Joint Stiffness Estimation in a Single-Link Soft Robot Driven by Pneumatic McKibben Muscles

Abstract

The possibility to control not only the position of a robot but also the stiffness of its joints has enabled safe human-robot collaboration and nature-like robot behaviour. However, since stiffness is not a measurable variable, one needs to perform either its extensive offline identification or apply the real-time stiffness estimators. To the best of the authors' knowledge, this paper proposes for the first time an online technique for the estimation of stiffness and elastic torque in an articulated soft robot joint driven by McKibben pneumatic artificial muscles. We address this problem in a two-phase process: first, we reconstruct the elastic torque by leveraging the theory of the delayed Unknown Input Observers, and then a Recursive Least Squares algorithm is used to determine the parameters of a stiffness approximation. Besides robot link dynamics, this approach requires information on link position and commanded pressures to the muscles. The solution is validated on the simulated single-link robot driven by a pair of McKibben muscles in an antagonistic setup.