A deep reinforcement learning-based approach for autonomous lane-changing velocity control in mixed flow of vehicle group level

As an important driving behavior, lane-changing has a great impact on the safety and efficiency of traffic flow interacting with surrounding vehicles, especially in mixed traffic flows with autonomous vehicles and human-driven vehicles. This study proposes a deep reinforcement learning-based lane-changing model to train autonomous vehicles to complete lane-changing in the interaction with different human driving behaviors. First, a mixed-flow lane-changing environment of vehicle group level is constructed with surrounding vehicle trajectories extracted from natural driving trajectories. Then, the state space and action space are determined, the reward function is designed to comprehensively consider safety and efficiency, so as to guide autonomous vehicles not to collide, and determine the acceleration and direction angle to complete lane-changing behavior, and a collision avoidance strategy is integrated into the proposed method to ensure the safety of longitudinal motion. Furthermore, the trained model can learn the experience of successful lane-changing, resulting in a 90% success rate without collision in testing. Finally, the driving performance of the proposed method is analyzed in terms of safety and efficiency evaluation indicators, which proves that the proposed method can improve the efficiency and safety of the lane-changing process.