Game Theoretic Modeling and Decision Making for Connected Vehicle Interactions at Urban Intersections

To ensure safe and efficient deployment in real world, autonomous vehicles (AVs) need to deal with complex interactions. This study deduced the rudiment of a meta decision-making model for connected vehicle interactions at urban intersections based on a game-theoretic framework. In this work, one of the key components is a newly proposed set of attributes, i.e. the Egoism, Aggressiveness and Rationality, abbreviated as the EAR. It has great potential to indicate how the interaction between two vehicle agents would progress further, which enables the multi-equilibria problem to be solved in a more efficient way. Besides, the Approximate-Equivalent-Trajectory method is utilized to ensure the generalization and computational efficiency of the model. Finally, the proposed method is validated using both simulations and real-world human driving dataset. The results and analysis demonstrate the feasibility and effectiveness of the proposed algorithms.