Sliding-Mode-Based Output Feedback Neural Network Control for Electro-Hydraulic Actuator Subject to Unknown Dynamics and Uncertainties

Unstructured dynamics, un-modeled parameters, and uncertainties in electro-hydraulic servo-valve-controlled actuators (EHSAs) bring difficulty in designing controllers for output tracking performance with stability and robustness satisfactions. Therefore, this article proposes an output feedback-based control for position regulation subject to fully unknown system behavior and uncertainties. With this idea, a coordinately transformed canonical system is utilized where all mismatched/matched uncertainties are lumped into one term with unknown dynamics. Then, a radial basis function neural network (RBFNN) with a norm of weighting vector estimation combined with a time-delayed estimation (TDE) is employed to effectively compensate for the system behavior. Accordingly, a second-order sliding-mode-based output feedback control is conducted to avoid following the step-by-step backstepping control (BSC) design. Interestingly, the proposed methodology requires only the measured output for the control law implementation with only one estimated variable for the system dynamics compensation due to using the hybrid RBF-based TDE (RBF-TDE). Moreover, to lower this approximated error, a modified sliding-mode-based nonlinear disturbance observer (DOB) is extensively involved. The closed-loop system stability is mathematically proven through the Lyapunov theorem with simulation and experiment on EHSA protocols to realize the effectiveness of the proposed algorithm.