Academic
Publications
Traffic Light Control by Multiagent Reinforcement Learning Systems

Traffic Light Control by Multiagent Reinforcement Learning Systems,10.1007/978-3-642-11688-9_18,Bram Bakker,Shimon Whiteson,Leon Kester,Frans C. A. Gr

Traffic Light Control by Multiagent Reinforcement Learning Systems   (Citations: 1)
BibTex | RIS | RefWorks Download
Traffic light control is one of the main means of controlling road traffic. Improving traffic control is important because it can lead to higher traffic throughput and reduced congestion. This chapter describes multiagent reinforcement learning techniques for automatic optimization of traffic light controllers. Such techniques are attractive because they can automatically discover efficient control strategies for complex tasks, such as traffic control, for which it is hard or impossible to compute optimal solutions directly and hard to develop hand-coded solutions. First the gen- eral multi-agent reinforcement learning framework is described that is used to con- trol traffic lights in this work. In this framework, multiple local controllers (agents) are each responsible for the optimization of traffic lights around a single traffic junc- tion, making use of locally perceived traffic state information (sensed cars on the road), a learned probabilistic model of car behavior, and a learned value function which indicates how traffic light decisions affect long-term utility, in terms of the average waiting time of cars. Next, three extensions are described which improve upon the basic framework in various ways: agents (traffic junction controllers) tak- ing into account congestion information from neighboring agents; handling partial observability of traffic states; and coordinating the behavior of multiple agents by coordination graphs and the max-plus algorithm.
Published in 2010.
Cumulative Annual
View Publication
The following links allow you to view full publications. These links are maintained by other sources not affiliated with Microsoft Academic Search.
    • ... control is fundamentally a problem of sequential decision making, and at the same time is a task that is too complex for straightforward computation of optimal solutions or effective hand-coded solutions, it is perhaps best suited to the framework ofMarkov Decision Processes (MDPs) and reinforcement learning (RL) or approximate dynamic programming (ADP), in which an agent learns from trial and error via interaction with its environment[5]...

    Chun-gui Liet al. Multi-intersections traffic signal intelligent control using collabora...

Sort by: