A collaborative target tracking algorithm for multiple UAVs with inferior tracking capabilities

Abstract

Target tracking is one of the hottest topics in the field of drone research. In this paper, we study the multiple unmanned aerial vehicles (multi-UAV) collaborative target tracking problem. We propose a novel tracking method based on intention estimation and effective cooperation for UAVs with inferior tracking capabilities to track the targets that may have agile, uncertain, and intelligent motion. For three classic target motion modes, we first design a novel trajectory feature extraction method with the least dimension and maximum coverage constraints, and propose an intention estimation mechanism based on the environment and target trajectory features. We propose a novel Voronoi diagram, called MDA-Voronoi, which divides the area with obstacles according to the minimum reachable distance and the minimum steering angle of each UAV. In each MDA-Voronoi region, the maximum reachable region of each UAV is defined, the upper and lower bounds of the trajectory coverage probability are analyzed, and the tracking strategies of the UAVs are designed to effectively reduce the tracking gaps to improve the target sensing time. Then, we use the Nash Q-learning method to design the UAVs’ collaborative tracking strategy, considering factors such as collision avoidance, maneuvering constraints, tracking cost, sensing performance, and path overlap. By designing the reward mechanism, the optimal action strategies are obtained as the control input of the UAVs. Finally, simulation analyses are provided to validate our method, and the results demonstrate that the algorithm can improve the collaborative target tracking performance for multiple UAVs with inferior tracking capabilities.

摘要

目标跟踪是无人机领域研究热点之一。本文针对无人机跟踪性能不占优, 以及目标具有灵活、智能运动特征的情形, 研究了多无人机协同目标跟踪问题。提出一种基于目标意图估计的多无人机协同跟踪策略。首先设计了一种具有降维和最大感知覆盖约束的轨迹特征提取方法, 以降低无人机跟踪代价, 并对目标典型的3类运动模式, 根据环境和目标轨迹主要特征, 设计了一种意图估计方法;然后, 设计了一种在障碍物环境中基于最小可达距离和最小转角代价的MDA-Voronoi图, 证明分析了目标被感知的概率;接着, 设计了无人机的协同跟踪策略, 以减小目标跟踪丢失的间隙, 增加目标被感知的时间;通过纳什Q学习方法, 在奖励函数中考虑了避障、跟踪代价、感知质量、飞行约束等因素, 将最优动作策略作为无人机的控制输入。最后, 通过仿真验证了本文方法能在无人机跟踪性能不占优的情况下提高跟踪质量。