A study on two-layer spatiotemporal corridor planning based on SVO

Abstract

High-density urban Unmanned Aerial Vehicle (UAV) logistics demand a sophisticated balance between limited airspace resources and dynamic multi-path coupling. This paper proposes a hierarchical spatiotemporal planning framework grounded in a “capacity-driven trajectory decoupling” mechanism leveraging Sparse Voxel Octrees (SVO). By transforming unstructured 3D space into a multi-resolution topological network, we quantify the traffic-bearing limits of heterogeneous voxels through a physical scale-based capacity mapping. The framework synergizes an upper-layer capacity-aware A* algorithm with an improved lower-layer Windowed Hierarchical Cooperative A* (WHCA*) algorithm, effectively mitigating localized congestion and ensuring deterministic intra-node conflict resolution. Experimental results in complex urban scenarios demonstrate that the SVO-based model achieves a 10.4x spatial compression ratio and a 98.5% mission success rate, significantly outperforming traditional Conflict-Based Search (CBS) and WHCA* baselines in scalability and global throughput.