RUSH: Rapid UAV Spatial Hierarchical Exploration via Regional Viewpoint Generation for Large-Scale Environments

Exploring large-scale environments quickly and autonomously using unmanned aerial vehicles (UAVs) remains a challenge. Two major issues, long-distance back-tracking and the UAV's low-velocity flight, significantly hinder exploration efficiency. To tackle this problem, we propose a spatial hierarchical exploration method combining rapid regional viewpoint generation. This involves dividing the exploration space into subregions using an online hgrid spatial decomposition, and determining the order of exploration for these subregions by solving a non-closed traveling salesman problem. Optimal viewpoints are chosen from these subregions based on a global loss function related to frontiers. By dividing the space into subregions and optimizing the path globally, we can reduce the UAV's back-tracking distance. Additionally, the selected optimal viewpoints allow UAVs to make smaller turns, avoid obstacles, and achieve better coverage, which helps decrease the occurrence of low-velocity movements and back-tracking. We also incorporate a velocity loss constrain to improve local trajectories, ensuring high-velocity flight. Our proposed method has been analyzed and validated through simulations and real-world tests, showing improved exploration efficiency compared to several leading methods, particularly in large-scale environments.