A new study from MIT’s Computer Science and Artificial Intelligence Laboratory (CSAIL) demonstrates a significant advancement in robotic manipulation. Researchers have developed a system that allows robots to perform complex, deformable object manipulation—such as shaping dough or arranging cloth—using a two-stage learning process. The method combines physics-based simulation with real-world adaptation, enabling robots to generalize skills learned in simulation to handle new, unseen tasks and materials. This approach addresses a major challenge in robotics, where the unpredictable behavior of soft materials has traditionally been difficult for machines to master. The technology has potential applications in domestic settings, manufacturing, and healthcare. For more details, read the full article at https://technologyreview.com/2024/07/18/robots-dough-cloth-manipulation.