Autonomous vehicles in mixed-autonomy traffic: game theoretic human-like decision making countermeasures

Human-like decision-making plays a pivotal role in enhancing the human-likeness of autonomous vehicles and ensuring seamless blending into human-driven vehicles-dominated traffic. Due to the ability to capture the interaction between drivers, it has great potential to apply game theory in the development of human-like decision-making algorithms. However, there are few reviews that systematically focused on the human-like decision-making strategy based on game theory. To this end, this paper is targeted to present a comprehensive and up-to-date summary of game theoretic human-like decision-making methods for autonomous vehicles in mixed-autonomy traffic by reviewing cutting-edge research conducted for various scenarios. The questions discussed in this article include: (1) What are the implications of social interactions for human-like decision-making development; and (2) How to establish the human-like decision-making algorithm with game theory, satisfying personalized requirements and coping with the uncertainty and randomness of complex traffic environment. To provide sound answers, the pivotal factors influencing decision performance are concluded based on the existing social interaction research and human-like decision methods. Through a comprehensive analysis, the development framework of the human-like game theoretic algorithm is proposed. Finally, the critical academic issues are concluded for indicating the future research directions.