In tests, the robot was able to successfully peel a banana 57% of the time. The whole process takes less than 3 minutes.
Credit: Heecheol Kim/University of Tokyo
A machine learning system developed by researchers at Japan's University of Tokyo can train a robot to peel a banana without crushing it.
The robot was trained on about 13 hours of data in which a human demonstrator peeled hundreds of bananas.
The machine learning model maps out a trajectory that involves copying the human when it comes to broad movements not likely to damage the fruit.
The system shifts to a reactive approach, in which it responds to changes in its environment, for more precise movements.
The robot successfully peeled a banana 57% of the time, with the process taking fewer than three minutes.
From New Scientist
View Full Article - May Require Paid Subscription
Abstracts Copyright © 2022 SmithBucklin, Washington, DC, USA
No entries found