Research On State Estimation Method Based On Multi-source Information Fusion

Multi-source information fusion is a multi-level and multi-faceted process,for instance,multi-source data detection,correlation,combination and estimation.In accordance of them,the accuracy is improved about state and identity estimation.Its high detection accuracy,good scalability,high reliability,low access to information and other advantages,are widely used in modern military and various civil areas.Besides,multi-source information fusion plays a vital role in the development of modern science and technology including promote people's lives.In many fields,the multi-source information fusion theory is especially under the military event,which has an obvious effect through the maneuvering target state accuracy.Based on the theory of multi-source information fusion,the research of some problems has a proper solution in maneuvering target state estimation.The main contributions are as follows:Aiming at the problem that bootstrapped measurement is easy to be affected through the uncertainty of the measurement noise in the single-sensor measurement bootstrapped distributed Kalman filter.There is an original algorithm is proposed about the multi-sensor measurement bootstrapped distributed Kalman filter,in accordance of the physical characteristics of multi-sensor measurement system,and the formation mechanism of bootstrapped measurement in single-sensor measurement bootstrapped distributed Kalman filter.This algorithm constructs a set of bootstrapped measurement in multi-sensor system,extending the application range of the bootstrapped measurement strategy from single sensor to multi-sensor system.Besides,we eliminated the influence of measurement noise uncertainty and realized the improvement of the filtering accuracy of the estimated system through extraction and utilization the redundant and complementary information of multi-sensor measurement.Furthermore,a selective sampling strategy is introduced in the sampling process of the bootstrapped measurement,which increases the reliability of bootstrapped measurement and improves the estimation accuracy of the system.Experiment illustrates that a more accurate estimation of target state has achieved as results of the original algorithm with smaller root mean square error instead of the distributed kalman filtering algorithm based on real measurement.Aiming at the linear estimated constraint in the scenario of Kalman consensus filter,combined with the cubature Kalman filter and consensus strategy,a novel Cubature Kalman consensus filtering algorithm is proposed.In the realization of algorithm,the distributed fusion framework is adopted.Firstly,measurement data form the capable-communication adjacent nodes are sampled,which are applied to cubature Kalman filter to achieve the distributed estimation of system state.Secondly,according to consensus strategy,these local state estimations in the whole sensor network are optimized.And then the estimation precision of system state is improved by enhancing the consensus of each sensor node.Compared to standard Kalman consensus filter,the algorithm makes consensus strategy extend to nonlinear system estimation.The theoretical analysis and experimental results verify the feasibility and efficiency of the proposed algorithm.Aiming at the problem of incomplete measurement in the condition of the two sensors to obtain the measurement information,we proposed a measurement reconstruction of distributed Kalman filtering algorithm,which takes into account the incomplete measurement.Firstly,a new set of two sensors is generated when the sampling sequence parity and the measurement information exchange have a mapping relationship.Then,based on the measurement set,the pair of Kalman filters are constructed,combined with the distributed weighted fusion techniques to obtain state estimates of targets.Experiments show that compared with the distributed Kalman filter algorithm,the novel algorithm can make better use of the measurement information without extra hardware cost.Expecially filtering accuracy is improved owing to smaller root mean square error.