Qualitative Geometrical Uncertainty in a Topological Robot Localization System

Despite the great advances in robotics made in recent years, there is not yet a universal model that allows a robot to move autonomously, safely and taking into account the most important elements of the surrounding environment. Autonomous and safe navigation requires that the robot knows its location and the location of obstacles and this is achieved through localization. Localization is the process of knowing and updating continuously a robot position with regards to its environment based on sensor information. The aim of this work is to develop a localization system that allows a mobile robot to know its location at each moment taking into account environmental uncertainty. In this paper, uncertainty estimation models are applied to a Topological Navigation algorithm using Hidden Markov Models (HMMs). HMMs offer a robust and efficient solution for the definition of discrete representations. HMMs are proposed as the basis to build an improved probabilistic algorithm using geometric information. Topological localization is used as it scales better to large environments and its navigation strategies are more natural. The main contribution is utilizing Hidden Markov Models and qualitative geometric information to infer robot localization in a topological map taking uncertainty into account. Finally, experiments have been carried out, their results show the feasibility of the localization system and the improvement due to the method proposed.