How boredengineering/isaaclab-arena works
Isaac Lab-Arena composes robot embodiments, scenes, and task objectives into Isaac Lab simulation environments, then runs policies and records evaluation results. Users can also define environments as graph specs, generate them through an LLM-assisted workflow, and review previews. The README documents composable assembly, sequential tasks, semantic placement, parallel heterogeneous evaluation, and Isaac Lab interoperability. Some relationships below are directly documented; links within unsampled subsystems are omitted or marked unverified.
Main entry points: isaaclab_arena_environment.py Environment model · isaaclab_arena_manager_based_env.py Arena environment · visualization_service.py Asset visualization · policy_runner.py Policy runners
Components 31
Environment authoring
- Environment graph
isaaclab_arena/environment_spec/arena_env_graph_spec.py - Environment model
isaaclab_arena/environments/isaaclab_arena_environment.py - Scene composition
isaaclab_arena/scene/scene.py - Robot embodiments
isaaclab_arena/embodiments/embodiment_base.py - Task library
isaaclab_arena/tasks/task_library.py - Sequential tasks
isaaclab_arena/tasks/sequential_task_base.py - Relation placement
isaaclab_arena/relations/relation_solver.py
Simulation runtime
- Environment builder
isaaclab_arena/environments/arena_env_builder.py - Arena environment
isaaclab_arena/environments/isaaclab_arena_manager_based_env.py - Scene variations
isaaclab_arena/variations/variation_base.py
Policy evaluation
- Metrics manager
isaaclab_arena/metrics/metrics_manager.py - Episode recording
isaaclab_arena/recording/episode_recorder_manager.py - Policy runners
isaaclab_arena/evaluation/policy_runner.py - Experiment runner
isaaclab_arena/evaluation/experiment_runner.py - Result analysis
isaaclab_arena/visualization/report_data.py
Agentic generation
- Spec inference
isaaclab_arena/agentic_environment_generation/spec_inference.py - LLM backend
isaaclab_arena/agentic_environment_generation/inference_backend.py - Prim path inference
isaaclab_arena/agentic_environment_generation/prim_path_inference.py - Spec validation
isaaclab_arena/agentic_environment_generation/spec_validation.py - Generation review UI
isaaclab_arena_examples/agentic_environment_generation/review_gui/editor_panel.py - Preview client
isaaclab_arena_examples/agentic_environment_generation/review_gui/simapp/client.py - Preview server
isaaclab_arena_examples/agentic_environment_generation/review_gui/simapp/server.py - Asset visualization
isaaclab_arena_examples/agentic_environment_generation/review_gui/visualization_service.py
Runtime foundations
- Bounding geometry
isaaclab_arena/utils/bounding_box.py - Pose utilities
isaaclab_arena/utils/pose.py - USD helpers
isaaclab_arena/utils/usd_helpers.py - USD prim tree
isaaclab_arena/utils/usd_prim_tree.py
Other components
- Researcher
- Inference service
- Isaac Sim
- Policy provider
Connections 34
- Researcher → Environment graph: defines
README.md - Researcher → Environment model: composes
README.md - Environment graph → Environment model: converts to
isaaclab_arena/environment_spec/arena_env_graph_spec.py - Environment model → Scene composition: includes
README.md - Environment model → Robot embodiments: includes
README.md - Environment model → Task library: includes
README.md - Researcher → Environment builder: builds
README.md - Environment builder → Arena environment: assembles
README.md - Task library → Sequential tasks: chains skills
README.md - Relation placement → Scene composition: positions assets
README.md - Arena environment → Scene variations: samples
- Researcher → Policy runners: evaluates
README.md - Policy provider → Policy runners: supplies policy
- Policy runners → Arena environment: acts in
- Arena environment → Episode recording: records episodes
isaaclab_arena/environments/isaaclab_arena_manager_based_env.py - Arena environment → Metrics manager: computes metrics
isaaclab_arena/environments/isaaclab_arena_manager_based_env.py - Policy runners → Experiment runner: runs in
- Experiment runner → Result analysis: reports results
- Researcher → Spec inference: requests generation
- Spec inference → LLM backend: requests structured output
isaaclab_arena/agentic_environment_generation/spec_inference.py - LLM backend → Inference service: sends inference
isaaclab_arena/agentic_environment_generation/inference_backend.py - Spec inference → Spec validation: validates spec
isaaclab_arena/agentic_environment_generation/spec_inference.py - Prim path inference → LLM backend: requests inference
isaaclab_arena/agentic_environment_generation/prim_path_inference.py - Prim path inference → USD prim tree: loads tree
isaaclab_arena/agentic_environment_generation/prim_path_inference.py - Researcher → Generation review UI: reviews spec
- Generation review UI → Asset visualization: requests thumbnails
isaaclab_arena_examples/agentic_environment_generation/review_gui/visualization_service.py - Asset visualization → Preview client: requests preview
isaaclab_arena_examples/agentic_environment_generation/review_gui/visualization_service.py - Preview client → Preview server: sends RPC
isaaclab_arena_examples/agentic_environment_generation/review_gui/simapp/client.py - Preview server → Isaac Sim: renders preview
- Relation placement → Bounding geometry: uses geometry
- Relation placement → Pose utilities: uses transforms
- Asset visualization → USD prim tree: resolves tree
isaaclab_arena_examples/agentic_environment_generation/review_gui/visualization_service.py - Asset visualization → Preview client: gets thumbnails
isaaclab_arena_examples/agentic_environment_generation/review_gui/visualization_service.py - USD prim tree → USD helpers: reads USD
isaaclab_arena/utils/usd_prim_tree.py