Model Predictive Displacement Control Tuning for Tap-Water-Driven Muscle By Inverse Optimization with Adaptive Model Matching and Analysis of Contribution

Abstract The tap-water-driven McKibben muscle has many advantages and is expected to be applied in mechanical systems that require a high degree of cleanliness. However, the muscle has strong asymmetric hysteresis characteristics that depend on the load, and these problems prevent its widespread use. In this study, a novel control method - model predictive control with servomechanism based on inverse optimization with adaptive model matching - is applied to the muscle based on a high-precision mathematical model employing an asymmetric Bouc-Wen model. The experimental results show that the proposed approach achieves a high tracking performance at a reference frequency of 0.3 Hz, with a mean absolute error of 0.13 mm in the steady-state response. Furthermore, an easier controller tuning can be achieved. Additionally, the authors evaluate the contributions of the elements of the proposed method. The results show that the contribution of the adaptive system is higher than that of the servo system. Furthermore, the effectiveness of adaptive model matching is reconfirmed.