Louis Thorpe-Monaghan Undergraduate Dissertation 2016/17

Mobile Robot Navigation

Supervised by R.Gaizauskas

Abstract

The ability to navigate in an environment is a key problem to solve when considering robots that are capable of autonomous behaviour . A hu man can solve similar problems, such as finding one's location on a map, by using knowledge of where they've been and what they've seen to determine their location. This level of reasoning can be expensive and difficult for a machine. Robot self-localisation shares a lot of these same ideas, such as mapping and landmark recognition. The main difference comes from the use of a probabilistic framework, based around random samples of environment locations, in order for a robot to estimate its whereabouts. The aims of this project are to understand and implement Monte Carlo Localisation in Python, and perform extensive evaluation of its performance on virtual robots. These were met with significant success, and the result is a version of MCL that performs well across different environments.