Modeling of Soft Pneumatic Actuators with Different Orientation Angles Using Echo State Networks for Irregular Time...

Area of Science:

• Robotics
• Machine Learning
• Materials Science

Background:

• Modeling soft robotics is challenging due to large material deformations.
• Accurate kinematic models are essential for controlling soft robots.
• Soft Pneumatic Actuators (SPAs) are a key component in soft robotics.

Purpose of the Study:

• To develop a data-driven machine learning model for predicting the kinematics of SPAs.
• To utilize an Echo State Network (ESN) for modeling SPA tip position in 3D space.
• To demonstrate a method for modeling SPAs without relying on feedback sensor data.

Main Methods:

• Collected experimental data from 3D printed SPAs under varying pressure inputs.
• Trained an Echo State Network (ESN) on the collected data to predict SPA tip position.
• Compared the ESN model against a Long Short-Term Memory (LSTM) network using experimental and Finite Element Analysis (FEA) data.

Main Results:

• The ESN successfully modeled the complex, non-linear behavior of SPAs using only control input.
• The model accurately predicted SPA kinematics across different orientation angles (θ).
• Both ESN and LSTM models showed effectiveness when tested on unseen FEA data.

Conclusions:

• The data-driven ESN approach provides a reliable method for modeling SPA kinematics.
• This methodology offers a generalizable strategy for modeling SPAs with diverse design parameters.
• The study highlights the potential of machine learning in advancing soft robotics control.