Multi-Agent Simulation Modeling and Optimization Strategies for Pedestrian–Vehicle Conflict Behavior at Intersections

Abstract. Amidst the increasing complexity of urban traffic networks and the frequent pedestrian-vehicle conflicts at intersections, this study addresses the challenges inherent in conventional methodologies, which often fail to accurately capture dynamic decision-making processes and are heavily reliant on empirical data. We propose an innovative simulation framework that synergistically integrates multi-agent behavioral modeling with dynamic non-cooperative game theory. This framework enables precise representation of pedestrian and vehicle behavioral strategies, conflict evolution, and spatial distribution patterns, complemented by conflict density mapping for enhanced visual risk assessment. Empirical results demonstrate the model’s capability to accurately identify high-risk zones and to substantially mitigate conflict frequency and severity through the modulation of vehicle yielding probabilities and the regulation of pedestrian behavior. By transcending the limitations of traditional static-rule-based approaches, the proposed framework offers a robust and theoretically grounded tool for traffic safety risk evaluation, thereby furnishing valuable insights for intersection safety management and the design of intelligent transportation systems.