How would you feel about having a robot help you with cooking and cleaning? This could become a reality thanks to a new robot training method developed by researchers at Carnegie Mellon University. They taught the robots how to perform various household tasks by simply showing them videos of people doing the same in their homes.

The method is called Vision-Robotics Bridge (VRB) and is based on the concept of affordances. Affordances are the possibilities of interaction with objects that are determined by human behavior. For example, when a robot sees a human open a drawer by the handle, it understands that this is a way to interact with the drawer. Thus, the robot can generalize its experience and open any boxes.

The two robots successfully mastered 12 household tasks, such as opening the oven door, removing the pot from the stove, or picking up the phone. “The robot can learn where and how people interact with different objects by watching the video,” said Deepak Patak, assistant professor of robotics at Carnegie Mellon University.

More detailed information is available at project website and in document presented in June at the Vision and Pattern Recognition Conference.