[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"project-1551":3},{"id":4,"name":5,"fullName":6,"owner":7,"repo":5,"description":8,"homepage":8,"htmlUrl":8,"language":9,"languages":8,"totalLinesOfCode":8,"stars":10,"forks":11,"watchers":12,"openIssues":13,"contributorsCount":14,"subscribersCount":14,"size":14,"stars1d":15,"stars7d":16,"stars30d":17,"stars90d":14,"forks30d":14,"starsTrendScore":18,"compositeScore":19,"rankGlobal":8,"rankLanguage":8,"license":20,"archived":21,"fork":21,"defaultBranch":22,"hasWiki":23,"hasPages":21,"topics":24,"createdAt":8,"pushedAt":8,"updatedAt":25,"readmeContent":26,"aiSummary":27,"trendingCount":14,"starSnapshotCount":14,"syncStatus":13,"lastSyncTime":28,"discoverSource":29},1551,"RoboLab","NVlabs\u002FRoboLab","NVlabs",null,"Python",283,26,3,2,0,14,49,83,42,4.29,"Other",false,"main",true,[],"2026-06-12 02:00:29","# RoboLab\n\n[![Website](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FWebsite-RoboLab-blue?logo=googlechrome&logoColor=white)](https:\u002F\u002Fresearch.nvidia.com\u002Flabs\u002Fsrl\u002Fprojects\u002Frobolab)\n[![arXiv](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FarXiv-2604.09860-b31b1b?logo=arXiv&logoColor=white)](https:\u002F\u002Farxiv.org\u002Fabs\u002F2604.09860)\n\n**RoboLab** is a task-based evaluation benchmark for robot manipulation policies built on [NVIDIA Isaac Lab](https:\u002F\u002Fgithub.com\u002Fisaac-sim\u002FIsaacLab). It provides 100+ manipulation tasks with automated success detection, a server-client policy architecture, and multi-environment parallel evaluation — designed for reproducible, large-scale benchmarking of generalist robot policies in simulation.\n\n\u003Cdiv align=\"center\">\n  \u003Cimg src=\"docs\u002Fimages\u002Frobolab.png\" alt=\"RoboLab Overview\" width=\"800\"\u002F>\n\u003C\u002Fdiv>\n\n## Key Features\n\n- **RoboLab-120**: An initial set of 120 brand new benchmark [tasks](robolab\u002Ftasks\u002FREADME.md) spanning pick-and-place, stacking, rearrangement, tool use, and more — each with language instructions and automated success\u002Ffailure detection via composable predicates.\n- **Bring your own robot**: Tasks are not tied to a specific robot embodiment, so you can plug in any robot compatible with IsaacLab!\n- **Rich Asset Libraries**: See a list of [objects](assets\u002Fobjects\u002FREADME.md), [scenes](assets\u002Fscenes\u002FREADME.md), and curated [backgrounds](assets\u002Fbackgrounds\u002FREADME.md) — everything you need to create new scenes and new tasks for your own evaluation needs.\n- **AI-Enabled Workflows**: Generate new scenes and tasks **in minutes** using natural language with the [\u002Frobolab-scenegen](skills\u002Frobolab-scenegen\u002F) and [\u002Frobolab-taskgen](skills\u002Frobolab-taskgen\u002F) Claude Code skills.\n- **Multi-Environment Parallel Evaluation**: Run multiple episodes in parallel across environments with vectorized conditionals and per-environment termination.\n- **Server-Client Policy Architecture**: Policy models run as standalone servers; RoboLab connects via lightweight inference clients (OpenPI, GR00T, and more).\n\n## Getting Started\n\nRequires [uv](https:\u002F\u002Fdocs.astral.sh\u002Fuv\u002Fgetting-started\u002Finstallation\u002F). Isaac Sim 5.0 and Isaac Lab 2.2.0 are installed automatically via `uv sync`. See [Requirements](#requirements) for hardware.