Active global localisation for a mobile robot using multiple hypothesis tracking

Abstract

In this paper we present a probabilistic approach for mobile robot localisation using an incomplete topological world model. The method uses multi--hypothesis Kalman filter based pose tracking combined with a probabilistic formulation of hypothesis correctness to on--line generate and track Gaussian pose hypotheses. Apart from a lower computational complexity, this has the advantage over traditional grid based methods that incomplete and topological world model information can be utilised. Furthermore, the method generates movement commands for the platform in order to optimise the information gathering for the pose estimation process. 1 Introduction The problem of global localisation---or "the kidnapped robot problem"---is that of, from little or no a priori pose 1 information, estimating the correct pose of a robot with respect to some global reference frame. This is fundamentally different from the pose tracking problem which is that of keeping the robot's pose est...