A Multi-Vehicle Game-Theoretic Framework for Decision Making and Planning of Autonomous Vehicles in Mixed Traffic

To improve the safety, comfort, and efficiency of the intelligent transportation system, particularly in complex traffic environments where autonomous vehicles (AVs) and human-driven vehicles (HVs) coexist, a game theoretic trajectory planning framework is proposed in this article. Firstly, the game framework including non-cooperative games between AVs and HVs, as well as partial cooperative games between ego AV and other AVs is constructed. Secondly, a longitudinal game strategy for HVs is established with consideration of the driver's longitudinal handling characteristics and personalized aggressiveness, with which the driving behaviors of HVs can be accurately predicted. Thirdly, the longitudinal game strategies for AVs at different positions are designed, with consideration of the safety constraints of AVs and surrounding vehicles, as well as the objectives of comfort and traffic efficiency. Then, the interactive games of ego AV, HVs, and other AVs in different types of mixed-driving environments are solved based on the Stackelberg game and partial cooperative game methods. Finally, the optimal lane-changing (LC) gaps and longitudinal speed trajectories of AVs are obtained. Human-in-the-loop experiment results demonstrate the effectiveness of the proposed trajectory planning framework in lane-changing scenarios involving HVs with different aggressiveness and response delays.