Towards an efficient inverse static model of a Festo actuator made of two antagonist muscles for hybrid control of its position and stiffness

04/27/2021 ∙ by Bertrand Tondu, et al. ∙ 0

The Festo air muscle is today one of the most known commercial version of the so-called McKibben pneumatic artificial muscle. A major advantage of handmade McKibben muscles, as well as its commercial versions, lies in the possibility it offers of realizing antagonist muscle actuator on the model of the biceps-triceps system. If pressures are independently controlled in each artificial muscle, it is then possible to define a position-stiffness control of the antagonist actuator by analogy with natural neural control of antagonist skeletal muscles. Such a control however requires a knowledge model of the actuator making possible a stiffness estimation provided by control pressures, while position closed-loop control is facilitated by a feedforward model of this highly nonlinear actuation device. We discuss this issue in the particular case of the antagonist Festo air muscle actuator, and we propose a simplified static actuator model derived from the classic static theoretical model of the McKibben artificial muscle, simple enough for a future MIMO position-stiffness controller be able to integrate it. If the proposed model highlights the ability of the antagonist Festo muscle actuator to mimic the stiffness neural control by the mean of the sum of pressures inside artificial muscles, it also highlights the difficulty to derive a simple but accurate model of the static force produced by Festo muscle in the full control range of pressure. The reported results suggest that it would be particularly interesting to derive advantage from numerous studies about Festo muscle modelling to go further in order to find such an inverse actuator model including an accurate estimation of the actuator stiffness.

READ FULL TEXT
POST COMMENT

Comments

There are no comments yet.

Authors

page 1

page 2

page 3

page 4

This week in AI

Get the week's most popular data science and artificial intelligence research sent straight to your inbox every Saturday.