Research on Driving Decision of Smart Vehicles Based on Reinforcement Learning

In order to ensure the smooth, reliable and safe driving process of intelligent driving cars, this paper analyzes the environmental information from the decision-making perception module of automatic driving and controls the car's own behavior to achieve the driving goal. The driving decision algorithm of intelligent vehicle based on reinforcement learning is given to ensure the safe driving of intelligent vehicle under complex constraints such as high speed, slip and roll. Through reinforcement learning, an iterative process of constantly interacting with the environment, getting rewards, updating strategies and then continuing to interact with the environment is given, and exploratory work is carried out by using reinforcement learning in TORCS simulator. Finally, the automatic driving system is built in simulink, and PreScan is used as the simulation environment for training and verification, which verifies the intelligent decision-making and control method of vehicles using reinforcement learning in the intelligent networked environment, and realizes the smooth, reliable and safe driving of intelligent vehicles.