Local area navigation using sonar feature extraction and model based predictive control

Abstract

This paper demonstrates a method to navigate an autonomous underwater vehicle in a local area using an acoustic sensor for position information derived from feature detection. A dynamic model of the vehicle response is used for control between location updates. Favorable results have been found using this approach, and precision positioning of the vehicle to centimeters has been accomplished. Another part of the problem is the need to provide a high frequency update of vehicle position in order to close the positioning servo loops. Using sonar image feature extraction is necessarily time consuming and therefore is performed in a tactical level process providing asynchronous data to the tactical navigator. It follows that some form of fast dead reckoning must be performed in the execution level either by INS/Doppler or by water speed and heading data or by a model based predictor. This paper deals with the use of a model based predictor technique where knowledge of the dynamic model of the vehicle provides state information to the vehicle positioning control function. The model uncertainty provides errors of course, but these are corrected through asynchronous updates from the feature extracted positions in the tactical level sonar manager. This concept has been verified by both simulation studies and by experimental validation with the NPS Phoenix vehicle.