基于再励学习和遗传算法的交通信号自组织控制

Self-organized control of traffic signals based on reinforcement learning and genetic algorithm

提出一种将再励学习与遗传算法相结合的遗传再励学习方法对交通信号进行自组织控制。再励学习是针对每一个道路交叉口交通流的优化，修正每个信号灯周期的绿性比；而遗传算法产生局部学习过程的全局优化标准，即是修正信号灯周期的大小。这种方法克服了现有的控制方法需要大量数据传输通讯、准确的交通模型等缺陷，将局部优化和全局优化统一起来。通过计算机仿真实验表明了方法的有效性。

A combination algorithm of reinforcement learning and genetic algorithm is proposed in this paper to self--organized control of the traffic signals. Reinforcement learning focuses on the optimization of an intersection traffic flow which modified the split of traffic signal cycle, while the genetic algorithm are intended to introduce a global optimization criterion to each of the local learning processes which modified the cycle itself of traffic signals. This method overcome the drawbacks in existing control methods such as huge data transfer and communication, accurately traffic model and so on, and unified the local optimization and global optimization. Through the computer simulation, the effectiveness of method is demonstated.