IEEE Trans Neural Netw Learn Syst. 2021 Jul;32(7):2952-2964. doi: 10.1109/TNNLS.2020.3009081. Epub 2021 Jul 6.
An innovative class of drive-response systems that are composed of Markovian reaction-diffusion memristive neural networks, where the drive and response systems follow inconsistent Markov chains, is proposed in this article. For this kind of nonlinear parameter-varying systems, a suitable gain-scheduled controller that involves a mode and memristor-dependent item is designed, so that the error system is bounded within a finite-time interval. Moreover, by constructing a novel Lyapunov-Krasovskii functional and employing the canonical Bessel-Legendre inequality and free-weighting matrix method, the conservatism of the finite-time synchronization criterion can be greatly reduced. Finally, two numerical examples are provided to illustrate the feasibility and practicability of the obtained results.
本文提出了一类由马尔可夫反应扩散忆阻神经网络组成的驱动-响应系统，其中驱动和响应系统遵循不一致的马尔可夫链。对于这种非线性时变系统，设计了一种合适的增益调度控制器，其中包含模式和忆阻器相关项，以使误差系统在有限时间间隔内有界。此外，通过构造一个新的李雅普诺夫-克拉索夫斯基泛函，并利用典范的贝塞尔-勒让德不等式和自由加权矩阵方法，可以大大降低有限时间同步判据的保守性。最后，通过两个数值例子验证了所得结果的可行性和实用性。
