Three-dimensional neural network tracking control of a moving target by underactuated autonomous underwater vehicles

This paper investigates three-dimensional target tracking control problem of underactuated autonomous underwater vehicles (AUVs) by using coordinates transformation and multi-layer neural networks. The passive-boundedness assumption of sway and heave velocities of underactuated AUVs is used to design a controller in the actuated directions. For this purpose, a new Euler–Lagrange formulation is proposed based on range and bearing tracking errors with respect to a moving target in the body-fixed frame. Then, a tracking controller is proposed to make range and bearing tracking errors converge to zero. Multi-layer neural networks (MLNNs) are utilized to approximate unknown nonlinear-in-parameter dynamics of the system, and adaptive robust control techniques are adopted to compensate for MLNN approximation errors and time-varying environmental disturbances which are induced by waves, wind and ocean currents. The stability of the proposed control system is analysed based on Lyapunov’s approach which shows that target tracking errors are semi-globally uniformly ultimately bounded and exponentially tend to a small neighbourhood around the zero. At the end, simulation examples are given to demonstrate the competency of the proposed target tracking controller.