Regarding assistance to disabled people for object manipulation and carryin
g, the paper focuses on the localisation for mobile robot autonomy. In orde
r to respect strong low-cost constraints, the perception system of the mobi
le robot uses sensors of low metrological quality, ultrasonic ring and odom
etry. That poses new problems for localisation, in particular. Among differ
ent localisation techniques, we present only off-line localisation. With po
or perception means, it is necessary to introduce a priori knowledge on sen
sors and environment models. To solve the localisation problem, the ultraso
nic image is segmented applying the Hough transform, well-adapted to ultras
onic sensor characteristics. The segments are then matched with the room, m
odelled and assumed to be rectangular. Several positions are found. A first
sort, based on a cost function, reduces the possibilities. The remaining a
mbiguities are removed by a neural network which plays the part of a classi
fier detecting the door in the environment. Improvements of the method are
proposed to take into account obstacles and nonrectangular room. Experiment
al results show that the localisation operates even with one obstacle.