基于遗传再励和博弈论的多Agent交通控制系统优化和协调

Optimized and coordinated multi-agent traffic control system based on genetic reinforcement and game theory

提出一种基于多Agent的交通信号控制方法.通过再励学习对交叉口的交通流进行优化,通过遗传算法产生局部学习过程的全局优化标准来优化信号灯周期和绿信比,最后用博弈论来进行区域协调.这种方法将局部优化和全局优化统一起来.研究结果表明,遗传算法优化结果优于爬山法,新的控制方法优于传统的控制方法.

The paper proposed a control method of the traffic signals based on multi-agents. The reinforcement learning focuses on the optimization of intersection＇s traffic flow. The genetic algorithm intends to introduce a global optimization criterion to each of the local learning processes that optimized the cycle of traffic signals and green ratio. Area-wide coordination was done by game theory. This method can combined local optimization with global optimization. Analysis result indicates that genetic algorithm is better than hill climbing method,and the new control method is better than the traditional control method.