Packet Loss System Distributed Information Fusion Estimation

As multi-sensor information fusion technology in military, civilian, and widelyon mine comprehensive monitoring, have attracted many domestic and foreign schol-ars attention. Yet a large number of studies have focused on an unlimited numberof lost packets on a single-sensor control system, the optimal estimate with limitedloss of linear discrete-time stochastic systems have been researched even few. But inpractice we often see observation data and/or known external input data with sys-tem of fnite consecutive packet. For example in networked control system (NCSs),production processes, manufacturing, communications, signal processing, etc. Thedata are transmitted through communication channels, such as from the sensors tothe estimators and/or from the controller to actuator. Since many problems, such asthe variability, the transmission channel of communication unreliability, the agingof components and not enough sensitivity cause packet dropouts, it will result insystem instability and system performance degradation. So it has more importantin theoretical meaning and applying values to study the distributed optimal fusionalgorithm of the multi-sensor system with fnite consecutive packet dropouts. In thisthesis, the linear discrete-time stochastic systems with the fnite consecutive packetdropouts are studied which can be described by two Bernoulli distributed randomvariables. And multi-sensor systems with the fnite consecutive packet dropouts inthe measurement data and/or known external input data, the distributed optimalweighted fusion estimation algorithm is developed. The main contents are as follows:For the network control system with limited packet dropouts, from the sen-sors to the estimators and/or from the controller to actuator. The original systemis transferred into one with the random parameters by using state augmentationmethod, based on the projection theory for bilateral packet dropouts, by introduc-ing two group of new variables. For the new augmented system, the local linearminimum variance optimal Kalman estimators are designed as well as steady-state minimum variance estimators and a sufcient condition for existence of steady-stateestimators.The cross-covariance matrix between any two sensor subsystems is derived formulti-sensor network control systems with bilateral packet dropouts. At last, thedistributed weighted fusion Kalman estimators are obtained by applying distributedweighted fusion estimation algorithms.The massive simulation examples have proven the accuracy and validity of theabove theory.