This paper explores a way of combining conventional inertial sensors with cold atom interferometers (CAI) in order to reduce the drift of the navigation solutions in velocity and orientation. Instead of complementing and improving the CAI with conventional sensors, in this approach the conventional IMU will be used as main sensor for a prediction of the kinematic state. The CAI is then used for the correction of systematic errors and offsets in the framework of an extended Kalman Filter. Monte Carlo simulation studies demonstrate an improvement of the navigation solution precision. In addition, most drifts of velocity and orientation can be eliminated and the uncertainty of the velocity solution can further be reduced by a factor of 30 or more compared to the conventional strapdown. The observability of the error states is discussed.