Adaptive Sensing for Localisation of an Autonomous Underwater Vehicle

FULL JURNAL

This paper demonstrates an improved robot localisation approach based on adaptive sensing.
Using a particle ¯lter to represent the uncertainty in location the approach minimises the
expected entropy after the next robot action.It is demonstrated in simulation that this ap-
proach is useful in sensor management of an Autonomous Underwater Vehicle when a digi-
tised elevation map of the environment is available.

1 Introduction
The sophistication and usefulness of Autonomous Underwater Vehicles (AUVs) has increased dramatically in recent years, with record feats of depth and endurance offering new opportunities for ocean science and exploration. However the majority of these vehicles have only a primitive level of autonomy, with current AUVs typically carrying out a mission by attempting to follow a pre-planned trajectory. Sensing is usually passive, in that the perceived sensor data does not in°uence the actions taken by the vehicle.1 We propose to provide the
AUV with a means to make sensor management decisions based upon the information it obtains during the course of the mission, in order to enhance navigation performance. There are also opportunities to improve battery life [Nygren and Jansson, 2004] and the overall effectiveness of a mission [Makarenko et al., 2002] by sensing intelligently rather than continuously and indiscriminately. In this paper we demonstrate the potential of adaptive sensing by applying it to a terrain-aided tracking algorithm [Williams, 2003].