A decentralized, evolutionary, hybrid controller for directing traffic along two-way streets

This paper tackles the excessive complexity facing the design of traffic controllers. This complexity, which is usually exponential in the number of agents, effectively limits planning to that of a small-size traffic. Here, an evolutionary approach to control design is suggested to free traffic controllers from such a limitation. Guidelines from self-organizing systems are used to derive a multi-agent motion controller that, among other things, has a computational effort that linearly increase with the number of agents; therefore, making it feasible to manage a large-size traffic. The controller is also required to be "open" in order for any agent to join or leave traffic without the others having to adjust the manner by which they sense, process information, or actuate motion. Along with self-organization, concepts from hybrid systems, and the potential field approach to control synthesis are utilized to realize the controller. Theoretical development as well as simulation results are provided.