Using backward adjustment with model predictive control for adaptive control of nonlinear soft artificial muscle

Soft artificial muscles possess inherent compliance and safety features, rendering them highly suitable for applications in wearable robots and unstructured environments. However, accurately modeling the nonlinearity of soft actuators proves to be a challenging task. In this paper, we present an adaptive control method that leverages model learning and model parameter backward adjustment. Our approach focuses on updating the dynamic model of the artificial muscles in two ways: by refining the input-output relation and by addressing prediction and control errors. To achieve this, we utilize tracking performance as a posterior evaluation metric for model parameter adjustment. Through a series of experiments, we demonstrate that our controller is capable of achieving reference tracking with a root mean square error (RMSE) of less than 5% across different stiffness levels. These experimental results validate the effectiveness of our proposed method in capturing the nonlinearity of soft artificial muscles, adapting to varying loads, and achieving precise reference tracking.