Distributed Recursive Fusion Estimation For Multi-sensor Asynchronous Sampling Systems

With the development of computer technology,multi-sensor information fusion technology has gained significant attention at home and abroad.Multi-sensor information fusion technology is to obtain the information measured by sensors and combine the information with certain criteria so as to gain the accurate estimation for the target.Due to the sensor aging,fault and the limited communication broadband,the random transmission delay phenomenon often happens when the data is received by the processing center.At the same time,each sensor may have different sampling rates under different cases.These factors can lead to asynchronous problems.In this paper,the distributed optimal recursive linear fusion estimation problem is studied for multisensor asynchronous sampling systems under the linear unbiased minimum variance criterion.The main research contents are as follows:The multi-sensor asynchronous linear discrete random systems with uncorrelated noise are studied in this paper.The asynchronous sampling systems are transformed into synchronous sampling systems by a pseudo-observation method and an iteration state equation method.The local filter and the estimation error cross-covariance matrices between local estimates are given,respectively.The estimation error cross-covariance matrices between the prior fusion estimate and local estimates are derived.Then,a distributed optimal recursive linear fusion filter without feedback is proposed.The proposed algorithm has higher accuracy than the distributed fusion filtering algorithm where local estimates are only weighted by matrices,but it has accuracy loss compared with the centralized fusion algorithm.In order to improve the fusion accuracy,a distributed optimal recursive linear fusion filter with feedback is proposed which has the same accuracy as the centralized fusion filter.Then,the optimality is proved.Simulation results show the effectiveness of proposed algorithms.The multi-sensor asynchronous uniformly sampling linear discrete systems with correlated noise are studied.The asynchronous systems are transformed into synchronous systems by using the pseudo-observation method.The local filter and the estimation error cross-covariance matrices between local estimates are given.The estimation error cross-covariance matrices between the prior fusion estimate and local estimates are derived.The distributed optimal recursive linear fusion predictors and filters without and with feedback are proposed,respectively.It is proved that the fusion estimate with feedback has the same accuracy as the centralized fusion estimation,namely,it has optimality.Simulation results show the proposed algorithm is effective.The multi-sensor asynchronous non-uniform sampling linear discrete random systems are studied.The state-weighted method is used to transform the asynchronous system to synchronous one.The local filter and the estimation error cross-covariance matrices between local estimates for asynchronous non-uniform sampling systems are given.The distributed optimal recursive linear fusion filtering algorithms without and with feedback are proposed.It is proved that the distributed optimal recursive linear fusion filtering algorithm with feedback has the same estimation accuracy as the centralized fusion,namely,it has optimality.Simulation results show the effectiveness of the proposed algorithms.