LONG–TERM AUTONOMY OF MOBILE ROBOTS IN CHANGING ENVIRONMENTS

Abstract

This habilitation thesis presents research that aims to enable long-term deployment of
\nmobile robots in changing environments. The presented approaches encompass methods
\nthat ensure robustness of autonomous visual navigation in outdoor environments for prolonged
\ntime periods, spatio-temporal representations that explicitly model the environment
\nchanges over time, and supporting software modules that enable robust and accurate robot
\nlocalisation.
\nThe main contribution of the thesis is a novel approach that allows to incorporate
\nthe notion of time into most stationary environment models used in mobile robotics.
\nThis is achieved by representing the uncertainty of the environment states not by fixed
\nprobabilities, but by probabilistic functions of time, represented in the frequency domain.
\nThe method allows to integrate unlimited numbers of sparse and irregular observations
\nobtained during long-term deployments of mobile robots into memory-efficient models that
\nreflect the persistence and recurrence of environment variations. The frequency-enhanced
\nspatio-temporal models allow to predict the future environment states, which improves the
\nefficiency of mobile robot operation in changing environments. In this thesis, we present a
\nseries of articles, which demonstrate that the proposed approach improves mobile robot
\nlocalization, path and task planning, activity recognition, human-robot interaction and
\nallows for life-long spatio-temporal exploration of perpetually-changing environments.