Impedance estimation of a pneumatic muscle as a mechanical transmission and actuation device

Abstract

Human friendly robots and devices like exoskeletons, active prostheses etc. are required to be compliant in their actuation system; here the transmissions are deliberately made flexible. This flexibility demands variability in order to improve upon the transmission bandwidth, as well as meeting specific requirements in task execution. Pneumatic Muscle Actuator (PMA) has an inherent ability to vary impedance of transmission in its actuation with varying pressure, resembling a biological muscle, when implemented in agonist-antagonistic arrangement. Regulation and control for stiffness/impedance requires knowledge of the impedance values; however, there is no transducer available to measure the impedance components. In this paper the issue of online estimation of impedance components of a pneumatic muscle actuator is addressed in terms of effective inertia, damping rate and stiffness of the elastic muscle. Devising a model free estimator is indeed difficult, especially in steady state. The present approach considers a physical model of the Pneumatic Artificial Muscle (PAM), suitable for practical implementation and in the same time detailed enough representing, as far as possible, all important behaviors (relating muscle force with displacement/contraction, muscle velocity and muscle pressure). Sensors with noise and varying behavior of passive components with time (and environment) can provide only approximate calibration with inconsistent and not-so-stable results. Online estimation becomes necessary here. This article proposes a first order Extended Kalman Filtering technique to estimate online the impedance parameters. Experimental results are presented to validate the proposed estimation algorithm.