Distributed Information Fusion Estimation For Singular Systems With Multi - Step Stochastic Hysteresis And Multiple Packet Loss

Descriptor system is a general description form of dynamical system. It’s more general than the normal state space system. It has been widespread used in the ?eld of economy, electric power and the robots. Meanwhile, with the rapid development of network communication technology, network transmission is widely used in various aspects of the national defense and national economy. However, the introduction of the network bring convenience to the control system, but some problems such as random delays and random packets dropouts, are unavoidable. Therefore, based on singular value decomposition method and optimal weighted fusion algorithm in the linear minimum variance sense, this article studies information fusion estimation algorithms for multi-sensors descriptor system with multiple random delay and multiple dropouts respectively for three kinds of situations. 1)The data can be sent repeatedly at the sensor side, and the estimators can only receive a packet or nothing at each time; 2)A packet at the sensor side is only sent once, the estimators only receive one packet or nothing at each time. And sensors have random failures; 3)A packet at the sensor side is only sent once. Possibly one or multiple data packets,or nothing arrives at the data processing center at each moment. Main contents are as follows:Firstly, the original descriptor system is transformed into a fast and a slow reduced-order subsystems based on singular value decomposition.For the fast reducedorder subsystem, random parameters are introduced into a new state variables. And the fast subsystem is transformed into an augmented stochastic parameter system.Based on this augmented stochastic parameter system, the local estimators and white noise estimators are developed in the linear minimum variance sense.Then,based on the local estimators and white noise estimators of the fast reduced-order subsystem, the local estimators of the slow reduced-order subsystem are derived.Further, the computation formulas for error covariance matrices between any two sensor subsystems for each reduced-order subsystems are derived to compute the fusion weights. Then, based on the three optimal weighted fusion algorithms in the linear minimum variance sense, three distributed fusion estimators of the fast and the slow reduced-order subsystems are obtained. Then, the distributed fusion estimators of the original descriptor system are obtained based on the original nonsingular linear transformation. The computation formulas for error covariance matrices between the fast and the slow reduced-order subsystems are derived, the corresponding fusion estimation error variance matrices are obtained. The MATLAB software is used to do the simulation research. The simulation studies show the e?ectiveness of the distributed fusion estimation algorithms.