基于神经网络的空间柔性机械臂PID快速学习控制  

Fast learning PID control over space flexible manipulators based on neural network

作　　者：张文辉[1,2] 马静[1] 高九州[2] 

机构地区：[1]徐州师范大学机电工程学院,江苏徐州221116 [2]哈尔滨工业大学航天学院,黑龙江哈尔滨150001

出　　处：《应用科技》2011年第5期1-4,共4页Applied Science and Technology

基　　金：中国航天科技创新基金资助项目(CAST-HIT09C01)

摘　　要：针对自由漂浮柔性空间机器人的轨迹跟踪问题,提出一种径向基函数(RBF)神经网络控制策略.首先建立漂浮基柔性空间机器人的非线性动力学方程,考虑到RBF神经网络良好的逼近能力,柔性臂的非线性逆动力学模型通过RBF网络来逼近,采用PID控制器与神经网络控制器来共同保证系统稳定性,其误差代价函数由PID控制器提供,采用固定中心参数,而扩展宽度采用启发式关系确定,网络权值采用改进的最优准则算法进行调整来实现快速学习能力.仿真结果表明了这种RBF神经控制器能够达到较快的误差收敛速度.For the trajectory tracking problem of free-floating space robot with flexible manipulators,this paper adopts an adaptive RBF neural network control algorithm.Firstly,nonlinear dynamic equation of free-floating space flexible manipulators is established.Considering good approach ability of RBF neural network,it is used to adaptively learn and compensate inverse-model,proportion integration differentiation（PID） controller and RBF neural network are used to guarantee stability of the system.Error cost function is offered by PID controller with fixed central data and the extension width is determined by heuristic relation.The network weight is adjusted by the improved optimal criterion algorithm to realize fast learning ability.The simulation results show that the presented controller can achieve the asymptotic convergence of tracking error.

关 键 词：RBF神经网络 柔性空间机器人 启发式学习 PID控制 

分 类 号：TP242[自动化与计算机技术—检测技术与自动化装置]