Sensors data fusion for the absolute localisation of a mobile robot by a possibility theory based method. Comparison with Kalman filter

摘要

In order to improve autonomous system, it is necessary to determine accurately its position . In this paper, a method basedudon possibility theory has been developed in the experimental framework of the localisation of a miniature mobile robot fromudodometry reading and exteroceptive sensors into an environment equipped with beacons . The data are modelled in the setting ofudthe possibility theory which provides interesting tools of representing imprecision and uncertainty . A comparison with a classical method (Kalman filter) taken as a reference is realised . Basically, the fusion procedure by Kalman filtering method can be seenudas a weighted average by the information uncertainties . Its principle is to favour the information with low uncertainty (i.e. lowudvariance) and to realise a fusion by "minimisation of variance" . On the other hand, the adaptive combination rule used in theudpossibilistic method takes into account the level of conflict between the sources and favour the redundancy of the informationudwhich are in agreement . Then, it is rather a fusion by "agreement" . In spite of these fundamental discrepancies, the outcomes ofudsimulation and/or of experiences on the real robot, obtained by both methods are satisfactory and quite close .