漂浮基双臂空间机器人自主跟踪、稳定捕获与智能化操作方法研究

漂浮基双臂空间机器人执行舱外在轨维护任务，具有拟人化、寿命长和安全性高等优势。但由于其具有非完整性以及空间任务不确定等难点，其自主跟踪、稳定捕获与协调操作方面还远未达到在轨维护要求，需进一步深入研究。自主跟踪方面，本项目拟提出基于增广视觉雅可比的目标轨迹跟踪方法，并解决非完整性引起的动力学奇异问题；目标稳定捕获方面，研究基于闭合-锁定策略的视觉/力混合控制，实现空间目标柔顺捕获并抑制碰撞力对组合体稳定性影响；协调操作方面，拟提出一种基于深度强化学习的智能化操作方法，提高其执行复杂不确定空间任务的适应性和操作协调性；设计一种可变构型灵巧夹持器并搭建混合地面实验系统来开展空间典型操作实验研究。本项目成果不仅可以为我国漂浮基双臂空间机器人实现舱外在轨维护任务提供理论与实验基础，而且对于类似非完整机器人的传感器伺服控制和智能学习控制等共性问题的研究，具有重要的科学意义和学术价值。

Free-floating dual-arm space robot has many advantages, such as humanoid, long life, and high safety in conducting extra-vehicular on-orbit services. Because of difficulties in nonholonomic constraints and uncertainties during space tasks etc., many research aspects, such as autonomous tracking, stable capture and coordinated manipulation, needs to be investigated to fulfill requirements of on-orbit services. For autonomous tracking, this project proposes a target trajectory tracking method based on generalized visual jacobian matrix in order to avoid dynamic singularity caused by nonholonomic constraints. For stable capture of the target, hybrid visual/force control based on close-lock strategy is proposed to compliantly capture the space object and suppress disturbance of the impact to the spacecrafts’ combination stability. For coordinated manipulation, an intelligent manipulation method based on deep reinforcement learning is proposed to improve the adaptability to complicated space tasks and the coordination of manipulation. A reconfigurable dexterous gripper is designed and a hybrid ground experiment platform is built to conduct typical manipulation experiment. This project can provide theoretical and experimental experience in extra-vehicular on-orbit services for free-floating dual-arm space robot. In addition, this project can offer scientific and academic values in common problems such as sensor control and intelligent learning control for nonholonomic robots.