Real-Time Whole-Body Contact Estimation of Continuum Robots Using Multiplexed Fibers for Embodied Actuation and...

Area of Science:

• Robotics
• Medical Devices
• Machine Intelligence

Background:

• Millimeter cable-driven continuum robots offer shape conforming and dexterous manipulation for confined environments like endoluminal interventions.
• Accurate force interaction quantification is crucial for shape adjustment and preventing damage to delicate luminal structures during manipulation.

Purpose of the Study:

• To develop a real-time, whole-body contact estimation framework for small-scale continuum robots.
• To enable safer and more precise manipulation in constrained environments through advanced sensing capabilities.

Main Methods:

• A framework combining actuation fibers and model-informed neural networks was developed, treating contact estimation as an inverse problem.
• Rod theory formulated the physical relationship between external contact, internal actuation, and shape sensing.
• Generative adversarial networks (GANs) were used for data augmentation, and an automatic data acquisition platform was created to minimize real-world data requirements.

Main Results:

• The framework achieved high accuracy in estimating 3D contact position (1.7 mm error) and contact force magnitude (8.7 mN error).
• Real-time estimation was demonstrated at a frequency of 25 Hz.
• Experiments on notched continuum robots validated the general applicability and accuracy of the approach.

Conclusions:

• The proposed framework enables precise, real-time force estimation for millimeter continuum robots.
• This technology facilitates safer operations in confined spaces by providing crucial sensory feedback.
• It paves the way for integrating simultaneous actuation and sensing using multiplexed fibers in continuum robots.