Research On Covariance Cross Fusion Algorithm Based On Genetic Algorithm

Multi-sensor fusion estimation is an important part of the field of state estimation,and has important application value in the fields of aerospace,navigation and guidance,target recognition and intelligent manufacturing.For multi-sensor systems,when the cross-covariances are known,the optimal fusion algorithms weighted by matrices,diagonal matrices and scalars can be obtained in the sense of linear minimum variance criterion.However,in practical applications,the cross-covariances are usually unknown,and how to effectively solve the fusion estimation problem with unknown cross-covariances becomes the focus and difficulty of research.Therefore,for the multi-sensor system with unknown cross-covariances,the main research work in this paper is as follows:（1）Under the framework of batch covariance intersection（BCI）fusion algorithm,the original scalar weighting coefficients are replaced by positive definite diagonal matrices,and a CI fusion algorithm weighted by diagonal matrices（DCI）is proposed.Compared with the classical CI fusion algorithm,this algorithm is more flexible in structure and has higher fusion accuracy.The unbiasedness,robustness and accuracy relationship of the algorithm are proved theoretically.（2）A joint optimization scheme based on genetic simulated annealing algorithm and BP neural network is proposed.The BP network is trained offline,and the trained neural network is used for online fusion.The optimization process of the weighting coefficients during fusion is avoided,and the local optimal value is "filtered",so the real-time performance and fusion accuracy of the system are improved.（3）A fast covariance intersection fusion algorithm weighted by diagonal matrices（FDCI）is proposed in this paper,and the unbiased and robust accuracy of the proposed FDCI algorithm is proved.The algorithm does not involve the optimization of the nonlinear cost function of multiple weight coefficients,which greatly reduces the computational burden,improves the real-time performance of the system,and is more suitable for nonlinear and other complex multi-sensor systems with unknown cross-covariances.Aiming at the situation of asynchronous data transmission,a sequential fast covariance intersection fusion algorithm weighted by diagonal matrices is proposed,and the batch form and related theorems of the algorithm are deduced.Secondly,combined with the CKF algorithm,a fast covariance intersection fusion weighted by diagonal matrices Cubature Kalman filter algorithm is proposed,which can effectively solve the fusion estimation problem of nonlinear multi-sensor systems with unknown cross-covariances.