Hou Nan, Dong Hongli, Wang Zidong, Liu Hongjian
IEEE Trans Neural Netw Learn Syst. 2021 Nov;32(11):5167-5178. doi: 10.1109/TNNLS.2020.3027252. Epub 2021 Oct 27.
In this article, the robust finite-horizon state estimation problem is investigated for a class of time-varying complex networks (CNs) under the random access protocol (RAP) through available measurements from only a part of network nodes. The underlying CNs are subject to randomly occurring uncertainties, randomly occurring multiple delays, as well as sensor saturations. Several sequences of random variables are employed to characterize the random occurrences of parameter uncertainties and multiple delays. The RAP is adopted to orchestrate the data transmission at each time step based on a Markov chain. The aim of the addressed problem is to design a series of robust state estimators that make use of the available measurements from partial network nodes to estimate the network states, under the RAP and over a finite horizon, such that the estimation error dynamics achieves the prescribed H performance requirement. Sufficient conditions are provided for the existence of such time-varying partial-node-based H state estimators via stochastic analysis and matrix operations. The desired estimators are parameterized by solving certain recursive linear matrix inequalities. The effectiveness of the proposed state estimation algorithm is demonstrated via a simulation example.
在本文中，针对一类时变复杂网络（CNs），研究了在随机接入协议（RAP）下，仅通过部分网络节点的可用测量进行鲁棒有限时域状态估计问题。底层的复杂网络受到随机出现的不确定性、随机出现的多重延迟以及传感器饱和的影响。采用若干随机变量序列来表征参数不确定性和多重延迟的随机出现情况。基于马尔可夫链采用随机接入协议在每个时间步编排数据传输。所解决问题的目的是设计一系列鲁棒状态估计器，这些估计器利用部分网络节点的可用测量，在随机接入协议下并在有限时域内估计网络状态，使得估计误差动态满足规定的H性能要求。通过随机分析和矩阵运算，为存在此类基于时变部分节点的H状态估计器提供了充分条件。通过求解某些递归线性矩阵不等式对期望的估计器进行参数化。通过一个仿真示例证明了所提出状态估计算法的有效性。
