A dynamic packet-based multi-agent approach for large scale multimodal network simulation

The object of the paper is to introduce a multiagent model for very large networks. The prominent feature of this model is that agents are regrouped in packets with identical characteristics and dynamics. Specific agents describe the transportation system and eventually the user information system. The dynamics of packets are recaptured either by Lagrangian discretization (vehicular traffic) or by multiagent simulated public transportation dynamics. The computation of an approximate user equilibrium is achieved by the crossentropy technique. A simulation-based approach for large multimodal networks is developed.