Did you have the chance to attend the 2022 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS 2022) in Kyoto? Here we bring you the papers that received an award this year in case you missed them.

Most robots have not generally been equipped for the task of folding clothes. But an international group of researchers say their new method could change that -- or at least speed up the process. Their robot is seen here in multiple exposures. Most robots have not generally been equipped for the task of folding clothes. But an international group of researchers say their new method could change that -- or at least speed up the process.

While it's possible that someone out there enjoys folding clothes, it's probably not a beloved pastime. Accordingly, researchers at UC Berkeley's AUTOLAB have developed a new robotic method of folding garments at record speed (for a robot) called SpeedFolding. Using machine vision, a neural network called BiManual Manipulation Network (BiMaMa-Net), and a pair of industrial robot arms, SpeedFolding can fold 30–40 randomly positioned garments per hour, usually finishing each within two minutes. While that rate does not sound impressive compared to a human, previous robotic garment-folding methods reached only "3-6 FPH" (that's "folds per hour") according to the researchers in a paper submitted for presentation at IROS2022 next week in Kyoto. Speed achievement aside, the paper is worth a read to enjoy how the researchers describe the garment-folding problem in technical terms.

Folding garments reliably and efficiently is a long standing challenge in robotic manipulation due to the complex dynamics and high dimensional configuration space of garments. An intuitive approach is to initially manipulate the garment to a canonical smooth configuration before folding. In this work, we develop SpeedFolding, a reliable and efficient bimanual system, which given user-defined instructions as folding lines, manipulates an initially crumpled garment to (1) a smoothed and (2) a folded configuration. Our primary contribution is a novel neural network architecture that is able to predict pairs of gripper poses to parameterize a diverse set of bimanual action primitives. After learning from 4300 human-annotated and self-supervised actions, the robot is able to fold garments from a random initial configuration in under 120s on average with a success rate of 93%. Real-world experiments show that the system is able to generalize to unseen garments of different color, shape, and stiffness. While prior work achieved 3-6 Folds Per Hour (FPH), SpeedFolding achieves 30-40 FPH.