Modeling of lane-changing behaviors of human-machine co-driving vehicles based on evolutionary game

To alleviate the conflicts resulting from lane-changing in human-machine co-driving, this study investigates four game scenarios: human-driven lane-changing vehicles (HLVs) and human-driven target vehicles (HTVs), HLVs and Autonomous target vehicles (ATVs), autonomous lane-changing vehicles (ALVs) and HTVs, and ALVs and ATVs. An evolutionary game model is formulated by integrating safety, efficiency, comfort, fatigue, and energy-saving utilities. The strategy selection process of decision-makers is simulated by replicator dynamics equations. Based on the Friedman method, the stability of equilibrium points in different situations is analyzed, and the sensitivity of evolutionary results to utility parameters is clarified. Results show that the strategy profile of (lane-changing, not-giving-way) that causes conflicts only exists temporarily under certain conditions and will not become a long-term stable point. The game characteristics under different scenarios are obviously different; that is, the game evolutionary paths for autonomous vehicles are the most stable, while the adjustment of game strategies for human-driven vehicles is more frequent. Additionally, energy-saving promotes the formation of the strategy profile of (lane-changing, giving-way), while fatigue limits the frequent lane-changing of human-driven vehicles. The evolutionary game model proposed can provide theoretical support for improving the performance of the human-machine co-driving system.