\n\n### Installation\n\n```bash\ngit clone https:\u002F\u002Fgithub.com\u002FNVlabs\u002FRoboLab.git\ncd robolab\nuv venv --python 3.11\nsource .venv\u002Fbin\u002Factivate\nuv sync\n```\n\nOn first run, Isaac Sim prompts you to accept the NVIDIA Omniverse EULA. Either accept it interactively, or set it once in your shell:\n```bash\nexport OMNI_KIT_ACCEPT_EULA=Y\n```\n\n> **Running without activating the venv**: if you don't `source .venv\u002Fbin\u002Factivate`, prefix every `python` command with `uv run` (e.g. `uv run python scripts\u002Fcheck_registered_envs.py`).\n\nVerify installation:\n```bash\npython scripts\u002Fcheck_registered_envs.py\n```\n\n### Run without a policy\n\n```bash\n# Run an empty episode with random actions\npython examples\u002Fdemo\u002Frun_empty.py --headless\n\n# Playback recorded demonstration data\npython examples\u002Fdemo\u002Frun_recorded.py --headless\n```\n\n### Run with a policy\n\nRoboLab uses a **server-client architecture**: your model runs as a standalone server, and RoboLab connects to it via a lightweight inference client. To quickly test RoboLab, try [Pi0-5 via OpenPI](docs\u002Finference.md#openpi-pi0--pi0-fast--pi05).\n\nEach inference client has its own lightweight Python dependency — e.g. Pi0 \u002F Pi0-fast \u002F Pi05 need `openpi-client`, which is **not** installed by `uv sync`. Install only the client(s) you need; see [docs\u002Finference.md](docs\u002Finference.md) for each backend. For example, to use the Pi0 family:\n```bash\n# Clone the OpenPI repo separately and install its client into the RoboLab venv\ngit clone git@github.com:xuningy\u002Fopenpi.git ..\u002Fopenpi\nuv pip install -e ..\u002Fopenpi\u002Fpackages\u002Fopenpi-client\n```\n\n1. Start your policy server in a separate terminal.\n2. Run evaluation:\n   ```bash\n   python examples\u002Fpolicy\u002Frun_eval.py --policy pi05 --task BananaInBowlTask --num-envs 12 --headless\n   ```\n3. Analyze results:\n   ```bash\n   python analysis\u002Fread_results.py output\u002F\u003Cyour_run_folder>\n   ```\n\n### Common CLI Options\n\n```bash\n# Run on specific tasks (these two are good for sanity checking)\npython examples\u002Fpolicy\u002Frun_eval.py --policy pi05 --task BananaInBowlTask RubiksCubeAndBananaTask\n\n# Run on a tag of tasks\npython examples\u002Fpolicy\u002Frun_eval.py --policy pi05 --tag semantics\n\n# Run 12 parallel episodes per task\npython examples\u002Fpolicy\u002Frun_eval.py --policy pi05 --headless --num-envs 12\n\n# Enable subtask progress tracking\npython examples\u002Fpolicy\u002Frun_eval.py --policy pi05 --headless --enable-subtask\n\n# Resume a previous run (skips completed episodes)\npython examples\u002Fpolicy\u002Frun_eval.py --policy pi05 --output-folder-name my_previous_run\n```\n\n## Documentation\n\nFull documentation is at **[docs\u002FREADME.md](docs\u002FREADME.md)**, covering:\n\n- [Objects](docs\u002Fobjects.md), [Scenes](docs\u002Fscene.md), [Tasks](docs\u002Ftask.md) — Creating and managing assets and benchmark tasks\n- [Robots](docs\u002Frobots.md), [Cameras](docs\u002Fcamera.md), [Lighting](docs\u002Flighting.md), [Backgrounds](docs\u002Fbackground.md) — Configuring simulation parameters\n- [Environment Registration](docs\u002Fenvironment_registration.md) — Combining tasks with robot\u002Fobservation\u002Faction