The hysteresis modeling of pneumatic artificial muscles using least squares support vector machine approach

Abstract

Reducing the modeling errors of hysteresis model is of great significance for improving the control accuracy of pneumatic artificial muscles. However, the asymmetry and complexity of its hysteresis loops limit the effect of existing modeling approach, especially for the irregular loops. This article extends the least squares support vector machine approach to the domain of asymmetric and irregular hysteresis characterization for pneumatic artificial muscles. Compared with the established hysteresis models, the significance of this approach is that it does not depend on the shapes of hysteresis loops and possess the advantages of both few identification parameters and high accuracy. The effectiveness and advantage of the presented model is compared with modified symmetric generalized Prandtl–Ishlinskii model. In addition, the length–pressure hysteresis experiment and modeling comparison on both of Festo commercial and self-made pneumatic artificial muscles are presented. The modeling errors of least squares support vector machine model for both loops are suppressed to a negligible level, and are not affected by the shape and complexity of hysteresis loops, which validates the effectiveness of this approach.