A relative navigation algorithm based on the adaptive extended Kalman filter is studied in this paper to solve problems of the decline of relative position precision, which is induced by the cause that the measurement noise cannot accurately know. Firstly, the Taylor series expansion is applied to transform the measurement matrix to the linearization model. Secondly, the covariance matrix of the measurement noise is estimated dynamically by using the estimation method of adaptive time-variant noise, and the covariance matrix of the state noise is got via the prior knowledge of inertial navigation. Finally, the simulation results indicate that the research method in this paper can provide continuous and smooth relative position information with highly precision, and the method still show the better navigation parameter estimation performance in the case of the varying measurement noise happens especially.