Adaptive Neural Network Finite-Time Tracking Control for Uncertain Hydraulic Manipulators

Motivated by the challenges associated with achieving rapid convergence and high accuracy in tracking errors for $\bm n$-DOF uncertain hydraulic manipulators, this article delves into an adaptive neural network (NN) finite-time control scheme. Initially, we propose an integrated finite-time dynamic surface control framework geared toward achieving swift convergence of tracking errors. Unlike conventional backstepping control schemes, this introduced dynamic surface control scheme adeptly circumvents computational complexity and singularity issues stemming from iterative derivatives of the virtual control input. Moreover, as accurate dynamic models of hydraulic manipulators are not readily available due to the substantial coupling between joints and the intricate nonlinearities, we introduce two adaptive NNs to encapsulate the complex coupled mechanical dynamics and uncertain hydraulic dynamics, respectively. The weight update laws for the adaptive NN are formulated utilizing convex optimization techniques and the gradient descent method, thereby accelerating the convergence of NNs compared to traditional weight update law construction methods. Finally, experimental studies conducted on a six-DOF hydraulic manipulator platform are presented to substantiate the efficacy of the proposed methodology.