Abstract: For the general requirement of space robot to have advantages such as long arms and huge load for space on orbit operation, the flexible problem of mechanical arm should be considered for analyzing space robot's dynamics and control. The coordinated motion control and vibration suppression problems of flexible arm space robot with unknown parameters were discussed. With the law of momentum conservation and assumed mode method, the Lagrange principle was utilized to model the dynamic function of flexible arm space robot. On these bases, mathematical models which were suit for control system design were established by using double time scale decomposition of singular perturbation theory. Aiming at the slow subsystem, Radial Basis Function (RBF) neural network control algorithm with uncertain parameters was designed to dominate the trajectory tracking of coordinated motion. The research purpose of neural network control algorithm was to improve the control accuracy of whole system based on good on line self learning of neural network. For fast subsystem, hierarchical fuzzy control algorithm was used to control the vibration of flexible link. The research purpose of hierarchical fuzzy control algorithm was to reduce the size of fuzzy rule base, and raises the calculation efficiency of fuzzy controller effectively. Computer simulation results illustrated the effectiveness and feasibility of proposed algorithms.

Key words: flexible arm space robot, singular perturbation theory, radial basis function, neural network, hierarchical fuzzy control algorithm

摘要： 由于太空中的在轨作业一般要求空间机器人具有手臂长、负载大，对空间机器人动力学及控制进行分析时必须考虑到机械臂杆件的柔性问题。为此探讨了参数未知柔性臂空间机器人载体姿态、关节协调运动控制及柔性振动抑制问题。由系统动量守恒关系及假设模态法，利用Lagrange方法建立了漂浮基柔性臂空间机器人的系统动力学方程。运用奇异摄动理论的双时间刻度分解，导出了适用于控制系统算法设计的奇异摄动数学模型。利用该模型，针对慢时标子系统——等价刚性空间机械臂，设计了系统参数不确定情况下的径向基函数神经网络补偿控制算法，以控制柔性臂空间机器人的载体姿态及机械臂关节铰协调地来完成各自在关节空间的期望运动；神经网络控制算法的研究目的是基于神经网络良好的在线自学习能力，大大提高整个系统的控制精度。针对快时标子系统——柔性臂的振动，设计了分级模糊控制算法来主动抑制柔性杆的振动。分级模糊控制算法的研究目的是为了减少模糊规则库的大小，有效提高模糊控制器的计算效率。通过计算机仿真验证了提出方法的有效性和可行性。

关键词: 柔性臂空间机器人, 奇异摄动理论, 径向基函数, 神经网络, 分级模糊控制算法