Related Experiment Video
Updated: May 14, 2025
00:08
A Random-displacement Measurement by Combining a Magnetic Scale and Two Fiber Bragg Gratings
Published on: September 30, 2019
6.2K
Length Estimation of Pneumatic Artificial Muscle with Optical Fiber Sensor Using Machine Learning
Yilei Ni1, Shuichi Wakimoto1, Weihang Tian1
1Graduate School of Environmental, Life, Natural Science and Technology, Okayama University, Okayama 700-8530, Japan.
Sensors (Basel, Switzerland)
|April 12, 2025
Summary
Researchers developed a smart artificial muscle with integrated optical fibers for motion sensing. Machine learning, specifically Long Short-Term Memory networks, significantly improved length estimation accuracy by accounting for complex sensor dependencies.
Area of Science:
- Robotics
- Materials Science
- Sensor Technology
Background:
- McKibben artificial muscles are flexible, lightweight soft actuators.
- Integrating optical fibers enables motion sensing in these actuators.
- Traditional methods struggle with sensor dependency on load and pressure.
Purpose of the Study:
- To develop a smart artificial muscle capable of accurate motion sensing.
- To overcome limitations in optical fiber sensing due to load and pressure variations.
- To enhance length measurement accuracy using advanced algorithms.
Main Methods:
- Integrated an optical fiber into the McKibben artificial muscle sleeve.
- Measured macrobending and microbending loss in the optical fiber.
- Employed a machine learning model, primarily Long Short-Term Memory (LSTM) neural networks, for length estimation.
Main Results:
- Optical fiber sensor characteristics were dependent on length, load, and air pressure.
- Machine learning effectively compensated for these complex dependencies.
- LSTM-based length estimation demonstrated significantly reduced error compared to traditional methods.
Conclusions:
- Machine learning offers a feasible and effective approach to enhance artificial muscle length measurement accuracy.
- Smart artificial muscles with integrated sensors show promise for advanced robotic applications.
- LSTM networks provide a robust solution for complex, multi-variable sensor data interpretation.
