冗余度卫星搭载机械臂神经网络优化控制

Optimal Control of the Space-Craft Based Manipulators with Neural Networks

研究了节点变换函数参数可调的多层神经网络 ,并将这种神经网络用于控制冗余度卫星搭载机器人系统 ;提出了卫星位姿最小摄动的冗余度机器人臂神经网络控制方案 ,在这种方案中将系统的正运动学方程作为神经网络学习的对象 ,由神经网络学习拟合系统逆运动学 ,实现了机器人臂运动和卫星运动的解耦 .在机器人臂的运动和卫星本体解耦的基础上 ,基于带动量项的BP算法 ,实现了系统的最小动能优化控制 .利用平面 4关节星载机器人系统进行了仿真 ,取得比较理想的效果 .

In this paper, the multi layer forward neural network with flexible sigmoid function was used in controlling the spacecraft based redundant manipulators. The optimal control scheme that can guaranty to minimize the disturbance of the spacecraft by manipulators is proposed. In the scheme, the forward kinematic equations of the system were needed, neural network simulates the inverse kinematic equations which decouple the motion between the manipulators and the spacecraft base. The problem of minimizing the kinetic energy of the system was also considered. A four body planar modal was used to simulate the system, the results verified that the method is successful.