acm-header
Sign In

Communications of the ACM

ACM TechNews

Robot See, Robot Do: How Robots Can Learn New Tasks By Observing


View as: Print Mobile App Share:
A robot mixes a cocktail after watching a person perform the task.

Researchers at the University of Maryland are developing robots that can learn a task by watching someone perform it.

Credit: Universithy of Maryland

The University of Maryland's (UMD) Autonomy, Robotics, and Cognition Lab is developing robots that can learn how to do a new job by watching others do it first.

For example, the UMD researchers have developed a cocktail-making robot that watches a human mix a drink by pouring liquid from several bottles into a jug. The robot copies these actions, grasping bottles in the correct order before pouring the proper quantities into the jug.

The new approach involves training a computer system to associate specific robot actions with video footage showing humans performing various tasks.

The researchers also developed a robot that can learn how to pick up different objects using two different systems by watching thousands of instructional YouTube videos. One of the systems learns to recognize different objects, while the other identifies different types of grasps. The learning system used for the grasping work relies on advanced artificial neural networks. The researchers note this technique is more efficient than programming a robot to handle countless different items, and it can enable a robot to deal with new objects on its own without human intervention.

From Technology Review
View Full Article

 

Abstracts Copyright © 2015 Information Inc., Bethesda, Maryland, USA


 

No entries found

Sign In for Full Access
» Forgot Password? » Create an ACM Web Account