💡 News

• [2025-04-10] RoboVerse gets accepted by RSS 2025!
• [2025-04-03] Code released! This codebase is actively evolving, and we are continuously working to improve it. Contributions from the open-source community are highly encouraged, and we welcome feedback and bug reports through the GitHub Issues page.

🚀 Getting Started

This repository contains both layers of RoboVerse: the simulation core MetaSim (packages/metasim, package roboverse-metasim, import metasim) and the content, learning code and examples (roboverse-py, repo root). Both are released together from one tag.

git clone https://github.com/RoboVerseOrg/RoboVerse.git
cd RoboVerse
python -m pip install -e "packages/metasim[mujoco]" -e ".[mujoco]"   # MetaSim first, then RoboVerse

For development:

python -m pip install -e "packages/metasim[dev,examples,mujoco]" -e ".[dev,mujoco]"

Repository layout (one directory per concern, no top-level scratch):

Directory	What lives there
packages/metasim/	the simulation core (roboverse-metasim): simulator handlers, scenario config, task registry, its own tests and docs
roboverse_pack/	content: tasks, robots, scenes, grounds, queries, randomization (roboverse-py)
roboverse_learn/	RL / IL / VLA training and evaluation code
examples/	the numbered tutorial chain (0_static_scene.py → 16_…) plus dexhands, RL, motion-planning and viewer demos
tests/	RoboVerse content and integration tests (pytest tests/)
tools/	maintained integrations and utilities: *_integration/ (LIBERO, ManiSkill, mjlab, robosuite, RoboTwin, EmbodiedGen), dashboard/, asset upload
scripts/	developer scripts (data collection, conversion, parity probes); being consolidated into tools/
docs/	the Sphinx wiki (roboverse.wiki)
third_party/	vendored / submodule dependencies (curobo, gsnet)

Please refer to the documentation for simulator-specific installation details.

We also provide detailed tutorials for getting started.

🥰 Contributing

Contributions are welcome! Please see CONTRIBUTING.md for more details on how to contribute to this project.

🎅 Wish List

If you have any feature requests (e.g., support for new simulators, new tasks, new workflows, or new features), feel free to add them to the Wish List section of our GitHub Discussions.

You can also upvote the requests you find most relevant or important. We'll prioritize updates based on the number of upvotes each request receives!

📖 License and Acknowledgments

The RoboVerse source code is licensed under the Apache License 2.0.

The MetaSim core package lives in this repository under packages/metasim and is published as roboverse-metasim.

RoboVerse makes use of the following simulation frameworks, renderers, and libraries:

• Isaac Lab, which is built on top of Isaac Sim
• Isaac Gym
• MuJoCo
• SAPIEN
• PyBullet
• Genesis
• cuRobo
• PyRep, which is built on top of CoppeliaSim
• Blender

RoboVerse also integrates data from the following projects:

• RLBench
• Maniskill
• LIBERO
• Meta-World
• robosuite
• GraspNet
• ARNOLD
• GAPartNet
• GAPartManip
• UniDoorManip
• SimplerEnv
• RLAfford
• Open6DOR
• CALVIN
• GarmentLab
• Matterport3D
• VLN-CE
• vMaterials
• HumanoidBench

Third-party code

Beyond the data above, RoboVerse includes or adapts source code from the projects below. Each is used under its own license; every affected file carries a header naming the upstream, its license, and our changes — the only exceptions are the components listed as Unresolved in THIRD_PARTY_NOTICES.md. THIRD_PARTY_NOTICES.md is the authoritative index — it maps each path in this repo to its upstream, license, and license text. See also NOTICE.

• FastTD3 (MIT) — roboverse_learn/rl/fast_td3/
• CleanRL (MIT) — roboverse_learn/rl/clean_rl/
• stable-baselines3 (MIT)
• diffusion_policy (MIT) — roboverse_learn/il/policies/dp/
• robomimic (MIT), robosuite (MIT)
• ACT (MIT) — roboverse_learn/il/policies/act/
• minGPT (MIT), DETR (Apache-2.0)
• BeyondMimic (MIT) — roboverse_pack/tasks/beyondmimic/
• ManiSkill & ManiSkill2-real2sim (Apache-2.0)
• MuJoCo Playground (Apache-2.0), mjlab (Apache-2.0)
• Isaac Lab (BSD-3-Clause), rsl_rl (BSD-3-Clause)

If you contribute code adapted from another project, follow the attribution rule in AGENTS.md — the header is required, not optional.

Asset licenses are documented per integration under docs/source/dataset_benchmark/; follow the original terms for each. Please contact us if you spot an attribution error or omission — we will correct it promptly.

Citation

If you find RoboVerse useful, please consider citing it:

@misc{geng2025roboverse,
      title={RoboVerse: Towards a Unified Platform, Dataset and Benchmark for Scalable and Generalizable Robot Learning}, 
      author={Haoran Geng and Feishi Wang and Songlin Wei and Yuyang Li and Bangjun Wang and Boshi An and Charlie Tianyue Cheng and Haozhe Lou and Peihao Li and Yen-Jen Wang and Yutong Liang and Dylan Goetting and Chaoyi Xu and Haozhe Chen and Yuxi Qian and Yiran Geng and Jiageng Mao and Weikang Wan and Mingtong Zhang and Jiangran Lyu and Siheng Zhao and Jiazhao Zhang and Jialiang Zhang and Chengyang Zhao and Haoran Lu and Yufei Ding and Ran Gong and Yuran Wang and Yuxuan Kuang and Ruihai Wu and Baoxiong Jia and Carlo Sferrazza and Hao Dong and Siyuan Huang and Yue Wang and Jitendra Malik and Pieter Abbeel},
      year={2025},
      eprint={2504.18904},
      archivePrefix={arXiv},
      primaryClass={cs.RO},
      url={https://arxiv.org/abs/2504.18904}, 
}