Robust Decoupled Motion and Stiffness Control for a Class of Variable-Stiffness Soft Manipulators

Maja Trumic, Franco Angelini, Kosta Jovanovic, Adriano Fagiolini

Keywords: soft robotics, variable stiffness actuation, adaptive control, motion control, robustness

Abstract: Variable-stiffness soft manipulators can regulate both motion and compliance, which is essential for robust contact-rich manipulation. However, model-based control remains challenging because motion and stiffness are strongly coupled, the dynamics are highly nonlinear, and tracking performance is sensitive to model uncertainty. We present a nonlinear adaptive cascade controller for electromechanical antagonistic variable-stiffness actuators that achieves closed-loop and decoupled tracking of link motion and joint stiffness. The key idea is to decouple the motion and stiffness channels through a perturbed actuation matrix that remains invertible, and to recover exact stiffness regulation through integral action. Robustness to parametric uncertainty is obtained through online adaptation, while a dead-zone improves tolerance to unmodeled effects such as hysteresis, disturbances, and sensor noise. We validate the method in simulation under 30\% parameter error and experimentally on articulated soft robots. Results show accurate motion and stiffness tracking, clear decoupling, and strong robustness, including nearly tenfold lower position error than computed torque control on hardware. This work has been recently published in Transactions on Control Systems Technology.

Submission Number: 38