A Personalized Lane-Changing Decision System Based on Improved Stackelberg Game and Traffic Flow Information

A lane-changing decision system based on improved Stackelberg Game and traffic flow information is proposed. The system consists of three parts: Lane-Changing Demand Assessment, Lane-Changing Condition Assessment and Multi-Lane Game Model. In Lane-Changing Demand Assessment, a lane-changing demand function considering the urgency and potential of lane change is proposed and traffic flow information is used as input to make assessment more forward-looking. Lane-Changing Condition Assessment is developed to assess the feasibility of lane change. Existing game models determine players before the game starts, but in multi-lane scenarios, they cannot determine in advance who will participate in the game. Inclusion of unrelated players will cause the game to collapse. To solve the problem of multi-lane game, we propose a new mechanism to improve the Stackelberg Game, and develop Multi-Lane Game Model based on the improved Stackelberg Game. The mechanism changes the classical game structure and determines actual players halfway through the game, which expands the application scope of game theory in the field of lane change. Additionally, a classifier is designed to incorporate driving style into Multi-Lane Game Model for enhancing its personalization. To verify the performance of the lane-changing decision system, three experiments were carried out. The results indicate that by considering traffic flow information, the assessment of lane-changing demand becomes more forward-looking, thereby avoiding unnecessary lane changes. By comparing with human drivers, it’s found that the decision-making of our system is more rational and reliable, mitigating the risks caused by the irrational decision-making of human drivers.