Infrastructure-Assisted Cooperative Decision Model With Priority Awareness at Unsignalized Intersections

Abstract Unsignalized intersections pose a significant challenge for multi-vehicle cooperative decision-making, where safety and efficiency must be ensured simultaneously under dynamic traffic conditions. To address this challenge, this study proposes a Vehicle-to-Infrastructure Integrated Distributed Agent Decision-Making (V2I-IDADM) framework, which leverages vehicle–infrastructure cooperation to enhance both coordination performance and safety at unsignalized intersections. Built upon a learning paradigm with centralized training and decentralized execution, the proposed V2I-IDADM framework utilizes the global perception of roadside infrastructure to assign passing priorities to connected and autonomous vehicles (CAVs), ensuring safe passage through intersections. Meanwhile, the framework’s global coordination mechanism constructs a unified and structured representation of intersection-level traffic states, enabling scalable, real-time decision-making. Specifically, a priority-based safety decision model is developed by jointly integrating passing-priority constraints with action optimization to promote multi-vehicle cooperation. To enhance training efficiency, a hierarchical weighted sampling strategy is introduced to emphasize high-value episodic experiences and accelerate iterative self-learning. Extensive experiments in pure CAV and mixed-traffic scenarios demonstrate that the proposed framework achieves superior safety and efficiency compared with state-of-the-art methods. Experiments conducted on both a miniature intelligent vehicle platform and a full-scale vehicle further validate the practical feasibility and deployment potential of the proposed framework.