程靖, 陈力. 空间机器人双臂捕获航天器后姿态管理、辅助对接操作一体化ELM神经网络控制[J]. 机器人, 2017, 39(5): 724-732. DOI: 10.13973/j.cnki.robot.2017.0724
引用本文: 程靖, 陈力. 空间机器人双臂捕获航天器后姿态管理、辅助对接操作一体化ELM神经网络控制[J]. 机器人, 2017, 39(5): 724-732. DOI: 10.13973/j.cnki.robot.2017.0724
CHENG Jing, CHEN Li. ELM Neural Network Control of Attitude Management and Auxiliary Docking Maneuver after Dual-arm Space Robot Capturing Spacecraft[J]. ROBOT, 2017, 39(5): 724-732. DOI: 10.13973/j.cnki.robot.2017.0724
Citation: CHENG Jing, CHEN Li. ELM Neural Network Control of Attitude Management and Auxiliary Docking Maneuver after Dual-arm Space Robot Capturing Spacecraft[J]. ROBOT, 2017, 39(5): 724-732. DOI: 10.13973/j.cnki.robot.2017.0724

空间机器人双臂捕获航天器后姿态管理、辅助对接操作一体化ELM神经网络控制

ELM Neural Network Control of Attitude Management and Auxiliary Docking Maneuver after Dual-arm Space Robot Capturing Spacecraft

  • 摘要: 讨论了空间机器人双臂捕获航天器后姿态管理和辅助对接操作的协调控制问题.首先,利用冲量定理、闭环约束几何及运动学条件获得了捕获操作后闭链混合体系统的动力学方程,并分析了混合体系统受到的冲击效应.其次,针对捕获操作后系统姿态受扰运动镇定及辅助对接操作需求,对闭链混合体系统提出了基于极限学习机(ELM)的自适应神经网络控制方案,极限学习机具有学习速度快、仅需调节网络输出权值等优点,可用于逼近系统的未知动力学模型.该方案不要求系统动力学方程关于惯性参数呈线性函数关系,并且不需要精确的系统动力学模型.通过李亚普诺夫方法设计了ELM网络的权值自适应律及鲁棒项,以保证系统的载体姿态受扰运动镇定与对接操作过程的位置及角度的精确控制,并证明了系统的稳定性.为保证各臂协同操作,运用加权最小范数法分配力矩.最后,通过系统数值仿真模拟了碰撞冲击效应及闭链系统的运动过程.所提控制方案可以有效完成载荷、载体运动控制及辅助对接操作.

     

    Abstract: The coordinated control problems of attitude management and auxiliary docking operation for a dual-arm space robot capturing a spacecraft are discussed. Firstly, the dynamic equations of the closed-chain composite system after the capturing operation are established by the theorem of impulse, and the geometrical and kinematic conditions of the closed-loop constraints, and the impact effect on the composite system is analyzed. Secondly, an adaptive neural network control scheme based on the extreme learning machine (ELM) is designed for the closed-chain composite system to implement the motion stabilization under attitude disturbance and the auxiliary docking operation of the system after capturing. The ELM is used to approximate unknown dynamical model of the system because it can implement fast learning and only needs to adjust the output weight values of the network. In the proposed control method, the dynamic equations of the system needn't be linearly dependent on inertial parameters, and the precise system dynamic model isn't required. The weight adaptive law of ELM network and the robust items are designed through Lyapunov method, to guarantee the motion stabilization of the base under attitude disturbance and the precise position and attitude control during docking operation. The system stability is proved. The torques are distributed by the weighted minimum-norm theory to ensure the cooperation of the manipulators. Finally, the collision impact effect and the movement process of the closed-chain system are demonstrated through numerical simulation. The proposed control scheme can efficiently complete the motion control of both the base and the load, as well as the auxiliary docking operation.

     

/

返回文章
返回