This master thesis investigates the performance of a car navigation system using lateral accelerometers, yaw rate and bearings relative three known radio beacons. Accelerometer, gyroscope and position data has been collected by an IMU combined with a GPS receiver, where the IMU was installed in the approximate motion center of a car. The bearing measurements are simulated using GPS data and the measurement noise model is derived from an experiment where the direction of arrival to one transmitter was estimated by an antenna array and the signal processing algorithm MUSIC.

The measurements are fused in a multi-rate extended Kalman filter which assumes that all measurement noise is Gaussian distributed. This is not the case for the bearing measurement noise which contains outliers and therefore is modelled as a Gaussian uniform noise mixture. Different methods to deal with this have been investigated where the main focus is on the principle to use the Kalman filter’s innovation for each bearing measurement as an indication of its quality and discarding measurements with a quality above a certain threshold.