Fusion Of Multi - Step Random Hysteresis And Multi - Packet - Loss Network System

With the development and fusion of network, communication and automatic control technologies, the networked control systems have become the focus of attention. Networked control systems have been widely used in national economy due to the advantages of low cost, making it convenient to share resources and remote operation. However, with the increasing of the scale of the networked control systems and the complexity of the system, random delays and packet dropouts are always exist in data exchange by the limitation of networked capacities and the communication bandwidths. In addition, environmental disturbances and sensor fadings can also make uncertainties of systems. Amiming at the above problems, this paper mainly focuses on the filtering fusion problems for networked stochastic uncertain systems with packet dropouts and random delays based on the projection theory and the distributed fusion algorithm in the linear minimum variance sense. The main research contents are as follows.For multi-sensor linear discrete-time stochastic systems with random delays and packet dropouts with/without time stamps, the error cross-covariance matrices between any two local filters that depend on the value of random variables and the probability of random variables are derived, respectively. Based on the existing local filter and filtering error cross-covariance matrices we derived, the distributed fusion filter and centralized fusion filter are designed. Further, for the distributed fusion filter and centralized fusion filter that depend on the probability of random variables, the steady state characteristic is anaylized.For multi-sensor discrete-time stochastic systems with random multiplicative noise, random delays and packet dropouts, the local linear optimal filter is presented in the linear minimum variance sense. As the random multiplicative noises are both in the state and observation equations, the distributed fusion filter weighted by matrices and centralized fusion filter are presented, and a sufficient condition for the existence of steady state fusion filters is given. Further, as the random multiplicative noise is only in the state equation, the CI fusion filter that can avoid calculating the filtering error cross-covariance matries is presented to reduce the computational burden.For multi-sensor discrete-time stochastic systems with uncertain observations, random delays and packet dropouts, the local optimal linear filter is presented in linear minimum variance sense, then the error cross-covariance matrices between any two local filters are derived, and a distributed fusion filter weighted by matrices is given. The existence of steady state fuser weighted by matrices is proven. For systems with one-step random delay, the cetralized fusion filter is proposed and a sufficient condition for the existence of steady state is given.