Multi-UAV 3D Hierarchical Coverage Path Planning in Dynamic Environments for Large and Complex Structures

In the field of structural inspection methodologies, Unmanned Aerial Vehicles have emerged as a suitable alternative to conventional inspections led by technicians due to their various advantages. However, suboptimal paths can pose risks to the structure under inspection, personnel, and nearby equipment. This paper addresses the challenge of path planning in dynamic environments, where people and equipment move near the structure during drone inspections, with a focus on minimizing inspection time, a hierarchical approach is proposed, combining Reinforcement Learning (RL) for the Vehicle Routing Problem (VRP) and OR Tools for the Travelling Salesman Problem (TSP). This method leverages Deep Reinforcement Learning (DRL) to adapt to various scenarios by adjusting environmental parameters, ensuring efficiency through parallelizability. The approach shows an average performance gap of 4.82% compared to Google Operations Research (OR) Tools in static settings and is evaluated for adaptability through experiments in various practical scenarios.