A chained form state representation for outdoor vehicle localisation

In this paper, we tackle the outdoor localization problem for autonomous vehicle navigation. A vehicle can be represented by a unicycle kinematic model which is a non holonomic one. The use of Kalman filter has been extensively used for multi-sensor data fusion for robotized vehicle localization. The non-linearity of evolution model gave raise to several extensions of classical Kalman filters such as extended Kalman filter and unscented Kalman filter or other approximative method. In this paper, we propose to use chained form state representation in order to avoid the approximative linearization made by approximate methods. Experimental vehicle equipped with a GPS, wheel encoders and gyroscope is used in order to test performance and validity of the proposed approach. Finally, performance comparison between Kalman filter applied on a chained form, extended Kalman filter and unscented Kalman filter is presented.