Research Of Filtering Algorithms For Discrete Stochastic Systems With Uncertain Observed Information

This thesis is concerned with optimal filtering problem for a class of discrete stochastic systems with different sources of noises, random time delays and multiple packet dropouts, the discrete stochastic system model with multiplicative noises, correlated noises, random time delays and multiple packet dropouts are proposed, and these system models are analysed deeply, the optimal filtering is investigated for these models. Different sources of noises(multiplicative noises, autocorrelated noises and crosscorrelated noises) are added to systems, the goal of which is to describe systems truly; the uncertain observed information(random sensor delays and multiple packet dropouts) is an inevitable and also a common problem in most systems. In order to control the system better and estimate the state of systems accurately, it is great significance in theory and practice to concern the systems with uncertain observed information.At first, the optimal filtering problem is researched for a class of discrete stochastic systems with multiplicative noises and random two-step sensor delays, the process noises and measurement noises of the systems are uncorrelated white Gaussian noises. Based on the innovation analysis approach and recursive projection formula, a new optimal filter is designed in the sense of the minimum mean square error, the simulation example is given to illustrate the effectiveness and feasibility of the proposed filtering scheme.Secondly, the globally optimal filtering problem is investigated for for a class of discrete stochastic systems with different sources of noises, random one-step sensor delay and multiple packet dropouts, the process noises of the systems are one-step autocorrelation. Based on the MMSE principle and the concept of optimal estimation, a new globally optimal filter is designed for the systems, the simulation example is shown to illustrate the effectiveness and feasibility of the proposed filtering scheme. Then, the problem is extended further, considering the process noises are finite-step autocorrelated, the globally optimal filtering problem is studied for a class of discrete stochastic systems with finite-step autocorrelated process noises, and the globally optimal filter is derived for the addressed systems, the simulation example is shown to illustrate the effectiveness and feasibility of the designed filtering.Finally, the optimal information fusion filtering problem is solved for a class of multi-sensor discrete stochastic systems with different sources noises, random one-step sensor delay and multiple packet dropouts, the process noises of the systems are finite-step autocorrelation, Based on the criterion weighted by matrices in the linear minimum variance sense, a optimal information fusion filtering is designed for the addressed systems, the simulation example is shown to illustrate the uniformity of the proposed filtering scheme.