A Collaborative Scheduling Lane Changing Model for Intelligent Connected Vehicles Based on Deep Reinforcement Learning

Zheyu Cui,Jianming Hu

DOI: https://doi.org/10.1061/9780784483053.178

2020-12-09

Abstract:The traditional lane-changing trajectory planning model divides road conditions into obstacle-changing trajectory design and barrier-free lane-changing trajectory design. Due to the constraints of dynamic obstacles, the obstacle-lane-changing trajectory design will introduce a larger amount of calculation than the barrier-free lane-changing trajectory design. This increases the delay and reduces safety. In order to overcome these shortcomings, this paper proposes a cooperative scheduling lane changing model for intelligent connected vehicle based on deep reinforcement learning. The collaborative scheduling algorithm acts on the trajectory planning layer to unify all lane-changing trajectory designs into barrier-free trajectory designs to reduce the amount of computation. The lane change decision maker based on deep reinforcement learning determines the optimal moment of the cooperative scheduling algorithm. Finally, the paper verifies that the proposed model has a significant improvement in traffic efficiency by comparing with the traditional lane changing model under different traffic density.