The robot localisation problem

Abstract

The interest in the use of robots to perform specific and routine operations is increasing at a rapid rate. If robots are able to move purposefully within an environment, their application domain can be greatly increased. In fact, navigation is a basic ability that mobile robots must possess. In order to navigate successfully within an environment, a robot must first be able to self-localise. In other words, the robot must possess a representation of the environment and be able to identify its current location within that representation. The Robot Localisation Problem is concerned with the problem of self-localisation using only the sensory perception of the robot of its environment. This problem has received considerable research attention. The goal of this project is to address the Robot Localisation Problem from a computational and algorithmic point of view. In this dissertation a number of existing techniques which provide a solution to the Robot Localisation Problem are investigated. A technique which solves this problem is also proposed. This technique does not utilise sophisticated data structures and processes, and it does not require complex pre-processing steps. However, the results obtained show that the system is efficient for practical usage. A fundamental sub problem of the Robot Localisation Problem concerns the computation of locations within the environment from where the robot can obtain sensory information which is useful to infer its location. The system developed as part of this project acts as a framework to allow the evaluation of techniques which solve this sub problem. A simple solution for this sub problem is also proposed. The results obtained show that it is also effective.