Autonomous Perching of Unmanned Aerial Manipulator Based on Visual Servoing Control

This study presents a collaborative control frame-work for UAM perching tasks, integrating image segmentation-based tracking with multimodal NMPC. Conventional perching strategies are susceptible to dynamic occlusions and illumination variations in complex environments, while also struggling to achieve an optimal trade-off between energy efficiency, tracking accuracy, and contact compliance. To address these challenges, this study introduces a dynamic task prioritization strategy and pixel-level perception enhancement techniques to enable precise, full-process regulation from target detection to perching execution. The proposed multimodal NMPC employs a phase-specific optimization of the weight function, facilitating precise control design tailored to different flight phases of the perching task. In terms of visual perception, the integration of image segmentation and motion compensation techniques significantly enhances the robustness of target recognition under dynamic occlusions and complex lighting conditions.