Fusion Estimation Of Networked Multi-sensor System With Communication Constraints

With the rapid development of network communication technology and the sensor technology,Networked Multi-sensor Fusion Systems(NMFs)is widely used in medical,military,national defense and high-tech research.Compared with the traditional multi-sensor fusion system,the information transmission mode changes fundamentally by the appearance of the communication network.NMFs can offer many advantages,such as flexible architectures,easier maintenance,simpler installation and low cost,but they also suffer from many communication constraints.Therefore,the fusion estimation of networked multi-sensor system with communication constraints has become one of the hot research topics.In this thesis,we study the following problems on NMFs with limited communication based on the sequential Kalman filtering algorithm and the optimal weighted fusion algorithm in the linear minimum variance sense.Firstly,a real-time sequential filtering fusion algorithm and a fault diagnosis method is proposed for NMFs with communication constraints.The local optimal filter is obtained for the sensors,which has access to the fusion center.The local optimal prediction is obtained for the sensors left according to the latest measurement.In addition,the optimal fusion estimation is derived from the fusion criterion weighted by matrices.The corresponding fault diagnosis method is used to locate thefault sensors.Therefore,sequential information fusion estimation is applied in the networked multi-sensor information fusion systems with communication constraints and missing sensor measurements.By introducing the fictitious noise,a novel stochastic model system is proposed to describe the missing measurement.Sequential Kalman filter method is used to obtain the real-time filtering of the sensors which have access to the fusion center.The local optimal predictor is obtained for the sensors left according to the latest measurement,based on which,the optimal fusion estimation is derived.Our algorithm is simulated by Matlab.Numberical examples illustrated the effectiveness of the fusion Kalman filtered estimator and compared with the local estimator with a single sensor.