Research On The Target Tracking Based Of Fisher Information Matrix

The research work is based on the theoretical of unscented Kalman filter(UKF)in WSNs.Since the present methods based on distance can't cluster adaptive nodes for target tracking in WSNs,a new clustering algorithm based on UKF and Fisher information matrix named UKF-F is presented that combines the UKF algorithm and the Cramer-Rao lower bound theory of stochastic vector estimation.In order to enhance the survivability of the network,the residual energy of nodes are added to the election criterion of the cluster heads,and the most suitable sensor node is activated as the tracking cluster when the information criteria of the current target is calculated according to the filtered error covariance matrix.By comparing the simulation with the nearest neighbor algorithm,it is proved that the dynamic clustering method based on Fisher information matrix can select the tracking node more suitably,and prove the decrease of the tracking error.In order to reduce the computational burden of cluster heads and the total traffic of WSNs in the centralized filter algorithm,a sequential UKF algorithm is proposed for multi-target tracking in the nonlinear Gaussian environment with data sorting algorithm named fuzzy C-means clustering,which meets the UKF-F node selection dynamic clustering algorithm in WSNs.This method allocates the cluster head's work to each member node,which greatly reduces the data transfer between the nodes,avoids the data collision and interference,and simulates and contrasts with centralized filtering algorithm,it reduces the network energy consumption to the maximum extent at the time improving estimated accuracy.