A high-level perception architecture: real-time visual navigation for autonomous robots in structured environments

Abstract

This thesis describes the design and implementation of a mobile robot which is able to perform a structural interpretation of indoor environments, using only visual and proprioceptive sensory information. The desired behaviour is real-time navigation based on this interpretation, instead of a reactive approach. The design is guided by a predictive criterion: the system must anticipate the consequences of its actions, showing a certain predictive understanding of the scene in which it moves.