Dynamic Gaming Lane-Changing Decision-Making for Intelligent Vehicles Considering Humanlike Driving Preferences

Regarding the traditional lane-changing decision theory around the vehicle’s intention to change lanes, less consideration of the behavioral interactions between vehicles, and the personalized driving preferences of different drivers, this paper proposes a dynamic game for a lane-changing decision-making method that considers human-like driving preferences. First, to match the multiperformance evaluation requirements in the lane-changing process of intelligent vehicles, a cost function including driving space, traffic efficiency, and driving comfort is constructed. Second, the analytic hierarchy process (AHP) and criteria importance through intercriteria correlation (CRITIC) methods are used to conduct subjective and objective analyses on the next generation simulation (NGSIM) traffic data set to obtain the weight coefficients of multiple performance indicators of human-like driving preferences. The effects of different driving behaviors on lane-changing intentions and performance indexes are also studied. Finally, fuzzy control theory and intelligent driver model (IDM) are used to predict the driving behavior of interacting vehicles in the target lane, and the master-slave dynamic game theory and the particle swarm optimization algorithm are used to realize the behavioral interaction between the main vehicle and surrounding vehicles and to make the optimal lane-changing decisions. The research results show that the dynamic game lane-changing decision-making method of intelligent vehicles as proposed in this paper, which considers human-like driving preferences, can effectively meet the personalized requirements of different driving behaviors on driving space and traffic efficiency in the process of lane changing and improve the safety of intelligent vehicle lane-changing driving.