基于不完全测量信息的随机忆阻神经网络的参数与状态估计问题研究

项目名称： 基于不完全测量信息的随机忆阻神经网络的参数与状态估计问题研究

项目编号： No.61503001

项目类型： 青年科学基金项目

立项/批准年度： 2016

项目学科： 自动化技术、计算机技术

项目作者： 刘宏建

作者单位： 安徽工程大学

项目金额： 20万元

中文摘要： 随机忆阻神经网络是当前国际上备受关注的研究热点，其参数及状态估计问题是非常重要的基础研究问题。本项目将充分考虑在网络化环境下由于受到外部随机扰动或物理设备限制而引起的测量信息不完全现象，深入研究网络化系统中的忆阻特性和随机因素的发生及变化规律，建立起能反映网络诱导的不完全信息、随机和忆阻等特性的动力学模型和信息测量模型。针对不完全测量信息情形下的随机忆阻神经网络系统，首先给出相应的参数估计方法，进而利用切比雪夫引理、遍历性定理和微分包含等数学工具，分析所给参数估计误差的相容性、一致收敛性和渐近正态性。然后，基于忆阻的特性和系统增广法构造网络系统状态估计器，利用李雅普诺夫泛函、拉萨尔不变原理等方法分析状态估计误差系统的稳定性和鲁棒性等动态性能。最后，利用计算机进行数值模拟仿真，验证所给估计方法与算法的有效性。本课题的研究成果将丰富随机忆阻神经网络的参数及状态估计理论。

中文关键词： 随机忆阻神经网络；参数估计；状态估计；不完全测量信息；随机系统

英文摘要： The stochastic memristive neural network is an extensive concerned research hot topic, and the design and synthesis issues of state estimator is a very important basic problem. The incomplete information caused by the network devices with the external stochastic disturbances or limit capacity is fully discussed and the memristive natures and stochastic factors as well as their variety laws are studied deeply. Then the kinetic model with measurement model is established that is capable of representing the networked-induced incomplete information and stochastic and memristive natures of the systems under consideration. For the stochastic memristive neural networks with incomplete measurement, the parameter estimation problems will be investigated firstly. Then by employing the techniques of Chebyshev lemma, ergodic theory and differential inclusion, the consistency, asymptotic properties, and local asymptotic normality will be studied by a set of conditions. Subsequently, based on memristive properties and the augmented method, the state estimators will be designed. And the dynamic performance such as robustness and stability will be analyzed for the error systems by Lyapunov functional, LaSalle invariance principle. Finally, the effectiveness and applicability of the proposed estimation theory will be demonstrated by the computer numerical simulation. The theory on parameter estimation and state estimation of stochastic memristive neural networks will be enriched by the investigation of this project.

英文关键词： stochastic memristive neural networks;parameter estimation;state estimation; incomplete measurement;stochastic systems