In this paper we propose a sonar based navigation system capable of detecting and avoiding collision with a moving obstacle. The main feature of the system IS the use of a Kalman filter for computing the presumed trajectory of the obstacle. Simulations show the efficiency of the system in general situations.

1. Introduction

In the coming years, the need for semi and fully autonomous underwater vehicles (AUVs) will become more and more evident. As the offshore oil and gas explorations move into increasingly deeper water, the conventional methods of inspecting underwater installations or doing underwater surveys (such as divers, remotely operated tools (ROTs) and remotely operated vehicles (ROVs» are no longer sufficient due to safety problems as well as to cost limitation (Hallset-R0dseth,1991; A.A., 1987). To enhance the autonomy of an underwater vehicle one needs to provide it with the capability of operating in an unstructured environment with little a priori information. Quite obviously, two relevant tasks in the first step toward autonomy are localisation and obstacle avoidance. By localisation we mean the process of tracking known objects (or beacons) to determine the vehicle position, while by obstacle avoidance we mean the action of detecting and avoiding collisions with unexpected fixed or moving objects. In this paper we are interested in tackling this second problem and in our approach we look for a simple and real-time solution. In developing a decisional scheme for guidance and obstacle avoidance, we have in mind a navigation system strucctured as that of the autonomous mobile vehicles Dolphin (Elves, 1987). More precisely, we assume that the control action is performed at three different levels, called respectively the sensory level, the geometric level and the symbolic level. The complexity of elaboration increases from the first level to the last one, as briefly described here on.

This content is only available via PDF.
You can access this article if you purchase or spend a download.