Cornell University researchers have developed a new robotic framework called RHyME that allows robots to learn tasks by watching a single how-to video.
RHyME significantly reduces the time, energy, and cost required to train robots, paving the way for faster development and deployment of robotic systems.
The approach of using how-to videos for training, known as imitation learning, can help robots learn multi-step tasks and adapt to real-world environments.
RHyME requires just 30 minutes of robot data for training, leading to a more than 50% increase in task success compared to previous methods.