Learning to sense three-dimensional shape deformation of a single multimode fiber

Area of Science:

• Photonics and Optical Sensing
• Machine Learning Applications
• Fiber Optic Technology

Background:

• Optical fiber sensors are crucial for healthcare, structural monitoring, and robotics.
• Existing sensors often involve complex structures and interrogation systems, limiting widespread adoption.

Purpose of the Study:

• To demonstrate that multimode optical fiber (MMF) output speckles inherently contain 3D geometric shape information.
• To develop a simplified and highly accurate deformation and shape-sensing system using a single MMF.

Main Methods:

• Utilizing the information-rich output speckles from a single multimode optical fiber.
• Applying a k-nearest neighbor (KNN) machine learning algorithm for data analysis and classification.
• Correlating speckle patterns with the fiber's 3D deformation and shape.

Main Results:

• Successfully demonstrated proof-of-concept 3D multi-point deformation sensing using a single MMF.
• Achieved classification accuracy close to 100% in identifying fiber shape and deformation.
• Highlighted the inherent sensitivity and resolution of the MMF-based sensing approach.

Conclusions:

• A single MMF can act as a simple, sensitive, and accurate sensor for deformation and shape monitoring.
• MMF speckle patterns provide rich data for 3D geometric sensing.
• This approach offers a promising, less complex alternative to existing optical fiber sensing technologies.