带有丢包的分布式网络系统的滚动时域融合估计策略  

Moving Horizon Fusion Estimation Strategy for Distributed Network Systems with Packet Loss

作　　者：范建明 薛斌强 FAN Jianming;XUE Binqiang(School of Automation, Qingdao University, Qingdao 266071, China)

机构地区：[1]青岛大学自动化学院,山东青岛266071

出　　处：《复杂系统与复杂性科学》2021年第3期75-79,87,共6页Complex Systems and Complexity Science

基　　金：国家自然科学基金青年项目(61603205);中国博士后基金项目(2017M612205);青岛市博士后应用项目(2016022)。

摘　　要：针对具有随机丢包分布式网络系统的状态估计问题,提出了一种带有信息融合策略的滚动时域估计方法。系统中每个节点能够通过网络与相邻节点传输数据,并在丢包时利用零阶保持器采用上一时刻数据来补偿测量值。基于滚动优化原理,传感器节点利用本地测量数据和相邻节点传来的数据将状态估计问题转化为固定窗口内的二次优化问题,并使用标量加权线性最小方差的方法对性能指标进行加权融合,从而得到一种新的估计方法。该方法保证了系统的一致性,并在保证准确度的情况下减少了计算时间。最后分析了估计器的稳定性并在仿真中与其它估计方法进行了比较,结果表明该方法具有更好的效果。A distributed robust moving horizon estimation method is proposed for distributed network systems with random packet loss.Each node in the system can transmit data over the network to its neighbors,and when a packet is lost it uses Zero-order hold data to compensate for the measurements.Based on moving horizon principle,sensor node transforms the state estimation problem into a quadratic optimization problem in a fixed window by using the local measurement data and the data from the adjacent nodes,the performance index is fused with the method of scalar weighted linear minimum variance.The method guarantees the consistency of the system and reduces the calculation time while the accuracy of the results is guaranteed.Finally,the stability of the estimator is analyzed and compared with other estimation methods in simulation.The results show that this method has better effect.

关 键 词：随机丢包 分布式网络系统 状态估计 滚动时域 

分 类 号：TP273[自动化与计算机技术—检测技术与自动化装置]