Hysteresis Modeling of Twisted-Coiled Polymer Actuators Using Long Short Term Memory Networks

Abstract

Twisted-coiled polymer actuators (TCAs) are a new promising type of artificial muscles that is light-weight, low-cost and provides high stroke. This paper presents a dynamic model based on long short term memory (LSTM) networks to predict the nonlinear behavior of an antagonistic joint driven by the hybrid TCA bundle made from Spandex and nylon fibers. Different from our previous work, the current model considers pre-strains of TCAs as its inputs, therefore, any change in pre-strains of both TCA actuators will not require new data and training. It was verified from the experimental results that when there are pre-strain changes, the present model has better performance in capturing the system’s behavior compared with that of the previous one. The proposed LSTM based model can estimate the joint angle with the mean error of 0.06∘ compared with that of 1.57∘ using the previous one within the working range of 30 % of the TCA. © 2022, The Author(s), under exclusive license to Springer Nature Switzerland AG.