In this paper, a new interference-tolerant fast convergence zeroing neural network (ITFCZNN) using a novel activation function (NAF) for solving dynamic matrix inversion (DMI) is presented and investigated. Compared with the original zeroing neural network (OZNN) models, the proposed ITFCZNN not only has the ability to converge to 0 within a fixed-time, but also resist different types of interference and noises in solving DMI problems. Besides, detailed mathematical analysis of convergence and robustness of the ITFCZNN are provided. Comparative numerical simulation verifications of the new ITFCZNN and the OZNN activated by other commonly used activation functions (AF) are also provided to demonstrate the better robustness, effectiveness and fixed-time convergence of the ITFCZNN. In addition, a mobile manipulator path tracking application example is given to verify the applicability and feasibility of the ITFCZNN with interference and noises. Both of the theoretical analysis and numerical simulation results verify the effectiveness and robustness of the ITFCZNN model.

本文提出并研究了一种使用新型激活函数（NAF）解决动态矩阵求逆（DMI）的新的抗干扰快速收敛归零神经网络（ITFCZNN）。与原始的归零神经网络（OZNN）模型相比，所提出的ITFCZNN不仅具有在固定时间内收敛到0的能力，而且在解决DMI问题时还可以抵抗不同类型的干扰和噪声。此外，还提供了ITFCZNN的收敛性和鲁棒性的详细数学分析。还提供了对新ITFCZNN和由其他常用激活函数（AF）激活的OZNN的比较数值模拟验证，以证明ITFCZNN具有更好的鲁棒性，有效性和固定时间收敛性。此外，给出了一个移动机械手路径跟踪应用示例，以验证ITFCZNN在干扰和噪声的情况下的适用性和可行性。理论分析和数值模拟结果均验证了ITFCZNN模型的有效性和鲁棒性。