作　　者： ; ;

机构地区： 华南理工大学自动化科学与工程学院

出　　处： 《自动化学报》 2013年第6期806-815,共10页

摘　　要： 针对产生回归轨迹的连续非线性动态系统,确定学习可实现未知闭环系统动态的局部准确逼近.基于确定学习理论,本文使用径向基函数(Radial basis function,RBF)神经网络为机器人任务空间跟踪控制设计了一种新的自适应神经网络控制算法,不仅实现了闭环系统所有信号的最终一致有界,而且在稳定的控制过程中,沿着回归跟踪轨迹实现了部分神经网络权值收敛到最优值以及未知闭环系统动态的局部准确逼近.学过的知识以时不变且空间分布的方式表达、以常值神经网络权值的方式存储,可以用来改进系统的控制性能,也可以应用到后续相同或相似的控制任务中,节约时间和能量.最后,用仿真说明了所设计控制算法的正确性和有效性. Deterministic learning can achieve locally-accurate approximation of the unknown closed-loop system dynamics while attempting to control a class of nonlinear systems producing recurrent trajectories. Based on deterministic learning, an adaptive neural control algorithm is proposed for unknown robots in task space using radial basis function （RBF） networks. The designed adaptive neural controller can not only guarantee all signals in the closed-loop system uniformly ultimately bounded, but also achieve convergence of partial network weights to their optimal values. It can also learn the unknown closed-loop system dynamics in a stable control process along recurrent tracking orbits. The learned knowledge stored as constant network weights can be reused in a same or similar control task to improve the control performance and to save time and energy. Simulation results demonstrate the effectiveness of the proposed approach.

关 键 词： 确定学习 径向基函数神经网络 自适应神经网络控制 机器人

领　　域： [自动化与计算机技术] [自动化与计算机技术]