Was that a long read? Let me explain it more simply…
What is the story
Researchers at the Massachusetts Institute of Technology (MIT) have unveiled a new robot that can perform kitchen tasks like peeling vegetables.
A recent demonstration video shows the robot deftly peeling a gourd, holding the vegetable in one hand and peeling it with the other.
This development is part of ongoing research into dexterous manipulation problems, specifically the reorientation of objects in the hand.
Robot system mimics human vegetable peeling
Researchers at MIT have designed a system that mimics the way humans peel vegetables, holding the vegetable in one hand and operating a peeler with the other.
“We propose a simple system for learning a reorientation controller that facilitates subsequent peeling tasks,” the researchers said.
The robotic system uses an Allegro hand mounted on a Franka robotic arm to orient the food for peeling, while another Franka robotic arm grabs the peeler.
Learning process and challenges
The Allegro hand reorientation controller is trained via reinforcement learning, and peeling is performed via teleoperation.
The researchers demonstrated the process using melon, sweet potato, and pumpkin.
The reorienting controller discussed in this study is a blind controller that does not utilize visual information but relies solely on proprioceptive sensory information.
It effectively redirects and holds heavy objects firmly, but performance could be improved with the addition of visual and tactile feedback.
The limits of MIT's robotic system
Current systems have some limitations, such as the risk that objects may slip out of the hand if the vegetable is small, or that the controller may fail, as fingers may not be able to make effective contact with small objects.
Why is peeling vegetables difficult for a robot?
Typically, people hold a vegetable in one hand and use a peeler in the other to peel the skin, then rotate the vegetable and peel again until no skin remains.
“These extra steps to achieve rotation are so easy for humans that we don't even think about them,” said Pulkit Agrawal, an assistant professor at MIT.
“But this is hard for a robot to do,” he added.
