🤖 AI Summary
To address low exploration efficiency, suboptimal task allocation, and inadequate traversability modeling in unknown indoor environments with multi-scale heterogeneous UAVs, this paper proposes a hybrid-scale UAV collaborative exploration framework. Methodologically: (1) it integrates frontier detection with SphereMap-based traversability modeling to assess point-of-interest (POI) accessibility and generate safe paths; (2) it introduces a dual-strategy task allocation mechanism—combining a real-time priority-weighted greedy algorithm with a globally optimal minimum-cost flow algorithm; and (3) it incorporates heterogeneous sensor fusion and distributed multi-UAV collision avoidance. Evaluated in both simulation and real-world indoor/outdoor experiments, the system significantly improves exploration coverage and time efficiency while enabling dynamic task reassignment and conflict-free autonomous navigation. Key contributions include the first hybrid-scale collaborative UAV framework, the SphereMap traversability modeling paradigm, and an adaptive dual-strategy task allocation mechanism.
📝 Abstract
Heterogeneous teams of Unmanned Aerial Vehicles (UAVs) can enhance the exploration capabilities of aerial robots by exploiting different strengths and abilities of varying UAVs. This paper presents a novel method for exploring unknown indoor spaces with a team of UAVs of different sizes and sensory equipment. We propose a frontier-based exploration with two task allocation strategies: a greedy strategy that assigns Points of Interest (POIs) based on Euclidean distance and UAV priority and an optimization strategy that solves a minimum-cost flow problem. The proposed method utilizes the SphereMap algorithm to assess the accessibility of the POIs and generate paths that account for obstacle distances, including collision avoidance maneuvers among UAVs. The proposed approach was validated through simulation testing and real-world experiments that evaluated the method’s performance on board the UAVs.The paper is supported by the multimedia materials available at https://mrs.felk.cvut.cz/case2024exploration.