Decentralized adaptive agents for control of traffic signals

摘要

In this paper we discuss the design and implementation details of DAARTS, a novel agent-based controller for decentralized adaptive control of traffic signals in an urban traffic network. DAARTS adopts a hierarchical multi agent-based architecture where the lowest level (intersection agents) involves individual intersection-traffic dynamics and phase selection based on "local" information, while higher levels take into account the supervisory (network-level) dynamics. Coordination between intersections is achieved in a decentralized manner at the lowest level. The decentralized control strategy is implemented in Cybele, a multi agent infrastructure that allows for distributed computing and inter-agent communication, hence allowing us to emulate real traffic controllers at individual intersections of a network. DSTK, the DAARTS Simulation Toolkit, allows for easy integration of our agent controller with CORSIM for simulation, visualization and evaluation of the controller.