A Potential Game-Based Framework for Cooperative Driving at Mixed Un-Signalized Intersections

With the continuous development of autonomous driving technology, future transportation will be in a long-term state of mixed traffic with both Connected and Autonomous Vehicles (CAVs) and Human-driven Vehicles (HDVs). Due to the uncertainties and uncontrollability of human driver behavior, achieving safe decision-making for CAVs in mixed traffic environments poses challenges, especially in potential conflict areas such as un-signalized intersections. Cooperative driving holds the potential to address the above issues. Considering the increased hazards posed by heterogeneous HDVs in real-world traffic, an Adaptive Weight Shapley Weighted Potential Game (AWSW-PG) is proposed to establish a cooperative driving framework for CAVs. Heterogeneous HDVs are modeled by a non-cooperative Bayesian game and utilized as background traffic. Then, the potential game that connects the individual reward and cluster reward generates the optimal cooperative solution by searching the Nash Equilibrium of the problem. Furthermore, the Shapley value is introduced to quantify the unsymmetrical impact of each vehicle. Finally, given the uncontrollable and unpredictable nature of HDVs, an adaptive weight method is formalized to adjust the estimation on HDVs dynamically. To evaluate the effectiveness and robustness of the proposed cooperative driving framework, three cases are conducted. Results demonstrate that the AWSW-PG framework exhibits good performance in terms of efficiency and safety under different rates of penetration. In addition, the quantification of vehicle impacts and dynamic adjustment of HDVs estimates have been proven capable of guaranteeing stability and efficiency during cooperation.