configs\n- [Inference Clients](docs\u002Finference.md) — Built-in policy clients and server setup (OpenPI, GR00T)\n- [Evaluating a New Policy](docs\u002Fpolicy.md) — Implementing your own inference client\n- [Running Environments](docs\u002Fenvironment_run.md) — CLI reference and evaluation workflows\n- [Analysis and Results](docs\u002Fanalysis.md) — Summarizing, comparing, and auditing results\n- [Subtask Checking](docs\u002Fsubtask.md), [Conditionals](docs\u002Ftask_conditionals.md), [Event Tracking](docs\u002Fevent_tracking.md) — Advanced task features\n\n## Example Tasks\n\nSee the full [Benchmark Task Library](robolab\u002Ftasks\u002FREADME.md) for all 120 tasks.\n\n\u003Cdiv align=\"center\">\n  \u003Cimg src=\"docs\u002Fimages\u002FMake_sure_all_the_white_mugs_are_upright_so_that_the_opening_is_facing_upwards_0_hstack_3X_fps24_width800.gif\" alt=\"Make sure all the white mugs are upright so that the opening is facing upwards\" width=\"800\"\u002F>\n  \u003Cbr>\u003Cem>\"Make sure all the white mugs are upright so that the opening is facing upwards.\"\u003C\u002Fem>\n\u003C\u002Fdiv>\n\n\u003Cdiv align=\"center\">\n  \u003Cimg src=\"docs\u002Fimages\u002FPut_all_plastic_bottles_away_in_the_bin_3_hstack_3X_fps24.gif\" alt=\"Put all plastic bottles away in the bin\" width=\"800\"\u002F>\n  \u003Cbr>\u003Cem>\"Put all plastic bottles away in the bin.\"\u003C\u002Fem>\n\u003C\u002Fdiv>\n\n\u003Cdiv align=\"center\">\n  \u003Cimg src=\"docs\u002Fimages\u002FPut_the_orange_measuring_cup_and_the_blue_measuring_cup_outside_of_the_plate_0_hstack_3X_fps24_width800.gif\" alt=\"Put the orange measuring cup and the blue measuring cup outside of the plate\" width=\"800\"\u002F>\n  \u003Cbr>\u003Cem>\"Put the orange measuring cup and the blue measuring cup outside of the plate.\"\u003C\u002Fem>\n\u003C\u002Fdiv>\n\n## Requirements\n\n| Dependency | Version |\n|---|---|\n| Isaac Sim | 5.0 |\n| Isaac Lab | 2.2.0 |\n| Python | 3.11 |\n| Linux | Ubuntu 22.04+ |\n\n- **Disk space**: ~8 GB (assets account for ~7 GB)\n- **GPU**: NVIDIA RTX GPU required. Recommend 48GB+ VRAM. See [Isaac Lab's hardware requirements](https:\u002F\u002Fisaac-sim.github.io\u002FIsaacLab\u002Fmain\u002Fsource\u002Fsetup\u002Finstallation\u002Findex.html#system-requirements) for recommended GPUs and VRAM.\n- **Speed**: 30 GPU hours \u002F 100 tasks, 1.4 it\u002Fs (assuming ~200ms inference step)\n\n## License\n\nThe RoboLab framework is released under [CC-BY-NC-4.0](https:\u002F\u002Fcreativecommons.org\u002Flicenses\u002Fby-nc\u002F4.0\u002F).\n\n## Citation\n\n```bibtex\n@misc{yang2026robolab,\n      title={RoboLab: A High-Fidelity Simulation Benchmark for Analysis of Task Generalist Policies},\n      author={Xuning Yang and Rishit Dagli and Alex Zook and Hugo Hadfield and Ankit Goyal and Stan Birchfield and Fabio Ramos and Jonathan Tremblay},\n      year={2026},\n      url={https:\u002F\u002Farxiv.org\u002Fabs\u002F2604.09860},\n}\n```\n\n## Contributing\n\nSee [CONTRIBUTING.md](CONTRIBUTING.md) for acknowledgements, issues, and how to contribute.\n","RoboLab是一个基于NVIDIA Isaac Lab构建的机器人操作策略评估基准平台。它提供了120多个涵盖拾取放置、堆叠、重新排列、工具使用等任务，每个任务都配有语言指令和自动成功\u002F失败检测功能。该平台支持多环境并行评估和服务器-客户端策略架构，便于大规模仿真测试通用型机器人策略。RoboLab允许用户自定义机器人模型，并提供丰富的资产库以创建新的场景和任务，同时通过AI驱动的工作流快速生成新场景与任务。此项目适用于需要在仿真环境中进行机器人操作算法开发与评估的研究者及工程师。","2026-06-11 02:44:36","CREATED_QUERY"]