动态规划求解空间双臂机器人非完整运动最优控制问题  

Dynamic Programming for Solving the Optimal Nonholonomic Motion Control Problem of the Bi-Arm Space Robot

作　　者：赵金刚[1] 戈新生[2] 

机构地区：[1]北京信息科技大学自动化学院,北京100192 [2]北京信息科技大学理学院,北京100192

出　　处：《力学季刊》2016年第2期225-233,共9页Chinese Quarterly of Mechanics

基　　金：国家自然科学基金(11472058)

摘　　要：通过自适应动态规划研究自由漂浮空间双臂机器人运动的最优控制问题.针对空间双臂机器人的非完整性运动,采用自适应动态规划(Adaptive Dynamic Programming,ADP)方法求解其最优控制问题.根据多体动力学理论,推导了载体位置、姿态均无控制条件下,双臂空间机器人满足的系统动量守恒关系的非完整约束方程,并将其转化为控制系统的状态方程,从而将双臂空间机器人的非完整运动规划问题转化为对非线性系统的控制问题.文中根据自适应动态规划网络结构,利用神经网络来近似性能指标函数,进而用龙格库塔法求解状态变量.并给出了适合该类问题的一种效用函数具体表达式,保证了空间双臂机器人到达期望位置后不再继续运动.实现了对空间双臂机器人非完整运动规划的最优控制.数值仿真实验验证了ADP对求解空间双臂机器人非完整运动规划最优控制问题的有效性.In this paper the optimal nonholonomic motion planning of the free-floating space robot system with two arms is solved using adaptive dynamic programming. Based on the principle of the translational and rotational momentum conservations in the multi-body dynamics theorem, the nonholonomic motion equations of the bi-arm space robot system are established, which are then transferred into the state equations of the optimal control for the system. In this way, the nonholonomic motion planning of the robot system is translated into the solution of a canonical nonlinear control problem. Based on the network of adaptive dynamic programming, the performance index function can be approximated within the neural network, and the state variables can be solved using Runge-Kutta integration algorithm. In addition, we provide an explicit expression of the utility function which is suitable for this kind of problem. The objective of the optimal control ofnonholonomic motion planning is achieved that the investigated bi-arm space robot no longer move once it reaches the desired location. The effectiveness of optimal control of nonholonomic motion planning for free-floating space robot system using ADP is verified by the numerical simulation experiments.

关 键 词：空间机械臂 非完整运动规划 最优控制 自适应动态规划 神经网络 

分 类 号：TP29[自动化与计算机技术—检测技术与自动化装置]