Research on Decision Model of Autonomous Vehicle Based on Deep Reinforcement Learning

The Deep Q Network (DQN) model has been widely used in autonomous vehicle lane change decision in highway scenes, but the traditional DQN has the problems of overestimation and slow convergence speed. Aiming at these problems, an autonomous vehicle lane changing decision model based on the improved DQN is proposed. First, the obtained state values are input into two neural networks with the same structure and different parameter update frequencies to reduce the correlation between empirical samples, and then the hybrid strategy based on e-greedy and Boltzmann is used to make the vehicles explore the environment. Finally, the model is trained and tested in the experimental scene built by the NGSIM dataset. The experimental results show that the Double Deep Q Network (DDQN) model based on the hybrid strategy improves the success rate of the autonomous vehicle's lane-changing decision and the convergence speed of the network.