New AI method lets robots detect human contact by analyzing shadows

Scientists from Cornell College have developed a manner for robots to determine bodily interactions simply by analyzing a person’s shadows.

Their ShadowSense system makes use of an off-the-shelf USB digicam to seize the shadows produced by hand gestures on a robotic’s pores and skin. Algorithms then classify the actions to deduce the person’s particular interplay.

Research lead writer Man Hoffman stated the strategy gives a pure manner of interacting with robots with out counting on giant and expensive sensor arrays:

Contact is such an necessary mode of communication for many organisms, however it has been just about absent from human-robot interplay. One of many causes is that full-body contact used to require an enormous variety of sensors, and was subsequently not sensible to implement. This analysis affords a low-cost different.

The researchers tried out the system on an inflatable robotic with a digicam beneath its pores and skin.

[Read: How Polestar is using blockchain to increase transparency]

They educated and examined the classification algorithms with shadow photographs of six gestures: touching with a palm, punching, touching with two fingers, hugging, pointing, and never touching.

It efficiently distinguished between the gestures with 87.5 − 96% accuracy, relying on the lighting.

Credit score: Hu et al.