Collaborative Decision-Making Algorithm for Autonomous Driving Based on Evolutionary Game Theory

Decision-making of autonomous driving is challenging due to high-density traffic and complex road environment. Based on vehicle to vehicle (V2V) communication, collaborative decision-making of multi-vehicles may enhance efficiency and safety. However, traditional game-theoretic collaborative decision-making approach assumes perfect rationality and therefore not capable to model behavioral uncertainty and irrationality. This study developed an evolutionary game theory approach for collaborative decisionmaking of autonomous driving. Vehicles are divided into multiple pairs of two vehicles for evolutionary game modeling, and a linear weighted income function composed of safety, efficiency, and comfort is designed. The test results of <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\mathbf{4, 2 8 8}$</tex> complex traffic scenarios at unsignalized intersections in the Lyft open dataset show that the proposed model based on the evolutionary game achieved higher accuracy compared with traditional game theory models. The model successfully captures the dynamic evolution characteristics of agent interaction under different game intensities.