Practice Makes Precise: At CoRL, NVIDIA and Partners Reveal How Robots Rehearse Delicate Work in Simulation
Web page · https://blogs.nvidia.com/blog/robotics-simulation-advancements-corl
Simulation, which has long been the proving ground for robots learning to walk, roll and hop, is now helping researchers teach robots tasks that require a finer touch: aligning gears, connecting cables and assembling small components. The shift from locomotion to manipulation is taking center stage at this week’s Conference on Robot Learning (CoRL) in […]
First seen by Oriel on .
AI summary
NVIDIA unveiled open world action models, simulation tools, benchmarks and agentic workflows at CoRL, aiming to train and evaluate precise robot manipulation for industrial deployment.
- GR00T 2 completed 40 of 50 tasks without task-specific demonstrations and averaged 71% success with one demonstration per task, rising to 95% with five.
- Isaac Lab 3.0 integrates Newton for deformable objects and articulated systems, delivering up to 2.5x Isaac Lab 2.3’s training throughput in a dexterous-manipulation benchmark.
- AlphaNIST learned all 20 NIST Assembly Task Board tasks with one policy, exceeding 98% training success while using roughly 39x less GPU time.
- Isaac Lab-Arena 0.4 scales repeatable policy evaluation; DexBench covers 15 tasks across four industrial scenes, while VABAR evaluates six industrial tasks.
- UC Berkeley’s Graph-as-Policy uses agents and simulation to compose robot skills into interpretable programs, test them and refine them from failures.
#robotics simulation #world action models #agentic robotics #reinforcement learning
Generated by gpt-5.6-sol. Check the original for details.
Aliases
| Source | External ID | URL |
|---|---|---|
| NVIDIA Blog | https://blogs.nvidia.com/?p=99751 | https://blogs.nvidia.com/blog/robotics-simulation-advancements-corl/ Read discussion |