LS-SVM Based Modeling and Model Predictive Control for a Water-Hydraulic Artificial Muscle Actuator

Artificial muscle actuators (AMAs) driven by the pressure of tap water are flexible and lightweight and are therefore safe for humans and suitable for wearable power-assist devices. This paper proposes a modeling approach for a water-hydraulic AMA based on least squares support vector machines (LS-SVMs). Modeling tests are carried out on this LS-SVM approach with experimentally acquired data and show that the proposed model can capture the hysteretic characteristics of the water-hydraulic AMA. In addition, we constructed a state-space model of the water-hydraulic AMA into which the proposed LS-SVM-based model was incorporated and propose a model predictive control system based on the state-space model. An experimental comparison of the proposed control system and a control system with an inverse model of the water-hydraulic AMA, used in our previous study, demonstrated that the proposed one exceeds the previous one in control performance. © Taylor & Francis Group, LLC 2020.