Adaptive decentralized and collaborative control of traffic lights

Abstract

We present an approach to design an improved control system for arbitrary urban traffic networks based on Organic Computing concepts. For control of these networks a decentralized structure consisting of small computers at every node of the net is proposed. Each node is viewed as an autonomous agent with limited sensory horizon. It tries to adapt itself such that the local traffic throughput is maximized. Additionally communication takes place between adjacent nodes which leads to an iterative global optimisation through collaboration of the agents. The paper discusses the approach in detail and introduces the controller architecture.