Nowadays, the automotive industry is in an era of great transformation. New technologies, new materials, new models, and so on are emerging endlessly, and traditional road testing methods for automobiles have long been phased out. Nowadays, car testing is generally conducted through road simulation experiments. Simulation has high flexibility and can simulate conditions that are difficult to achieve in reality, and road simulation experiments can quickly obtain results, greatly improving the efficiency of automotive testing. With the development of computer technology, the problems of automotive road simulation control systems are becoming increasingly prominent, including inconvenient data storage and sharing, difficulty in expansion, and insufficient number of signal analysis and processing channels. Nowadays, artificial intelligence technology is becoming more and more mature. Deep reinforcement learning can realize direct control from the original input to the output through end-to-end learning, and use cloud computing to realize the path. Theoretically, using deep reinforcement learning can effectively solve the problem of road simulation test control system. In theory, using Deep reinforcement learning can effectively solve the problem of road simulation test control system. Therefore, this paper studies the control system of road simulation experiment, and uses Deep reinforcement learning in artificial intelligence to improve the control system and control method of road simulation experiment through Deep reinforcement learning. The experiment proves that using Deep reinforcement learning to improve the road simulation test control system can effectively improve its control method and solve the problems exposed. While improving data collection accuracy by about 10%, it also reduces storage consumption by 50%.