漂浮基空间机器人的径向基神经网络鲁棒自适应控制  

The RBF neural network robust adaptive control of a free-floating space robot

作　　者：张文辉[1] 高九州[1] 马静[2] 齐乃明[1] 

机构地区：[1]哈尔滨工业大学航天学院,黑龙江哈尔滨150001 [2]东北农业大学工程学院,黑龙江哈尔滨150001

出　　处：《智能系统学报》2011年第2期114-118,共5页CAAI Transactions on Intelligent Systems

基　　金：中国航天科技集团创新基金资助项目(CAST09C01)

摘　　要：针对一类同时具有参数及非参数不确定性的自由漂浮空间机器人系统的轨迹跟踪问题,采用了一种RBF神经网络的自适应鲁棒补偿控制策略.对于系统的参数不确定性,通过对径向基神经网络来自适应学习并补偿,逼近误差通过滑模控制器消除,神经网络权重的自适应修正规则基于Lyapunov函数方法得到;而非参数不确定通过鲁棒控制器来实时自适应估计,且未知上界不需要先验的知识.该方法从整个闭环系统的稳定性出发设计的神经网络动态补偿的鲁棒控制器,并通过引入PD反馈来便于工程应用,这种鲁棒的神经网络控制器,可以有效提高收敛速度并保证其控制精度.试验结果进一步证明了这种自适应神经网络控制算法的有效性.The trajectory tracking of a class of free-floating space robot manipulators with parameter and non-parameter uncertainties was considered.An adaptive robust compensation control algorithm was proposed based on an RBF neural network.Neural networks are used for adaptive learning and compensating the unknown system for parameter uncertainties.The approaching error was eliminated by a sliding controller.The neural network weight adaptive correction laws were obtained based on the Lyapunov analysis approach,which can ensure the convergence of the algorithm.Non-parameter uncertainties were estimated and compensated in real time by a robust controller.The unknown upper bound was shown not to need priori knowledge.This control scheme is easy to use in engineering by introducing a PD feedback and designing a robustness controller in which the neural network is dynamically compensated based on the stability of the whole closed loop system.It was proven that the controller can guarantee the asymptotic convergence of tracking errors,good robustness,and the stability of a closed-loop system.The simulation results show that the presented method is effective.

关 键 词：神经网络 鲁棒控制 空间机器人 自适应控制 

分 类 号：TP242[自动化与计算机技术—检测技术与自动化装置]