Modeling and model identification of autonomous underwater vehicles

Abstract

As autonomous underwater vehicles (AUVs) are deployed in more complex operational scenarios (e.g., multi-vehicle operations or information gathering in cluttered littoral zones), accurate control of these platforms is of particular importance. However, the design of accurate controllers and these complex systems in general require accurate models. This research is focused on the identification of rigid body and hydrodynamic modeling parameters of the THAUS (a modified SeaBotix vLBV300) and the Hydroid REMUS100 AUVs. A hydrodynamic model is adopted that accounts for vehicle-specific properties, including symmetry and anticipated flow properties. An experimental setup, based on a quadrifilar pendulum, is developed to measure the moments of inertia of the vehicle. System identification techniques, based on Recursive Least Squares estimation with modifications for learning the parameters of dynamic systems, are applied in two approaches to learn the parametric models of the platforms: an individual channel excitation approach and a free decay pendulum test. The former is applied to THAUS, which can excite the system in individual channels in four degrees of freedom. These results are verified in the free decay pendulum setup, which has the advantage that the approach is independent of the platform actuation. The latter test is also applied to the REMUS AUV.