Diffusion Policy tagged posts

A Technique for more Effective Multipurpose Robots

Four photos show, on top level, a simulation of a robot hand using a spatula, knife, hammer and wrench. The second row shows a real robot hand performing the tasks, and the bottom row shows a human hand performing the tasks.
Caption: Three different data domains — simulation (top), robot tele-operation (middle) and human demos (bottom) — allow a robot to learn to use different tools.
Credits:Image: Courtesy of the researchers

With generative AI models, researchers combined robotics data from different sources to help robots learn better. MIT researchers developed a technique to combine robotics training data across domains, modalities, and tasks using generative AI models. They create a combined strategy from several different datasets that enables a robot to learn to perform new tasks in unseen environments.

Let’s say you want to train a robot so it understands how to use tools and can then quickly learn to make repairs around your house with a hammer, wrench, and screwdriver...

Read More