<ul><li>NVIDIA has introduced DreamGen, an engine designed to scale robot learning using synthetic video data instead of human operators.</li><li>DreamGen generates neural trajectories and photorealistic robot videos paired with motor action labels to train robots on new tasks and environments.</li><li>The technology enables robots to dream inside a video generative model, teaching them new behaviors without direct training and achieving high success rates on novel tasks.</li><li>DreamGen bypasses the need for physical deployment in numerous homes by creating synthetic training data through a 4-step pipeline, training robots to perform tasks zero-shot.</li></ul>

NVIDIA Unveils DreamGen for Robot Learning Through Digital Video ‘Dreams’

Discover more