Abstract:

A node three-dimensional estimation algorithm that combines the weighted centroid localization algorithm and simplified maximum correntropy unscented Kalman filter is proposed for solving the problems that the observation noise, which appears in estimating the three-dimensional states of the nodes, is heavy-tailed or has some sudden change in the wireless sensor network. First, the algorithm obtains the observation distance of beacon nodes and sensor nodes by using the method of signal strength ranging and gets the approximate estimation of nodes with the centroid localization method. And then a simplified maximum correntropy unscented Kalman filter algorithm is deduced by combining the node estimation model and the robustness of the maximum correntropy criterion for non-Gaussian and nonlinear problems. Finally, the accurate estimation is obtained by using it. Simulation results show that the new algorithm has a better performance for the three-dimensional state estimation of nodes than the classic methods in sensor networks with heavy-tailed non-Gaussian noise. It not only reduces the time complexity of the general maximum correntropy unscented Kalman filter, but also improves the accuracy of node estimation.

Key words: sensor networks, unscented Kalman filter, maximum correntropy criterion, state estimation