The difficulties that unscented Kalman filter encounters when it is introduced in reentry vehicle trajectory were investigated. These difficulties include abnormal measurement, the inaccuracy of measurement random error model and dynamic model. The current study uses an adaptive robust filter that can produce the estimation of equivalent weight flexibly according to measurement noise and state noise, can distinguish abnormal measurement from normal measurement, and can estimate the variance of Wiener model self-adaptively. The simulated results testified that the filter is easy to implement and can reduce the bad influences derived from inaccuracy of measurement random error model and dynamic model.