Architecture of a generic vehicle for evaluating localization algorithms

Development of a powerful localization system has suggested number of works. Localization is always done in a two phases process: prediction and estimation. Prediction (which integrates the odometric data) is quite similar whatever the method is. But, the estimation could be done in very different ways. Thus, we saw the emerging of several technics using probabilistic, interval analysis and Markovian methods. Depending on the chosen method, the localization can be performed in complex, dynamic or badly mapped environments. So as to compare and test some localization methods (which where never been used in real environment), we decided to design a generic mobile platform. This platform had to deal with several specificities like to be controlled by a distant operator, to give an estimation of the localization on an environment map and to provide a visual return to the user.