城市交通干线的Q-学习控制算法

针对城市交通干线协调控制的要求,提出了利用Q-学习控制算法和模糊算法的分层递阶控制的方法.采用两层结构,第1层为控制层,针对单个路口,对下一个时间段内路口各个方向的相位饱和度进行预测,并在此基础上计算出下一个时间段内各个路口的周期、各个方向上的绿信比;第2层是协调层,采用Q-学习控制算法对干线各个路口间的相位差进行调整.采用TSIS交通分析软件对由5个路口组成的交通干线进行仿真,Q-学习控制算法与定时控制和遗传算法进行比较,结果表明:Q-学习控制算法具有明显的优越性.

Q-Learning Control Algorithm for Urban Trunk Road Control

Q-learning control algorithm and the fuzzy algorithm are used to solve the problem of urban trunk road coordination control.The first layer of the network is a manipulative layer that forecasts the traffic flow saturation degree of a phase during the next period,and based on that,the single cycle and split of each direction are calculated.The second layer is a coordinated layer.Q-learning control algorithm is used to correct the phase difference of intersections.This paper adopts the TSIS microscopic traffic analysis software to implement the simulation of traffic trunk roads which consist of 5 intersections.A comparison of fixed-time control with genetic algorithm indicates that Q-learning control algorithm has an obvious advantage.