With the rapid development of computer technology and artificial intelligence, unmanned vehicles have become a new hot spot. In this paper, a verification model of the automatic car is proposed to simulate the unmanned vehicle, and the deep Q network (DQN) algorithm is used to control the automatic car. The algorithm uses reinforcement learning and neural network technology, in the case of less prior knowledge, it can train the neural network according to the obtained sensor information ,then make the right decision to achieve the control of the vehicle and the effect of avoiding obstacles. In addition, this paper verifies the control effect of DQN algorithm on automatic trolley by experimenting in simulated environment. Experimental results show that, after a certain period of training, DQN algorithm can effectively control the automatic car.