Affordance-based Agent model for road traffic simulation

Existing traffic simulations often consider normative driver behavior. Drivers do not always use physically delineated lanes: sometimes drivers use the entire road surface. Thus, current traffic simulations do not reproduce all observed urban and suburban traffic phenomena. To improve the validity of urban and suburban traffic simulations, we propose to consider driving context and driver behavior in terms of occupied space. We endow driver agents with an ego-centered representation of the environment based on the concept of affordances and virtual lanes. Affordances thus identify the possible space occupation actions afforded by the environment and by other agents. The proposed model was implemented using our ArchiSim tool. We show that this model is more efficient and realistic than existing models. The experiments also reproduce real traffic situations and compare simulated data to real data.