【摘要】 针对传统自适应控制需要满足匹配条件、激发信号存在以及逼近误差有界等条件,提出一种新的基于神经网络的一类非线性系统自适应反步控制器设计方案。使用三层神经网络逼近系统的非线性特性,通过网络权系数自适应调整来不断的在线估计未知的逼近误差上界,采用有σ修正项的自适应律以放松持续激励条件。给出了基于Lyapunov意义上的闭环系统稳定性分析,证明跟踪误差收敛于原点的一个ε领域内。仿真结果表明了所提反步控制器的正确性。更多还原

【Abstract】 Traditional adaptive control requires some assumptions, such as match conditions, the existence of triggering signal and the bounded approximation error. This paper proposes a novel scheme of adaptive backstepping control for a class of nonlinear systems based on neural networks. The scheme utilizes the nonlinearity of 3-layer neural network approximation systems, continuously estimates the unknown upper bound of approximation error by the self-adaption of network weights, and employs adaptive law with the amendmentσ to relax persistent excitation condition. Moreover, we give an analysis of closed-loop system stability in the sense of Lyapunov, which proves that the tracking errors will be convergent to the range of the original point with the variation ε . Simulation results verify the effectiveness of the proposed scheme.更多还原