Stackelberg Game Based on Trajectory Prediction for Lane Change in Mixed Traffic

To address lane-changing conflicts between intelligent and human-driven vehicles in mixed traffic environments, this study proposes a Stackelberg game-based decision-making method for autonomous vehicles. A Stackelberg game framework is established between autonomous vehicles (leaders) and target-lane human-driven vehicles (followers) in three typical scenarios. The method develops a utility function for human-driven vehicles incorporating driving styles and safety-comfort-efficiency factors, with a corresponding cost function for autonomous vehicles. An improved Stackelberg game model integrates trajectory prediction of human-driven vehicles, while a bi-level optimization algorithm combining model predictive control and genetic algorithms jointly optimizes acceleration sequences and lane-change timing. Simulations demonstrate 60% reduction in heading angle variation and 67.59% decrease in yaw rate compared to non-game strategies, confirming enhanced safety, comfort, and efficiency of the proposed method.