Researchers at the University of California, Berkeley (UC Berkeley) have developed visual foresight, a learning technology that enables robots to imagine the future of their actions so they can determine how to manipulate objects they have never interacted with before.
Robots can use visual foresight to predict what their cameras will see if they perform a specific sequence of movements, and they can learn to perform these tasks without human assistance because the visual imagination is learned entirely from scratch from unattended, unsupervised exploration.
After the exploration phase, the robot develops a predictive model of the world, and it uses this model to manipulate new objects that it has not seen before.
The system is based on convolutional recurrent video prediction, or dynamic neural advection.
The goal of this research is to enable a robot to learn about how the world works via autonomous interaction, says UC Berkeley professor Sergey Levine.
From Berkeley News
View Full Article
Abstracts Copyright © 2017 Information Inc., Bethesda, Maryland, USA
No entries found