This year’s Westworld TV series painted a picture of a world in which robots can fulfill our every wish. With that in mind, step forward and take a bow “autonomous clothes folder,” an international robotics project created by researchers in the United Kingdom, Greece, and the Czech Republic.
OK, so it’s torturously slow, but it means you won’t have to do any more T-shirt folding yourself — and if you understand what it’s doing it actually represents a pretty impressive example of cutting-edge robots in action.
“In this work, we managed to develop a solution for folding clothes using a dual-arm robot, which can pick up a crumpled garment, unfold it, lay it on the table, and finally fold it,” Andreas Doumanoglou, a doctoral student in robotic vision at Imperial College London, told Digital Trends.
“In the past, we, as well as other groups, tried to solve each of the steps of this process individually, but to deliver an end-to-end solution one has to overcome more difficulties. As far as I know, this is the first attempt to solve and implement on a real robot an end-to-end folding pipeline.”
As Doumanoglou notes, there is still room for improvement to achieve human-level performance, but it greatly reduces the execution time compared to other state-of-the-art robotic methods and once again demonstrates that robots are capable of increasingly complex work.
“The exciting thing about this project is that with a ‘not-so-flexible’ robot, having only two flat fingers and a long hand that restricts the robot’s working space, we were able to come up with algorithms to manipulate and fold regular-sized garments,” Doumanoglou continued. “Clothes are one of the most difficult types of objects for vision and robotic manipulation since they can be deformed in almost infinite number of ways, making recognition hard, and slight deviations in detecting the correct grasping positions can lead to large errors in the unfolding process.”
Jan Stria of the Center for Machine Perception at Czech Technical University in Prague — a co-author on the paper — told us that folding clothes represents a “toy problem” on which algorithms can be tested as relate to similar flexible materials. “The main outcome is learning new methods how to deal with soft objects,” he said.
As such, this is more likely to be a research project than a real-world tool available to buy any time soon.
“At the beginning of the project, we thought that it would be possible to use the final product commercially,” Stria said. “However, the robustness and speed of the folding pipeline is far from the expected industrial standards. They usually use specialized preprogrammed lines that will be probably always more efficient than an autonomous dual arm robot. However, I can imagine that people will have some home assistant robots in the future. Such robot will have some general manipulators to perform various tasks. It will need to act autonomously — so there is a room for commercialization in the future.”