Multisource risk fusion and safe path planning for low-altitude campus inspection

Abstract

Low-altitude campus inspection requires unmanned aerial vehicles (UAVs) to fly through dense buildings, temporary no-fly constraints, moving human and vehicle exposure, uneven communication and Global Navigation Satellite System (GNSS) quality, and local weather disturbances. This paper presents a multisource risk fusion and safe path planning method for such missions. The campus airspace is discretized as a three-dimensional (3D) grid. Obstacle, no-fly-zone, human/vehicle, communication and Global Navigation Satellite System (GNSS), and weather risks are normalized and fused into a searchable risk field, while buildings and no-fly areas are treated as hard constraints. A risk-aware three-dimensional A-star (3D A*) planner then combines distance, fused risk, and altitude-change costs, with inspection-order optimization and conservative path pruning to reduce redundant turns. During execution, future-path risk is monitored and local replanning is triggered when dynamic events increase exposure. Experiments on a 500 m × 500 m simulated campus show that the proposed method reduces integrated risk while preserving task reachability. In 30 randomized multitask missions, the method reduces mean integrated risk by 73.81%, shortens mean path length by 26.23%, and increases mean obstacle clearance by 9.51 m relative to distance-only A*. The results suggest that fusing heterogeneous risk sources with task-level ordering and dynamic replanning can provide an interpretable and reusable safety layer for campus-scale low-altitude inspection.