Answers are grounded in this paper and cite their sources.
Lane-change decisions in complex traffic environments involve strategic interactions with surrounding vehicles. This paper proposes a dynamic game-theoretic lane-change decision model that integrates lane-change intention recognition, game initiation and termination conditions, and a multi-objective payoff function incorporating safety, efficiency, and comfort. Expert trajectories representing both succeed and failed lane-change attempts are extracted from the NGSIM dataset to estimate the initial payoff weights through maximum entropy inverse reinforcement learning. A Bayesian update subsequently adjusts these weights according to real-time traffic states, adapting the model’s decision preferences to changing conditions. MATLAB case studies illustrate the model’s behavior under different initial interaction conditions. Under the same favorable conditions, the proposed model completes the lane change within five game rounds, whereas the defensive game model aborts the maneuver. Under unfavorable conditions, the proposed model identifies the maneuver as infeasible and abandons the attempt. Under an AV penetration rate of 50%, SUMO simulations are conducted to further evaluate the impact of the proposed model on road traffic capacity. Compared with the MOBIL model, the proposed model achieves an approximately 33% increase in average vehicle throughput, while improving the average driving speed by at least approximately 10 km/h.
No takes yet. Share an insight, caveat, or question.
Zhang et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: