Neural network for 3D trajectory tracking control of a CMG-actuated underwater vehicle with input saturation

This paper considers the trajectory tracking problem of an underactuated underwater vehicle actuated by control moment gyros (CMGs) in three-dimensional (3D) space, with the constraints from input saturation, partial parameter uncertainty and unknown external disturbance. First, utilizing a physical translation of the motion equations, the overall system can be decomposed to an input decoupling system. Then a modified virtual velocity guidance law is derived to transform the tracking error signals into the controllable velocity signals. Subsequently, the Gaussian error function is employed to update the common saturation model. To avoid complex derivations of the virtual control signals, first-order sliding mode differentiator is explored in the dynamic control layer. Then, the adaptive neural network (NN) control method is introduced into the backstepping procedure to account for nonlinear uncertainties and bounded disturbances. Among this, the constrained steering law is used to steer the CMG system to avoid its inherent singularity and fulfill the global tracking control. It is proved that the proposed controller can guarantee all closed-loop signals converge to a small neighborhood of the origin. Finally, two case studies are presented to illustrate the tracking performance of the proposed design.