Here are the key points from the Hacker News comments, summarized in a neutral and informative tone:
The general idea of making materials more "self-aware" or inspectable is appreciated as a neat and sci-fi-like concept.
One commenter's previous research used complex-valued neural nets to learn the transmission matrix of an optical fiber, which relates to this work as bending the fiber requires relearning a new matrix. It may be possible to learn a parameterized characterization of the fiber to model its shape.
The research seems thorough and well-documented. The advisor, Cindy Harnett, is likely aware of the conceptual similarity to time-domain reflectometry.
There are questions about how the sensor handles multiple bends. It seems that with the current setup, multiple bends would result in the sum of the log attenuations, potentially requiring as many strands as bend locations to disambiguate. Clarification is needed on whether this is intended for single-bend cases only.
Potential refinements could make this technology mass-producible with high accuracy. Applications include cost-effective 2D or 3D touch sensors for robots, directionally sensitive proprioception for flexible tubes, and detecting localized temperature differences.
The technology seems similar to the Nintendo Power Glove, which used light down a tube to trigger key presses based on finger/hand bending.
Having an awesome advisor is highlighted as the difference between a miserable and a good experience.
One commenter is interested in using this sensor (or multiple sensors together) to accurately detect a golf club swing for a golf launch monitor that doesn't require hitting a ball.
Suggestions for improvement include:
Reducing inter-sensor distance to fix "dead zones" in the sensor data
Dividing the tube into three sections to sense the direction of the bend
Finding solutions to sense multiple bends, as the current setup would result in invalid readings
The work reminds another commenter of distributed acoustic sensing (DAS) using fiber optic cables for various monitoring tasks, though they haven't seen it used for soft robotics before.
The manufacturing challenge of requiring log2 fibers and different encodings in each junction is acknowledged, but not considered a problem at the research/proof-of-concept phase.
Hacker News 의견
Here are the key points from the Hacker News comments, summarized in a neutral and informative tone:
The general idea of making materials more "self-aware" or inspectable is appreciated as a neat and sci-fi-like concept.
One commenter's previous research used complex-valued neural nets to learn the transmission matrix of an optical fiber, which relates to this work as bending the fiber requires relearning a new matrix. It may be possible to learn a parameterized characterization of the fiber to model its shape.
The research seems thorough and well-documented. The advisor, Cindy Harnett, is likely aware of the conceptual similarity to time-domain reflectometry.
There are questions about how the sensor handles multiple bends. It seems that with the current setup, multiple bends would result in the sum of the log attenuations, potentially requiring as many strands as bend locations to disambiguate. Clarification is needed on whether this is intended for single-bend cases only.
Potential refinements could make this technology mass-producible with high accuracy. Applications include cost-effective 2D or 3D touch sensors for robots, directionally sensitive proprioception for flexible tubes, and detecting localized temperature differences.
The technology seems similar to the Nintendo Power Glove, which used light down a tube to trigger key presses based on finger/hand bending.
Having an awesome advisor is highlighted as the difference between a miserable and a good experience.
One commenter is interested in using this sensor (or multiple sensors together) to accurately detect a golf club swing for a golf launch monitor that doesn't require hitting a ball.
Suggestions for improvement include:
The work reminds another commenter of distributed acoustic sensing (DAS) using fiber optic cables for various monitoring tasks, though they haven't seen it used for soft robotics before.
The manufacturing challenge of requiring log2 fibers and different encodings in each junction is acknowledged, but not considered a problem at the research/proof-of-concept phase.