This ShadowSense system uses a normal, off-the-shelf USB web-cam to capture the shadows that are produced by hand gestures on a robot’s skin. Then, there are algorithms that classify the movements to figure out the user’s specific interaction.
The study’s lead author, Guy Hoffman, said that their method provides a more natural way of interacting with a robot. Another benefit is that this method does not rely on large, expensive arrays of sensors.
Related post: Dominion Drones/SpiderMonkey FPV, YouTube
Touch is an extremely important part of communication for nearly all organisms. It has, however, been almost completely absent when it comes to human-to-robot interactions. One reason for this is that older methods of full-body touch required too many sensors. This has historically made this technology impractical to implement. With this research, we finally get a low-cost alternative.
The researchers hooked the system up to an inflatable robot that had a camera underneath its skin.
They trained and AI classification algorithms with shadow images composed of six gestures. Those include touching with a palm, punching, touching with two hands, hugging, pointing, and not touching.
Depending on the lighting conditions, the system was able to successfully distinguished between the gestures accurately between 87.5% to 96% of the time
The sort of research is paving the way for mobile guide robots that can respond to different gestures, like turning to face a person when it detects someone poke them, or something like moving back when it feels a tap on the back.
Also, interactive touch screens could be added to inflatable robots that could make home assistant droids more privacy-friendly.
If the robot can only see you in the form of your shadow, it can detect what you’re doing without taking high fidelity images of your appearance. That gives you a physical filter and protection, and provides psychological comfort.HOFFMAN
If you are interested in the full paper, you can check it out here.