Abstract
One of the problems in mobile robotics is the estimation of the robot position in the environment. In this paper we propose a model, called positioning model, for estimating a confidence interval of the robot position, in order to compare it with the estimation made by a dead-reckoning system. Both estimations are fused with heuristic rules. The positioning model is useful to estimate the robot position with or without previous knowledge of the previous position. Furthermore, it is possible to define the degree of previous knowledge of the robot position, allowing to make the estimation adaptive by varying this knowledge degree. This model is based on a one-pass neural network which could adapt itself in real time conditions and could learn the relationship between exteroceptive sensors measurements and the robot position.
