NVIDIA Research is making breakthroughs in robot training through areas like multimodal generative AI and synthetic data generation, to be showcased at the International Conference on Robotics and Automation (ICRA).
Research innovations aim to advance autonomous vehicles and humanoid robots by bridging data gaps, improving safety, and enhancing control.
Key NVIDIA-authored papers at ICRA include advancements like DreamDrive for generating 4D driving scenes, system-level safety monitoring for autonomous vehicles, and DexMimicGen for dexterous manipulation datasets.
Workshops at ICRA will delve into topics like human motion prediction, trustworthy embodied AI, real-world planning and control, integrating vision-language models, and human-centered robot learning.