考虑运动衍化与抓捕策略的空间非合作目标最优抓捕点确定研究

One of the key technologies in the implementation of on-orbit servicing and maintenance is using space manipulator to capture a non-cooperative target autonomously. Due to the nonlinear and chaotic characteristics of the non-cooperative target, how to determine an optimal capture point on the target only based on the observation data and the space manipulator model, become a great challenge in the field of on-orbit servicing. This investigation considers the constraints of uncertain capture opportunity, restrict joint drive ability and output range, possible collision etc. and explores the issues of the non-cooperative target motion prediction and space manipulator capture strategy design. First of all, the target motion state transition probability distribution is determined by using autoregressive hidden Markov model. The target dynamic safety zone is also calculated in a finite time domain. Secondly, the capability map of the space manipulator is obtained using Monto-Carlo method when the dynamic coupling effect between the space manipulator and its base is taken into account. The feasible capture point set can therefore be found on the target. Finally, the capture strategy design issue under the constraints of uncertain capture opportunity, restrict joint drive ability and output range is transformed to a multi-object nonlinear optimization problem under dynamic multi-constraints. The optimal capture point is now chosen based on the designed capture strategy to fulfill the requirements of safety, feasibility and optimality. The proposed analytical theory and computational methods will provide characteristic solutions to the present challenging academic issue of implementation of space manipulator autonomous capture under dynamic environment, which also lay a foundation for the engineering applications.

利用空间机械臂自主抓捕非合作目标是实现在轨服务与维护的关键技术之一。鉴于非合作目标的运动呈现出复杂非线性与混沌特性，如何仅基于目标观测数据与空间机械臂模型，自主确定目标上的最优抓捕点，成为航天领域的极大挑战。本项目围绕抓捕时机未定、关节驱动能力及输出受限、存在可能碰撞等约束，探索非合作目标运动衍化与空间机械臂抓捕策略设计问题。首先，结合自回归隐Markov模型，推导目标运动状态迁移概率分布，进而在有限时域内确定动态安全区；其次，考虑空间机械臂与其基座间的动力学耦合，利用蒙特卡洛法构建空间机械臂能力映射图形，确定目标上可行的抓捕点集；最后，将抓捕时机未定、关节驱动能力及输出受限下的机械臂抓捕策略设计转化为动态多约束下的多目标非线性优化问题，据此确定目标上满足安全性、可行性与最优性需求的抓捕点。本研究对动态环境下空间机械臂实现自主抓捕这一学术难题，提出有特色的方法，同时为工程应用提供理论基础。

本项目以空间机械臂自主抓捕非合作目标为研究背景，构建非合作目标最优抓捕点自主确定的新型理论与计算方法。主要研究了基于历史观测数据的空间非合作目标运动状态预测、空间复杂拓扑构型机械臂的能力映射分析、基于非线性优化理论的动态多约束空间机械臂轨迹规划。取得的研究进展包括：..①基于隐Markov过程并通过蒙特卡洛方法进行多次打靶，给出目标潜在的位姿与速度，分析了抓捕概率；此外，基于稀疏高斯过程对目标运动进行预测，在保证与标准高斯过程同样的预测精度时，所提方法的预测效率是标准高斯过程的近10倍。..②采用空间向量与图论相结合，构建空间多臂机器人基于Lagrange力学体系的多体动力学模型，提出了动力学耦合因子与耦合椭球的概念，用于评估机械臂-基座之间的耦合程度。..③提出了机械臂工作空间网格化方法与工作球的概念，基于逆运动学求解和统计分析，构建了空间机械臂的操作能力图谱。..④结合目标运动预测、空间机械臂的工作空间与操作能力，设计了抓捕可行性评估的判定条件，并给出了抓捕时机初确定的一维搜索方法。..⑤基于RRT*算法，协调规划末端执行器的初始可行路径，采用样条曲线平滑设计路径，将抓捕策略转化为跟踪所设计路径的非线性二次规划问题，并采用基于任务优先级的QP进行求解。..⑥利用B样条的高度灵活性和局部修正特性，机械臂各关节轨迹基于Clamped B样条进行参数描述，同时考虑抓捕过程中的关节角、速度、避撞等动态约束，将抓捕策略转化为多约束、多目标非线性优化问题，并采用启发式智能优化算法求解。..本项目研究可应用于空间高动态目标的自主在轨捕获，对于提高我国空间的战略威慑力、丰富空间对抗的手段具有重要的意义，也可拓展其应用于地面机器人的复杂操作。