具有马尔可夫通信过程的分布式估计器设计  

Design of a Distributed Estimator with Markovian Communication

作　　者：綦祥[1] 刘成成[1] 蔡云泽[1] 

机构地区：[1]上海交通大学自动化系系统控制与信息处理教育部重点实验室,上海200240

出　　处：《控制工程》2013年第S1期190-194,共5页Control Engineering of China

基　　金：国家自然科学基金(61004088);上海航天科技创新基金(SAST201237)资助;航空科学基金(20100157001)联合资助

摘　　要：针对一类可以描述成为离散线性时不变系统的智能体编队系统,研究分布式估计器设计的方法。编队系统中的每个智能体的分布式估计器都对全局系统状态进行估计,各个分布式估计器之间实现的通信过程均视为一阶马尔可夫过程,同时通信过程受到噪声影响。对每一个分布式估计器利用估计误差状态方程和马尔可夫过程的平稳分布构成双线性矩阵不等式的稳定条件,对双线性矩阵不等式分解得到的两个线性矩阵不等式进行交替求解得到一个分布式估计器的估计误差协方差矩阵。反复迭代求解系统中所有分布式估计器,最终得到分布式估计器的反馈增益和通信增益。并仿真验证了设计方法的有效性。For a formation of agent system can be described as discrete linear time-invariant system,research on design method for distributed estimation in this article.Each distributed estimator of agent in the formation system estimate the state of the global system.Communication procedures between distributed estimators are regarded as a first-order Markov process and the communication process are affected by noise.Using estimate error state equation and the stationary distribution of the Markov process for each distributed estimators constitute a bilinear matrix inequality as stable condition.Bilinear matrix inequality can be decomposed into two linear matrix inequalities.Alternately solving two inequalities get covariance matrix of estimate error.Iteratively solving all distributed estimation of the feedback gain and a communication gain can be obtained.Finally,the effectiveness of the design method is verified by simulation.

关 键 词：分布式 估计器 线性矩阵不等式 马尔可夫过程 

分 类 号：TP13[自动化与计算机技术—控制理论与控制工程]