NVIDIA Isaac Lab-Arena
NVIDIA Isaac™ Lab-Arena is an open-source framework for large-scale policy setup and evaluation in simulation that gives you streamlined APIs to simplify task curation and diversification. You can also run large-scale, GPU-accelerated, parallel evaluations.
It’s built on NVIDIA Isaac Lab, enabling rapid prototyping across diverse embodiments, objects, and environments without complex system building. This makes large-scale evaluation accessible, helping you better publish unified evaluation methods and benchmarks.
How It Works
Isaac Lab-Arena delivers both a framework for creating and executing complex, diverse benchmarks and a growing library of ready-to-use community benchmark content. It makes large-scale simulation-based experimentation much more efficient and accessible.
Benefits for You
You now have a simpler, more effective way to simplify robotics development. Rapid prototyping of complex tasks in simulation can be accomplished without building underlying systems from scratch, while also supporting scalable evaluation across multiple robots and scenarios.
This framework also provides unified access to established community benchmarks and GPU-accelerated evaluations, streamlining the path from research to deployment across diverse environments.
Create tasks from modular building blocks and scale them across robots, objects, scenes, and parameters, with natural-language task authoring coming soon.
Capture extensible metrics and data for deeper evaluation, analysis, and iteration, while agentic evaluation and root cause analysis is coming soon.
Benchmark any policy across thousands of GPU-accelerated simulation environments.
Unify community benchmarks and shared evaluation methods on a common framework.
Run locally or in cloud-native CI/CD workflows, with paths to leaderboards and platforms like LeRobot.
Isaac Lab - Arena Learning Library
Simplify Generalist Robot Policy Evaluation in Simulation with NVIDIA Isaac Lab-Arena
Learn how to install Isaac Lab-Arena, configure tasks and robots, run evaluations across multiple environments in parallel, and benchmark generalist robot policies with detailed performance metrics and visualizations.
Generalist Robot Policy Evaluation in Simulation with NVIDIA Isaac Lab-Arena and LeRobot
This blog introduces NVIDIA Isaac Lab-Arena's integration with Hugging Face's LeRobot Environment Hub, enabling developers to efficiently evaluate generalist robot policies, like GR00T N, through GPU-accelerated simulation that reduces evaluation time from days to under an hour.
Large-Scale Robot Policy Evaluation with NVIDIA Isaac Lab-Arena
This video provides technical deep-dives into Isaac Lab-Arena's modular code architecture — the affordances system that enables generic task definitions across different objects and workflow integration with teleoperation, data generation, and policy training tools.
