PCDCT: Perception-Complementarity-Driven Collaborative Trajectory Generation for Vision-Based Aerial Tracking

This article proposes a perception-complementarity-driven trajectory generating method for multiple unmanned aerial vehicles (UAVs), which can effectively enhance the visibility of the target for UAVs in unknown environments. Traditional methods often rely on prior maps or additional sensors to assist with obstacle avoidance. Nevertheless, these methods are not only costly but also fail to effectively avoid occlusions caused by obstacles. Different from existing methods, the trajectory planned by the proposed method not only enables the vision-based UAVs to maintain the perception of obstacles and the target on the one hand, but also preserves topological equivalence with the predicted target trajectory on the other hand. Specifically, a vision-based mutual observation approach among UAVs is proposed to enhance the overall perception capability of the UAVs system. On this basis, a target-guided collaborative trajectory planning method is proposed to ensure the planned collision-free trajectory for other UAV in the formation maintains target visibility. In addition, a trajectory feasibility assessment method is proposed to obtain the collaborative trajectory planned by the UAV at the optimal observation location. Finally, comparative simulations are conducted with three state-of-the-art methods, demonstrating the advantages of the proposed method in maintaining target visibility and tracking efficiency during the vision-based aerial tracking. The real-world experiment demonstrates the feasibility of the proposed method. Note to Practitioners—Most existing vision-based multi-UAV target tracking methods require additional prior maps or laser sensors to assist in obstacle avoidance. In practical tracking scenarios, the limited perception range of cameras poses significant challenges for UAVs in synchronously observing moving target and environmental obstacles. The article proposes a trajectory generation method based on perceptual complementarity to ensure that the generated trajectory effectively perceives surrounding obstacles while enhancing the observation capability of UAVs towards moving target. The key insight of this work is to utilize mutual observation among multiple UAVs to assist in planning collision-free tracking trajectories, ensuring safety during the tracking process. Building upon this, by preserving topological equivalence with the predicted target trajectory, there is an improvement in the target visibility ratio during the tracking process. Furthermore, a trajectory feasibility assessment method is proposed to obtain the optimal collaborative trajectory from the UAV positioned at the optimal location in the formation. The effectiveness of the proposed method in improving flight safety and target visibility is validated through comparative experiments.

I. Introduction

Over the past few decades, robots have seen increasing application in the field of automation science and engineering. The rapid advancement in autonomous systems, artificial intelligence, and sensor technology has greatly expanded their capabilities. These advancements have not only increased the operational scope but also enhanced their ability to perform complex tasks with high precision and efficiency.