Adaptive Robust Multi‐Layer Neural Network Feedback‐Linearizing Fixed‐Time Control of Six‐Degrees‐of‐Freedom Underactuated Autonomous Underwater Vehicles

ABSTRACT

This paper presents a novel adaptive robust multi‐layer neural network feedback‐linearizing fixed‐time controller for six‐degrees‐of‐freedom (6‐DOF) underactuated autonomous underwater vehicles (AUVs) subjected to nonlinear dynamics, parametric uncertainties, and environmental disturbances which is rarely addressed in previous studies. The feedback linearization technique, prescribed performance control, multi‐layer neural network and the fixed‐time convergence are successfully integrated into one control package to propose a robust high‐performance trajectory tracking controller for the 6DOF AUVs. For this purpose, the vehicle's kinematics and dynamics are firstly combined with a new set of output position variables based on the look‐ahead control concept to obtain an input‐output dynamics of the system and, then, the input‐output feedback linearization is applied to transform the nonlinear system into a partially linearized form. Residual uncertainties and unmodeled dynamics are approximated online by using a multi‐layer neural network (MLNN), while a robust compensator ensures resilience against approximation errors and disturbances. To guarantee that the tracking errors remain within predefined bounds and converge within a fixed time, prescribed performance functions are incorporated into the control design. The stability of the closed‐loop control system is rigorously proven through a Lyapunov stability analysis, demonstrating boundedness of all the signals and satisfaction of performance guarantees. Simulation studies are conducted to evaluate the effectiveness of the proposed controller. The results clearly show superior tracking performance and robustness when compared with a conventional fixed‐time controller without prescribed performance constraints, confirming the advantages of the proposed control approach in terms of convergence speed, accuracy, and robustness.