Automotive radar has an advantage over other sensors in that it is better at operating in bad weather conditions. To see the extent of the effect that adverse weather conditions might have on the statistics of the data a statistical analysis was performed on real measurement data. During heavy rain there is a shift that can be observed in the Radar Cross Section (RCS) of the target. The average RCS of the target increases slightly when it is raining. In (Multiple Input Multiple Output) MIMO radar it is important to calibrate the radar system as there can be both amplitude and phase distortions between the channels that can give unexpected results. These are usually estimated in a predetermined setting for known targets. However instead it might be feasible to estimate this from objects of opportunities that are regularly appearing in the radar field of view.
To tackle this problem a method is used that tries to estimate these calibration coefficients from measurement data. The method needs to know the angle at which the target is located, however the range of the target can remain unknown. It uses the ideal steering vector and one of the antenna elements as a reference element. The method can recreate the phase errors very well, but relies on the reference element for the amplitude estimation. Therefore the performance is based on what element is chosen as a reference. To choose the right reference element some pre-processing is done. Then the estimation of the calibration coefficients was implemented in a Simultaneous Localization And Mapping (SLAM) framework. This was solved by using an Extended Kalman Filter (EKF). The EKF is a nonlinear form of the normal Kalman filter that will be used to make an estimate for both the location of the radar, the location of the objects of opportunity and the estimation of the calibration coefficients based of these landmarks at the same time. The resulting algorithm proves that it is feasible to calibrate the radar while driving in this way.