AbstractTheNAVSTAR GlobalPositioningSystem(GPS) is the satellite navigation system of the future. It is now in the development stage of its validation phase. Precise estimation and prediction of navigation satellite ephemerides and frequency standard characteristics are essential to projected user navigation accuracy. This paper presents the approach being implemented to achieve that accuracy.The approach blends the successful batch estimation technique (CELEST) developed by the Naval Surface Weapons Center with an on‐line Kalman estimation technique to be implemented at the GPS Master Control Station (MCS). Both techniques use the same precisely measured and smoothed pseudoranging data collected by geometrically spaced Monitor Stations (MS) located at Hawaii; Elmendorf AFB, Alaska; Guam; and Vandenberg AFB, California.CELEST employs detailed force models in a batch‐weighted least‐squares estimation of the ephemerides. It then predicts the ephemerides and partial derivatives for yet another long period. These ephemerides and partial derivatives provide excellent predicted references and simple ephemeris models for the linear recursive Kalman estimator to be resident at the Master Control Station.The Kalman estimator operates about the predicted reference to “fine tune” the ephemerides and estimate frequency standard characteristics in a timely on‐line manner. Timeliness is necessary to update the frequency standard drift characteristics. The ephemeris and frequency standard models are also used to predict ephemerides and characteristics on a daily basis using last‐minute estimates. Precise representation of the predictions are then transmitted to the satellites which, in turn, transmit current dat