Example Environments#

Isaac Lab Arena ships a catalog of ready-to-run environments under isaaclab_arena_environments/. Environments can be provided in two ways:

  • Python registered environments: small compositions of the building blocks introduced in Environment — Scene, Embodiment, and Task — wrapped in an ExampleEnvironmentBase subclass and registered with the global EnvironmentRegistry. The registered Task ID is passed as the positional example_environment argument to scripts such as isaaclab_arena/evaluation/policy_runner.py.

  • Environment graph YAML specs: ArenaEnvGraphSpec files that describe the same scene, embodiment, task, objects, and relations declaratively. These are passed with --env_spec and can be generated from prompts by the Agentic Environment Generation workflow.

The environments are grouped into three catalogs:

Robolab-Inspired Benchmark#

RoboLab environment graph YAMLs live under isaaclab_arena_environments/robolab/. They are generated from natural-language prompts and consumed with --env_spec instead of the positional example_environment name.

See RoboLab Task Catalog for the list of RoboLab tasks currently supported in Arena.

Kitchen Benchmark#

Kitchen benchmark environment graph YAMLs live under isaaclab_arena_environments/kitchen_bench/. They define DROID manipulation tasks across Lightwheel RoboCasa and Replicator kitchen layouts.

See Kitchen Benchmark Catalog for all 21 environment specs and their Pi policy executions.

Python Environment Catalog#

Python registered environments wrapped in an ExampleEnvironmentBase subclass and consumed via the positional example_environment name. They span pick-and-place, articulated-object manipulation, sorting, assembly, goal-pose / lift (RL), sandbox, and sequential / composite tasks.

See Python Environment Catalog for the full list with per-environment Key Specifications tables.

See Also#

  • Environment — the Scene / Embodiment / Task building blocks used by every environment listed here.

  • First Arena Environment — walkthrough of the pick_and_place_maple_table environment.

  • Arena in Your Repository — how to register your own ExampleEnvironmentBase subclass alongside the built-in ones.