Robot Localization from Minimalist Inertial Data Using A Hidden Markov Model

Hidden Markov Models (HMMs) are applied to interoceptive data (in this case the sense of rotation by way of a gyroscope) acquired by a moving wheeled robot when contouring an indoor environment. We demonstrate the soundness of HMMs to solve the problem of robot localization in a topological model of the environment, particularly the kidnapped robot problem and position tracking. In this approach, the environment topology is described by the sequence of movements a robot executes when contouring the environment. Movements are described in a fuzzy domain using distance traveled and curvature as features.