Localization and tracking of multiple quadrotors with collision avoidance : theory and experiment

Abstract

Multiple quadrotors target encirclement is widely used in the intelligent field, as it can effectively monitor and control target behavior. However, it faces the danger of collision, as well as difficulties in localization and tracking. Therefore, we propose a complete target encirclement method. Firstly, based on Hooke's law, a collision avoidance controller is designed to maintain a safe flying distance among quadrotors. Then, based on the consensus theory, a formation tracking controller is designed to meet the requirements of formation transformation and encirclement tasks, and a stability proof based on Lyapunov was provided. Besides, the target detection is designed based on YOLOv5s, and the target location model is constructed based on the principle of pinhole projection and triangle similarity. Finally, we conducted experiments on the built platform, with 3 reconnaissance quadrotors detecting and localization 3 target vehicles and 7 hunter quadrotors tracking them. The results show that the minimum average error for localization targets with reconnaissance quadrotors can reach 0.1354 m, while the minimum average error for tracking with hunter quadrotors is only 0.2960 m. No quadrotors collision occurred in the whole formation transformation and tracking experiment. In addition, compared with the advanced methods, the proposed method has better performance.