带时间相关噪声多传感器网络化系统的分布式融合估计算法研究

In this project, we will carry out the research on the linear optimal, suboptimal and distributed fusion estimation new algorithms for multi-sensor networked systems with time-correlated additive and multiplicative noises. Differently from the correlated noises in the most existing literatures, here the correlated noises are described by a general autoregressive moving average (ARMA) model which includes both finite-step and infinite-step correlation. At the same time, we consider the possible random packet dropouts from sensor to local estimation and from local estimation to the fusion center. When a packet is lost, its predictor is used for compensation. Differently from the state augmentation and measurement differencing methods in the most existing literatures, here we will propose new non-augmentation methods. First, based on the innovation analysis theory, three new local estimation algorithms will be proposed for single sensor system by state differencing, high-order moving average (MA) process fitting and some fictitious states, respectively. Then, for multi-sensor system, based on the three proposed local estimators, the corresponding three distributed fusion estimation new algorithms will be proposed in the linear minimum variance sense. Further, the condition for the existence of the steady-state fusion estimation will be explored. The research on this project will provide feasible theoretical basis and technology methods to solve some information fusion estimation problems for systems with time-correlated noises.

本项目将对带有时间相关加性和乘性噪声的多传感器网络化系统，开展线性最优、次优和分布式融合估计新算法的研究。不同于目前大多数文献中考虑的相关噪声，本项目采用具有一般性的自回归滑动平均(ARMA)模型刻画噪声的相关性，综合考虑了有限步和无限步相关。同时考虑传感器数据在传输到局部估计以及局部估计在传输到融合中心的过程中可能存在的随机丢失现象，基于观测预报和融合的状态预报补偿丢失的数据。不同于目前大多采用的状态增广方法和观测差分方法，本项目将提出新的非增广方法。首先，基于新息分析理论，对单传感器系统，拟分别采用状态差分、高阶MA过程拟合和虚拟状态三种方法，提出三种局部估计新算法。然后，对多传感器系统，基于所提出的三种局部估计，在线性最小方差意义下，提出相应的三种分布式融合估计新算法，并探索稳态融合估计存在的条件。本项目的研究将为解决带时间相关噪声系统的融合估计问题提供可行的理论依据和技术途径。

本项目研究了带时间相关加性和/或乘性噪声系统的状态估计问题，其中相关噪声由具有一般性的自回归滑动平均(ARMA)模型刻画，即综合考虑了有限步和无限步相关。同时考虑传感器数据在传输过程中可能存在的随机丢失现象，通过设计观测预报器补偿丢失的观测数据。基于新息分析和非增广方法在线性最小方差意义下开展了线性最优估计和信息融合估计新算法的研究。主要成果包括：1) 提出一种新的非增广方法处理带时间相关噪声系统2)基于所提出的非增广方法，对带时间相关加性噪声或乘性噪声以及随机数据包丢失的多传感器不确定系统，分别提出了局部最优、分布式最优融合和稳态融合估计新算法。并对上述所提算法进行了稳态分析，给出稳态估计存在的一个充分条件；3)对带时间相关观测噪声或乘性噪声的多传感器系统，提出了全局最优序贯融合和带反馈的全局最优分布式估计新算法，并对所提算法的全局最优性进行了严格的理论性证明。即所提算法在相同初值条件下可获得与集中式融合估计相同的估计精度。以上研究成果进一步丰富了状态估计理论，为解决带时间相关噪声多传感器系统的信息融合估计问题提供可借鉴的方法。