A combined power activation function based convergent factor-variable ZNN model for solving dynamic matrix inversion

Abstract

The application of zeroing neural network (ZNN) to solve multifarious time-varying problems, especially the dynamic matrix inversion (DMI), is widely used in recent years. As the core components of ZNN model, the activation function (AF) and convergent factor (CF) always occupy a momentous position in its development. In this paper, a convergent factor-variable ZNN (CFVZNN) model with a novel combined power activation function (CPAF) and a time-varying adjustable CF is proposed for online DMI solution. Unlike other existing conventional ZNN (CZNN) models, the proposed CFVZNN model has the advantages in both fixed-time convergence and anti-noise property, and these superiors of the proposed CFVZNN model are verified by strict mathematical derivation. Besides, several successful examples for solving DMI problems and tracking control of mobile manipulator in noisy environment further validate the practical application prospects of the proposed CFVZNN model.