Agile Multi-Agent Model Architecture for Intelligent Intersection Traffic Simulation

Abstract

Urban traffic can naturally be modeled as a distributed and heterogeneous multi-agent dynamic system. To allow a scalable solution for simulation, control and information management, a simple but powerful generic model architecture is proposed in this work, supporting various different traffic participant types. It is shown that even though the agent actions can be restricted to a fairly small set of basic operations, a remarkably rich global system functionality and complex information flow can be modeled. The resulting traffic dynamics and information model is inherently flexible and agile with respect to participating agents and their evolution of system dynamics while enabling efficient and parallelizable simulation, control, and prediction computations. Selected use cases for traffic monitoring, management, and control are outlined, and the model capabilities are demonstrated.