Probablistic approaches for intelligent AUV localisation
Abstract
This thesis studies the problem of intelligent localisation for an autonomous underwater
vehicle (AUV). After an introduction about robot localisation and specific
issues in the underwater domain, the thesis will focus on passive techniques for AUV
localisation, highlighting experimental results and comparison among different techniques.
Then, it will develop active techniques, which require intelligent decisions
about the steps to undertake in order for the AUV to localise itself. The undertaken
methodology consisted in three stages: theoretical analysis of the problem, tests with
a simulation environment, integration in the robot architecture and field trials. The
conclusions highlight applications and scenarios where the developed techniques have
been successfully used or can be potentially used to enhance the results given by current
techniques. The main contribution of this thesis is in the proposal of an active
localisation module, which is able to determine the best set of action to be executed,
in order to maximise the localisation results, in terms of time and efficiency.