结合Q学习和模糊逻辑的单路口交通信号自学习控制方法*

针对城市交通系统的动态性和不确定性,提出了基于强化学习的信号交叉口智能控制系统结构，对单交叉口动态实时控制进行了研究。将BP神经网络与Q学习算法相结合实现了路口的在线学习。同时，针对交通信号控制的多目标评价特征，采用基于模糊逻辑的Q学习奖惩信号设计方法，实施对交通信号的优化控制。最后，在三种交通场景下，应用Paramics微观交通仿真软件对典型十字路口进行仿真实验。结果表明，该方法对不同交通场景下的突变仍可保持较高的控制效率，控制效果明显优于定时控制。

Self-learning traffic signal control method of isolated intersection combining Q-learning and fuzzy logic

To address the dynamics and uncertainty in unban transportation system, this paper proposed a traffic signal control system based on reinforcement learning, which was suitable for real-time control in isolated intersection. The proposed method was capable of online learning through a combination of BP neural network and Q-learning algorithm. Furthermore, due to the multi-objective property in traffic signal control, this paper developed a reward design method for Q-learning based on fuzzy logic. Conducted simulated experiments in three traffic scenarios, using the Paramics microscopic traffic simulation software. Experimental results show that the proposed method has high control efficiency in different traffic scenarios, and is significantly better than fixed timing control method.