Multiagent model of VICS

VICS is becoming well-known as one of the representative applications of ITS, and this can be regarded as an interesting test-bed for the analysis of complexity emerging in a multiagent system. That is, differences of individual drivers' attitudes to the provided information will make affairs more complex than that of being treated by usual multiagent systems where the individual agents' behaviors could be predicted in some deterministic ways. Different from the conventional approaches to design a whole group of agents as a single system, we do not assume that any agents can make any peer-to-peer communication with others, but that they can pick up their own roles from common messages broadcasted over a group of agents. Based on our proposed bi-referential model explaining how agents validate their own role or their niche by referring to both self and others through the self-closed learning mechanism, the emergent properties caused by the change of individuality of the drivers as well as by the environmental status of traffic jams (i.e., their frequencies and geographical scales) are discussed.