Q-learning based multi-intersection traffic signal control model

In multi-intersection urban traffic environment, conventional fixed-time traffic signal control methods expose low performance when face with complex and stochastic traffic conditions which caused by the interaction among multiple intersections. A Q-learning based traffic signal control model is proposed to deal with time-varying and stochastic traffic flow problem, which takes advantage of the specialty of autonomous learning inherent in Q-learning. The capacity of discovering autonomously optimal control policy corresponding to varying traffic conditions and no fixed mathematic control model is needed are the major advantages of this method. The experiment results in simulation environment also demonstrate this method is applicable and effective.