UKF-Based Navigation System for AUVs: Online Experimental Validation

Modern autonomous underwater vehicles (AUVs) are currently involved in complex tasks and scenarios, and require accurate and robust navigation systems to estimate their position. However, since the Global Positioning System cannot be exploited underwater, the AUV position is not directly measurable in real time (unless using dedicated acoustic-based sensors), making the availability of a reliable navigation system even more crucial. In this context, the main role is played by the filter used to estimate the AUV motion, usually relying on simple kinematic vehicle models and equations linearization. A navigation strategy specifically thought for AUVs and based on an unscented Kalman filter is proposed and experimentally validated by the authors. Preliminary tests of the developed strategy have been carried out by running the navigation filter on experimental data acquired during the FP7 European ARROWS project. This initial validation has been performed totally offline. The AUVs navigated in dead reckoning without using navigation filters, whereas the proposed strategy has been compared to standard extended Kalman filter based ones, highlighting encouraging performances. To further validate the proposed navigation system, suitable sea tests have been performed. The navigation filter has been implemented online on an AUV and the vehicle controller relied only on it to navigate. The new validation procedure, whose results are reported in this paper, showed again the good performance of the chosen strategy, yielding satisfying results in terms of accuracy of vehicle position estimation.