This paper presents the practical use of Kalman filtering by combining global navigation satellite system signals with inertial sensor measurements in a sensor fusion scheme for rockets. A robust Schmidt–Kalman filter formulation is derived to consider the effects of parameter uncertainty in the sensor models. In this scheme, some uncertain parameters are added to the state vector and recursively estimated, whereas other uncertain parameters are not estimated, but the effect of their uncertainty is explicitly accounted for. The inertial measurement unit provides accelerometer and rate gyro measurements along and about the vehicle’s body axes. Both sensors have parametric uncertainty on the scale factor, misalignment, nonorthogonality, and bias. Furthermore, these parameters can drift with time, temperature, and vibrations. The global navigation satellite system receiver is subject to non-white-noise sequences and temporary loss of signal. This paper presents a comparison between different inertial sensor grades and demonstrates the efficiency and consistency of the estimation technique, both in simulation and using recorded flight data from the fourth flight of the Vega launch vehicle: the European Space Agency’s small launcher.