Cognitive Hierarchy Game-Based Method for Multi-Agent Lane Change Intention Recognition

In intelligent driving systems, the accurate prediction for lane-changing intentions of traffic participants is crucial for enhancing driving safety and traffic efficiency, especially in complex traffic environments. In this paper, we propose a novel cognitive hierarchy game-based lane-changing intention recognition (CHG-LCIR) method to dynamically predict the vehicle intentions, which innovatively integrates game theory with probabilistic inference. A hierarchical game model is first developed to describe the dynamic coupling interactions among vehicles. Then, the probabilistic inference approach is incorporated to estimate potential cognitive states, significantly improving the adaptability of the model to dynamic scenarios. Experimental validation based on the NGSIM dataset demonstrates that the proposed CHG-LCIR effectively achieves a good balance between model interpretability and intention performance.