This robot could clean up your mess

Robots may have a future in household chores. A new wave of machines could excel at picking up your shoes, matching them in pairs and putting them away – while you nap.

Researchers at the MIT Computer Science and Artificial Intelligence Laboratory developed a system to give robots a better understanding of the random objects that fill our world and sometimes our closets. Similar technology could eventually help automate work in Amazon's packaging warehouses or help clean up private homes, working quietly with Roombas.

"We want robots to learn for themselves how to understand in a very rich and visual way a lot of useful objects for many tasks," said Pete Florence, co-author of the paper.

Robots are already very present in warehouses and assembly lines, but they can not usually learn and improvise. They can take and move objects, but they must be in defined positions and locations.

The type of work that humans do well is general assembly, such as packaging fathers at random in e-commerce warehouses. But with the technology of MIT, a robotic arm could help: he will be able to grab a cup by the handle or shoe by the tongue. The system learns more about objects over time and ends up placing items in a box specifically.

To form the robotic arm, researchers placed groups of similar objects, such as shoes, hats and cups, in front of them. The robotic arm moved the products, examining their characteristics and learning what they had in common. After analyzing different shoes, he learned where the laces should be.

The test was conducted mainly on men's shoes, from the staff of the CSAIL laboratory. But if they place a stack of stilettos in front of the robot, Florence thinks the robot could develop a good visual representation of the high heels.

The research builds on earlier advances in computer vision, in-depth learning and robotics, says Florence. The paper will be presented at the conference on robot learning in Switzerland in October.

CNNMoney (San Francisco) First published on September 10, 2018: 11:55