非线性关联系统自适应神经网络输出反馈分散控制

Adaptive neural network output-feedback decentralized control for nonlinear interconnected systems

针对一类带有完全未知关联项的非线性大系统,提出一种自适应神经网络输出反馈分散控制方法.采用神经网络逼近未知的关联项,因此对关联项常做的假设如匹配条件,被上界函数所界定等不再要求.在神经元输入中采用参考信号取代关联信号,从而成功地避免了对关联信号的微分.保证了闭环系统所有信号半全局一致最终有界,证明了跟踪误差收敛于一个包含原点的小残集.

An adaptive neural network output-feedback decentralized control scheme is proposed for a class of largescale nonlinear systems with completely unknown interconnections. Neural networks are employed to approximate to the unknown interconnections, eliminating the common assumptions on interconnections such as matching condition, being bounded by upper bounding functions. By replacing the interconnected signals in neural inputs with the reference signals, the differentiation of interconnected signals is then successfully avoided. Moreover, all signals in the closed-loop system are guaranteed to be semi-globally uniformly ultimately bounded, and the tracking errors are proved to converge to a small residual set around the origin.