With the advancement of urbanization, urban traffic becomes more and more congested. As an important node of traffic flow convergence and diversion, the intersection efficiency will affect the operation efficiency of the urban transportation system to a large extent. The existing intersection signal control mostly uses the empirical fixed signal timing, and the empirical fixed signal timing seriously affects the traffic efficiency of the intersection. This paper starts with the signal control method of a single intersection in the city, mainly improves the operation efficiency of the intersection by optimizing the signal control method of the intersection; this paper proposes the intersection signal control method based on reinforcement learning, and adopts the Q-learning reward and punishment signal design method, implementing optimization of intersection signal control, and finally, the simulation experiment of the intersection is carried out by PTV- Vissim9.0 software to verify the feasibility of the theory. Keywords: Q-learning; Traffic signal control; Reinforcement learning; Simulation

CICTP