Cleaning a sink might not be the most exciting robot application we’ve seen over the last few years, but this seemingly simple task is actually the result of some deceptively complex machine learning carried out at Austria’s Automation and Control Institute at TU Wien. The team at TU Wien developed a method to “teach” their robot how to complete the task, which involved attaching sensors to a cleaning sponge which was used to clean the sink, allowing the robot to learn the correct angles to use, the correct amount of force to apply, how quickly to scrub, and a multitude of other variables.
“Capturing the geometric shape of a washbasin with cameras is relatively simple,” says Professor Andreas Kugi, part of the team at TU Wien. “But that’s not the crucial step. It is much more difficult to teach the robot: Which type of movement is required for which part of the surface? How fast should the motion be? What’s the appropriate angle? What’s the right amount of force?”
Kugi expounds that the sink cleaning task is a simple example, and that this programming method could be used to allow robotic arms to carry out a variety of labor tasks. Most interestingly, Kugi floats the idea that eventually, robotic arms could then learn how to complete tasks from other robotic arms, a system known as “federated learning.”
“Let’s imagine many workshops use these self-learning robots to sand or paint surfaces,” says Kugi. “Then, you could let the robots gain experience individually with local data. Still, all the robots could share the parameters they learned with each other.”
See also: Toyota Research Institute partners with Boston Dynamics to help robots learn
