IllustrationNo official image on file. Showing a schematic category illustration.
Collaborative system for large-scale robot policy evaluation in simulation, co-developed with Lightwheel.
Key specs
| Specifications | Not stated |
|---|
Technical details
Isaac Lab-Arena is NVIDIA's open-source framework for large-scale robot policy evaluation and benchmarking in simulation, built on Isaac Lab with evaluation and task layers designed with Lightwheel. It lets developers compose diverse environments from robot embodiments, objects and scenes, and connects to established benchmarks such as LIBERO and RoboCasa to standardize testing before deployment. It is available in LeRobot alongside GR00T N models for fine-tuning and evaluation.
| Architecture | Extension framework on NVIDIA Isaac Lab providing a composable system for building simulation environments (robot embodiments, objects, scenes) and evaluating robot learning policies at scale; evaluation and task layers co-designed with Lightwheel. |
|---|---|
| Inputs outputs | Policies and task/environment definitions in; benchmark evaluation results out; connects to benchmarks such as LIBERO and RoboCasa. |
| Embodiments supported | Various robot embodiments supported by Isaac Lab. |
| Hardware requirements | Requires NVIDIA Isaac Lab / Isaac Sim (NVIDIA RTX GPU). |
| How to try | github.com/…/IsaacLab-Arena |
| Sources |
