individually, while taking into account specific attributes such as income, or distance to public transit. The fact that this is now feasible makes it necessary to better understand the individual decision processes which generate the dynamics of the transportation system as a whole. This is an important research task for the coming years.