Cooperative passage and active gap adaptation for CAVs in mixed unsignalized intersections

The management of mixed traffic at unsignalized intersections, where connected and automated vehicles (CAVs) interact with human-driven vehicles (HDVs) and pedestrians, presents a significant challenge for intelligent transportation systems. Existing centralized motion planning methods often adopt conservative yielding strategies, ensuring safety but causing frequent stops and efficiency loss. To address this, we propose a novel cooperative management framework that integrates centralized CAV planning with interactive models for HDVs and pedestrians. It comprises three components: a mixed integer linear programming (MILP) motion planner, a game-theoretic CAV-HDV model, and an active gap adaptation model enabling CAVs to proactively create pedestrian crossing opportunities. Extensive simulations demonstrate that the proposed framework consistently outperforms both the conservative baseline and the MARL benchmark in terms of efficiency measures such as delay and queue length across varying CAV penetration rates and traffic demand levels, validating its substantial improvements in efficiency and safety for mixed intersections.