Neural-Net based Robust Adaptive Control for 3D Path Following of Torpedo-type AUVs

This paper considers the 3D path following problem for a class of torpedo-type AUVs (autonomous underwater vehicles), where only three control inputs (surge force, pitch and yaw moments) are available for its 6DoF (degree-of-freedom) motion in the water. For this typical underactuated system, two spherical coordinate transformations are introduced in this paper so as to transform the vehicle's path following model into certain three-input-three-output 2nd-order strict-feedback form. On the other hand, for underwater vehicles, due to their complicate and high nonlinear hydrodynamics, it is difficult to extract all of the exact hydrodynamic coefficients using conventional estimation methods such as CFD (Computational Fluid Dynamics) or PMM (Planar Motion Mechanism) tests. Moreover, considerable noises are unavoidable in the practical sensor measurements. For these reasons, in this paper, a sort of neural-net based on-line estimation method combined with robust scheme is applied to solve the above three-input-three-output strict-feedback form of path following problem. Proposed robust adaptive scheme can guarantee the uniform ultimate boundedness (UUB) of closed-loop system in terms of the spherical coordinates.