End-to-end two-layer planning and optimization method for UAV swarm spider web-inspired coverage search

To address the challenges of balanced optimization between resource allocation and search execution in unmanned aerial vehicle (UAV) swarm collaborative coverage search, and to improve the balance, flexibility, and response speed of regional coverage while reducing resource consumption, an end-to-end two-layer planning and optimization method for UAV swarm spider web-inspired coverage search was proposed. The first layer focused on UAV resource allocation optimization for multi-target regions. A multi-target balanced UAV resource allocation optimization model was constructed, and a deep learning-based end-to-end network was built. The search region features and UAV parameter encoding served as the input matrix, and the optimal collaborative quantity scheme of UAV swarms for multiple regions was directly output, to ensure the coverage task requirements under reliability constraints. The second layer achieved spider web-inspired coverage path optimization. Based on the maximum resource quota from the first layer's quantity allocation result, using the structure of radial threads and capture threads in spider webs, an arbitrary convex quadrilateral region was divided into adaptive sub-regions. Through the combination of radial and parallel paths, coverage optimization was realized and parallel search of UAV swarms was supported. The results demonstrate that the objective optimization performance of the proposed deep learning-based resource allocation network is comparable to the genetic algorithm (GA), and outperforms the deep learning network with linear loss combination of the same structure and the hybrid-loss single-step reinforcement learning method. Its strategic equilibrium is improved by 84.62% compared with GA, and the solution time is greatly reduced. The planning and optimization method for UAV swarm spider web-inspired coverage search is superior to its counterpart methods in terms of the association degree between base stations and sub-regions and the flexibility of search paths. The path equilibrium of sub-regions is increased by more than 75.45%, and the larger the UAV swarm scale, the more significant the optimization effect. The UAV swarm collaborative coverage search framework considers both resource and path optimization. It can improve resource utilization rate and task reliability in urban inspection, emergency rescue, and other scenarios.