Welcome to Isaac Lab Arena!

Isaac Lab Arena extends Isaac Lab with composable tools for creating robotics simulation environments and running them efficiently at scale.

Note

This is the development version of Isaac Lab Arena. It contains the newest features but may not be fully tested yet. For the tested version, please refer to the release/0.2.1 branch.

Modular Environments

Compose scenes, embodiments, and tasks as reusable building blocks instead of duplicating full environment definitions.

Swappable assets

Run-time swapping of building blocks such as target objects and backgrounds allows parallel evaluation covering many variations of one task without code changes.

Automatic Object Placement

Define layouts with semantic spatial relations rather than hand-coded poses. Arena solves and validates candidate placements against object geometry, collisions, and task constraints.

plate.add_relation(On(table))
banana.add_relation(On(plate))
bowl.add_relation(On(table))
bagel_1.add_relation(On(table))
bagel_2.add_relation(On(table))

Parallel Evaluation

Evaluate a policy across many parallel environments with aggregated metrics for high-level summary as well as per-episode recording for detailed analysis.

Agentic environment creation

Turn a natural-language task request into a runnable Arena environment. Edit the environment specification and review the randomized layouts in a GUI interactively.

Built-in Evaluation Environments

Run policies against built-in registered environments.

Environmental Variations

Enable build-time and run-time variations to sample controlled changes such as but not limited to HDR backgrounds, light properties, camera intrinsic and extrinsic parameters, and object mass.

Multi-node Evaluation

Scale-out evaluation across multiple compute nodes.

Sensitivity Analysis

Analyze which environment factors are associated with policy success or failure.

Teleoperation

Teleoperate robots using IsaacTeleop in Arena-defined environments.

License

This code is under an open-source license (Apache 2.0).

Acknowledgments

NVIDIA Isaac Lab-Arena is an open-source framework, available on GitHub, that provides a collaborative system for large-scale robot policy evaluation and benchmarking in simulation, with the evaluation and task layers designed in close collaboration with Lightwheel.

Isaac Lab-Arena was built in collaboration with the authors of Robolab (website, paper).

Contributing

For more details on how to contribute to Isaac Lab Arena, please refer to the Contributing Guidelines.

TABLE OF CONTENTS

• Home

Isaac Lab Arena Overview

• Isaac Lab Arena Overview
• Variational Approach to Robot Policy Evaluation

Set Up

• Installation

Getting Started

• First Arena Environment
• First Arena Experiment
• Exploring Environment Variations
• Running a Real Policy

Arena in Your Repo

• Arena in Your Repository
  • Installing IsaacLab-Arena in Your Repository
  • Defining Environments in Your Repository
  • Your Own Tasks and Embodiments

Example Workflows

• Example Environments
• Evaluation
• Agentic Environment Generation
• Imitation Learning
• Reinforcement Learning

Concepts

• Environment
• Agentic Environment Generation
• Scene
• Task
• Embodiment
• Object and Robot Placement
• Policy
• Arena Experiments
• Variations
• Sensitivity Analysis

Advanced

• Omniverse Authentication
• Assets Management
• Jupyter Notebooks

References

• Release Notes
• Citing Isaac Lab-Arena

Development Team Internal

• Development Team Internal