Mobile Robot Navigation Through An Unknown Environment Towards A Predefined Target

Abstract

This thesis project describes an implementation of a Human Computer Interaction(HCI) system for intelligent mobile robots. It aims to provide the ability to indoormobile robots to understand and execute direction orders described in a pseudo-humanlanguage. The robot is positioned in an unknown indoor environment and it is givendirections like "Go to the third office on your right".To achieve that, the robot should be able to create a map of its environment andlocalise itself in the map. Thus, Simultaneous Localisation And Mapping (SLAM) tech-niques are used. Two key novel challenges are confronted. Firstly, the navigation towardsa target in a yet unknown map, and secondly, how informative and useful the map is.The main objective of existing navigation techniques, is to drive the robot towardsthe unknown area of the map in order to minimise the yet unseen environment. That isnot enough if the robot must move towards a predefined target. In order to achieve thiskind of behaviour, an existing navigation technique is augmented.Secondly, the robot should be able to recognise entities that signify the way-pointsgiven by the human. To cope with that, a map of entities like door, chair, box, etc.is created and provided to to the robot. To identify these entities, pattern recognitiontechniques may be employed. The result is a map of objects that can be used not onlyfor navigation and path planning, but also for a better illustration of the actual map.