Square-root filtering and its applications in target tracking
HU Guang-da · Ha'erbin gongye daxue xuebao · 2008
The propagation of the covariance matrix in those most commonly used nonlinear filtering algorithms,such as the extended Kalman filter and the unscented Kalman filter,leads to the lose of its positive definiteness in the update process and makes the filters invalid.To lessen the effect of numerical errors,the square-root approach can be used.Based on Joseph form of the covariance updating,a new concise square-root extended Kalman filter is presented.The square-root unscented Kalman filter is also employed in the simulation.Performance and implementation in target tracking on reentry show their efficiency.