A robust neural network controller for a TITO interactive nonlinear system

This paper presents a robust NN control scheme for diving behavior of an autonomous underwater vehicle (AUV) whose dynamics can be simplified as a second-order TITO (two-input-two-output) nonlinear function. Because of singularity problem, above dynamics can't be properly solved using general backstepping method although it is in a well known strict-feedback form. Furthermore, the dynamics is in an interactive form so the traditional noninteracting control methods also can not be directly applied. In this paper, the value of one of two virtual inputs is derived from predefined vehicle's desired trajectory instead of stability point of view so the singularity problem can be avoided. Proposed scheme can guarantee all of the signals in the closed-loop system are semi-global uniformly ultimately bounded (SGUUB).