Abstract: When tracking maneuvering reentry vehicles, an extended Kalman filter with maneuvering reentry vehicle (MaRV) model is usually used in a traditional method. To increase the performance of maneuvering target tracking two efforts are made. On the one hand, a more appropriate multiple model method is used in describing movement properties of the maneuvering target, on the other hand, the Unscented Kalman filter implying higher order precision is adopted in respective matching model. For numerical problems which are often encountered in implementing this interactive multiple model Unscented Kalman filter, a numerical robust solution is given by using square root filtering.