Dynamic Priority-Driven Multi-Agent Path Planning for Warehouse Logistics: Spatiotemporal Constraint Optimization

To address challenges in multi-agent path planning (MAPF) for high-density warehouse logistics scenarios-including frequent spatiotemporal conflicts, narrow-passage deadlocks, and inefficiencies in large-scale cluster scheduling - this paper proposes a hierarchical cooperative planning architecture. The architecture integrates upper-layer dynamic priority scheduling with spatiotemporal constraint propagation and lower-layer safe interval path planning (SIPP) guided by reverse resumable A* (RRA*) global heuristic search, achieving conflict preemption and path optimization. Core algorithmic innovations include: (1) A sliding temporal window-based dynamic priority adjustment strategy to alleviate long-term blocking of low-priority agents; (2) Narrow-passage conflict prediction and alternative path pre-planning mechanisms to reduce deadlock risks; (3) RRA* historical state reuse and spatiotemporal constraint recursion to compress redundant search spaces. Experiments demonstrate that in high-density warehouse maps, the algorithm achieves collision-free planning success rates in multi-agent scenarios, with significant makespan reduction and improved computational efficiency compared to existing MAPF methods. Despite local path suboptimality caused by temporal window discretization, its advantages in continuous task support, dynamic environment adaptability, and balanced computational efficiency provide a reliable solution for large-scale cluster scheduling in logistics automation.