基于Q学习的交通信号自学习控制方法的研究

为了减少车辆通过交叉口的平均延误时间,将Q学习与模糊推理相结合对基于智能体的单交叉口进行信号配时优化,以适应动态变化的交通流。在模糊控制规则集的基础上,通过遗传算法优化模糊推理中的隶属度函数参数,克服传统隶属度函数设计的主观性和盲目性。在此基础上,通过Q学习算法对其在线学习,以实现单交叉口交通信号控制智能体的自学习能力。仿真表明,该方法相比于传统的定时控制与模糊控制,具有较好的控制效果。

Study on self-learning traffic signal control method based on Q learning algorithms

In order to reduce the average delay time of vehicles passing intersection, to optimize the signal timing of agent controlled intersection by Q learning method and fuzzy reasoning to adapt dynamic variable traffic flow. On the basis of fuzzy rule set for signal control, to improve the effect of signal control and self - learning of signal control agent in an single intersection through Q learning, which is based on optimizing fuzzy control membership function＇s parameters with genetic algorithms and avoiding the sub- jectivity and blindness of designing the traditional ones. The result of simulation illustrates that the signal control method based on Q learning is better than fixed - time control and fuzzy control.