Displacement Estimation of Twist String Actuator Using Neural Network Introducing Physical Model

Abstract

It is difficult to estimate the displacement of a twisted string actuator (TSA), which is a type of linear actuator, using only a mathematical model because the output is nonlinear due to the nature of the viscoelastic resin string. Therefore, in this study, the displacement of the TSA was estimated using a neural network in which the mathematical model of the TSA was introduced. The experimental results showed that the neural network using the mathematical model as cue information improved the accuracy of displacement estimation compared to the results obtained by estimating displacement using only the neural network.