Homography-Based Visual Servoing of a UAM for Payload Delivery Onto a Dynamic Target
提出一种动态单应性视觉伺服控制策略，通过构建旋转鲁棒的单应性矩阵和高阶滑模观测器，实现无人机机械臂在动态目标上的精确载荷投放，无需直接速度测量。
This article proposes a dynamic homography-based visual servoing (HBVS) control strategy for payload delivery by an uncrewed aerial manipulator (UAM). Unlike traditional approaches that assume planar-target motion, this strategy constructs a rotation-robust dynamic homography matrix to decouple image dynamics from the adverse effects of target rotation. Ahigh-order sliding mode observer (HOSMO) based on decoupled error signals is developed to address unmeasurable relative velocities, lumped disturbances, and the susceptibility of existing observers to target rotation. The proposed HOSMO isolates the estimation process from target rotation, thereby considerably improving estimation accuracy. This observer enables the UAM to achieve stable tracking within the HBVS framework without requiring direct velocity measurements or prior knowledge of external disturbances. Moreover, a virtual camera-based dual-loop HBVS architecture is proposed to dynamically coordinate the UAM and the onboard manipulator, overcoming the restricted fields of view (FOVs) and heterogeneous visual features that limit existing manipulator control strategies. Isomorphic visual features are unified to dynamically correct UAM deviations, improving payload delivery accuracy. The stability of the proposed control strategy is rigorously analyzed using Lyapunov theory. Finally, comparative simulations and outdoor experiments are conducted to validate the effectiveness of the proposed strategy.