Yuchen Liang
The field of artificial muscles has progressed significantly in developing robotic systems that mimic the human musculoskeletal system. Pneumatic artificial muscles (PAMs) are valued for their lightweight and flexible properties, but traditional PAMs often involve complex fabrication and multiple components, complicating miniaturization and design.
To address these issues, the "GeometRy-based Actuators that Contract and Elongate" (GRACE) series has been created. GRACEs utilize a single-material pleated membrane for contraction and elongation without additional strain-limiting elements, enabling diverse lifelike movements and allowing for cost-effective 3D printing of functional devices.
However, controlling PAMs remains challenging due to inadequate sensing mechanisms. Conventional sensors often fail to provide the real-time feedback needed for precise control in robotics. This thesis presents a novel sensing approach for GRACE artificial muscles, integrating fiber optics and machine learning. By exploiting the macro bending loss of optical fibers, we developed a sensing mechanism that delivers accurate, real-time feedback on actuator performance. Our findings demonstrate a data-driven setup that predicts force and length changes in GRACE actuators, enhancing their controllability and potential applications in robotics, with the possibility of extending this approach to other soft actuators.
[More information will be disclosed after the submission of the relevant journal]