Backstepping-Sliding-Mode-Based Neural Network Control for Electro-Hydraulic Actuator Subject to Completely Unknown System Dynamics

Requirements of the system dynamics and unknown disturbance and uncertainty suppression bring certain challenges in developing advanced control algorithms for electro-hydraulic actuators (EHAs). Usually, radial basis function neural network (RBFNN) is employed to address unknown nonlinear dynamics; however, traditional approaches require more effort in adopting multi-estimated parameters when designing adaptive laws, especially in the case of completely unknown system dynamics. Hence, this paper proposes a novel adaptive neural network control, based on the backstepping control (BSC) framework, to compensate for unknown system dynamics and also to reduce the problem of the multi-estimated parameters based on the norm estimation technique. To facilitate the proposed control implementation, a new system transformation is first expressed in such a way that the same key properties as the original system are upheld. Besides, extended state observers (ESOs) are employed to fundamentally address the redundant remaining dynamics of unknown terms, disturbance, and uncertainty. In addition, the command filter (CF) technique is also involved to deal with the explosion complexity in the BSC design. The stability of the closed-loop system and the effectiveness of the proposed controller are theoretically guaranteed via mathematical proof with comparative experiments.Note to Practitioners—This article was motivated by the shortcomings of the published works in addressing the problem of completely unknown system dynamics for Electro-hydraulic Actuators subject to disturbances and unstructured uncertainties. To overcome these concerns, an approximation technique for the dynamical behavior compensation and extended state observers for the disturbances and uncertainties suppression have been carried out in such a new way that reduces a number of estimated parameters compared to the conventional approximation mechanism. Moreover, this new-way approximation also facilitates a combination of the command-filter technique to cope with the complexity explosion issue, which always exists when employing the backstepping control scheme to guarantee the closed-loop system stability. From the derived control strategy and with good experimental results, this proposed method can be considered a premise to expand to other topics in the field of automatically controlled systems and contributes to broad interest in both academic and industrial applications of system modeling and dealing with unexpected impacts.