Decision Making for Overtaking of Unmanned Vehicle Based on Deep Q-learning

Overtaking decision under dynamic environment is one of the key research directions. Traditional methods are difficult to solve such a complex optimization decision-making problem. In recent years, reinforcement learning algorithm is developing continuously, which can be applied to solve the overtaking decision problem of unmanned vehicles. Reinforcement learning generates new data through the continuous interaction between agent and environment, and feeds back to the agent for corresponding learning behavior, so as to improve its own behavior strategy. In this paper, Deep Q-learning (DQN) is used to solve the overtaking decision-making problem on one-way through two lanes. A one-way through two lane traffic environment is built in Python to train and test the algorithm.