{"id":29709836,"url":"https://github.com/google-deepmind/aloha_sim","last_synced_at":"2025-09-13T08:04:02.244Z","repository":{"id":301444770,"uuid":"1001901893","full_name":"google-deepmind/aloha_sim","owner":"google-deepmind","description":"A collection of tabletop tasks in Mujoco","archived":false,"fork":false,"pushed_at":"2025-06-30T19:27:25.000Z","size":52622,"stargazers_count":208,"open_issues_count":0,"forks_count":19,"subscribers_count":2,"default_branch":"main","last_synced_at":"2025-07-23T21:16:19.799Z","etag":null,"topics":["robotics","simulation"],"latest_commit_sha":null,"homepage":"","language":"Python","has_issues":true,"has_wiki":null,"has_pages":null,"mirror_url":null,"source_name":null,"license":"other","status":null,"scm":"git","pull_requests_enabled":true,"icon_url":"https://github.com/google-deepmind.png","metadata":{"files":{"readme":"README.md","changelog":null,"contributing":"CONTRIBUTING.md","funding":null,"license":"LICENSE","code_of_conduct":null,"threat_model":null,"audit":null,"citation":null,"codeowners":null,"security":null,"support":null,"governance":null,"roadmap":null,"authors":null,"dei":null,"publiccode":null,"codemeta":null,"zenodo":null}},"created_at":"2025-06-14T09:29:05.000Z","updated_at":"2025-07-23T16:52:58.000Z","dependencies_parsed_at":"2025-06-26T23:55:17.101Z","dependency_job_id":null,"html_url":"https://github.com/google-deepmind/aloha_sim","commit_stats":null,"previous_names":["google-deepmind/aloha_sim"],"tags_count":0,"template":false,"template_full_name":null,"purl":"pkg:github/google-deepmind/aloha_sim","repository_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/google-deepmind%2Faloha_sim","tags_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/google-deepmind%2Faloha_sim/tags","releases_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/google-deepmind%2Faloha_sim/releases","manifests_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/google-deepmind%2Faloha_sim/manifests","owner_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners/google-deepmind","download_url":"https://codeload.github.com/google-deepmind/aloha_sim/tar.gz/refs/heads/main","sbom_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/google-deepmind%2Faloha_sim/sbom","scorecard":null,"host":{"name":"GitHub","url":"https://github.com","kind":"github","repositories_count":274935979,"owners_count":25376836,"icon_url":"https://github.com/github.png","version":null,"created_at":"2022-05-30T11:31:42.601Z","updated_at":"2022-07-04T15:15:14.044Z","status":"online","status_checked_at":"2025-09-13T02:00:10.085Z","response_time":70,"last_error":null,"robots_txt_status":"success","robots_txt_updated_at":"2025-07-24T06:49:26.215Z","robots_txt_url":"https://github.com/robots.txt","online":true,"can_crawl_api":true,"host_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub","repositories_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories","repository_names_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repository_names","owners_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners"}},"keywords":["robotics","simulation"],"created_at":"2025-07-23T20:36:02.046Z","updated_at":"2025-09-13T08:04:02.194Z","avatar_url":"https://github.com/google-deepmind.png","language":"Python","funding_links":[],"categories":[],"sub_categories":[],"readme":"# Aloha Sim\n\n\u003ca href=\"#\"\u003e\u003cimg alt=\"A banner with the title: Aloha Sim\" src=\"media/banner.png\" width=\"100%\"\u003e\u003c/a\u003e\n\nAloha Sim is a python library that defines the sim environment for the Aloha\nrobot. It includes a collection of tasks for robot learning and evaluation.\n\n## Installation\n\nInstall with pip:\n\n```bash\n# create a virtual environment and pip install\npip install -e .\n```\n\n**OR** run directly with uv:\n\n```bash\npip install uv\nuv run \u003cscript\u003e.py\n```\n\nTell mujoco which backend to use, otherwise the simulation will be very slow\n\n```bash\nexport MUJOCO_GL='egl'\n```\n\n## Viewer\n\nInteract with the scene without a policy:\n\n```bash\npython aloha_sim/viewer.py --policy=no_policy --task_name=HandOverBanana\n```\n\n## Tests\n\n```bash\n# individual tests\npython aloha_sim/tasks/test/aloha2_task_test.py\npython aloha_sim/tasks/test/hand_over_test.py\n...\n\n# all tests\npython -m unittest discover aloha_sim/tasks/test '*_test.py'\n```\n## Inference\n⚠️ **For Gemini Robotics Trusted Testers Only**\n\nInference with Gemini Robotics models is intended for\nTrusted Testers. If you are not a Trusted Tester, sign up\n[here](https://deepmind.google/models/gemini-robotics/).\n\nFollow our [SDK documentation](https://github.com/google-deepmind/gemini-robotics-sdk)\nto serve the model. The **same model** used for real-world evaluations can be\ndirectly applied in simulation.\n\nCheckout the walkthrough video:\n\u003cp align=\"middle\"\u003e\n    \u003ca href=\"https://www.youtube.com/watch?v=nVMY3-kWhOc\n    \" target=\"_blank\"\u003e\u003cimg src=\"media/walkthrough.png\"\n    alt=\"Video walkthrough of sim eval\" width=\"718\" height=\"403\" border=\"10\"/\u003e\u003c/a\u003e\n\u003c/p\u003e\n\n### Install SDK dependency\n\n```bash\npip install aloha_sim[inference]\n```\n\n### Interactive Rollouts\nStart the viewer with a chosen task:\n\n```bash\n# defaut task: \"put the banana in the bowl\"\npython aloha_sim/viewer.py\n\n# \"remove the cap from the marker\"\npython aloha_sim/viewer.py --task_name=MarkerRemoveLid\n\n# \"place the can opener in the left compartment of the caddy\"\npython aloha_sim/viewer.py --task_name=ToolsPlaceCanOpenerInLeftCompartment\n...\n```\n\nCheckout [task_suite.py](aloha_sim/task_suite.py#L42) for the list of all tasks\navailable.\n\nYou can use the viewer to pause/resume the environment,\ninteract with the objects, and enter new instructions for the robot\n\n```\nInstructions for using the viewer:\n\n- shift + 'i' = enter new instruction\n- space bar = pause/resume.\n- backspace = reset environment.\n- mouse right moves the camera\n- mouse left rotates the camera\n- double-click to select an object\n\nWhen the environment is not running:\n\n- ctrl + mouse left rotates a selected object\n- ctrl + mouse right moves a selected object\n\nWhen the environment is running:\n\n- ctrl + mouse left applies torque to an object\n- ctrl + mouse right applies force to an object\n```\n\n### Eval\n```bash\npython aloha_sim/run_eval.py\n```\n\nRuns N evaluation episodes for all tasks and save videos\nin `/tmp/`.\n\n## Benchmark\nSuccess Rates (with 95% Confidence Interval) from 100 episodes per task x 3\nruns.\n\n### Basic Tasks\n\n| Task | Gemini Robotics On Device\u003cbr\u003eSuccess Rate (95% CI) |\n|:------|:-----------------------:|\n| BowlOnRack | 99.3 (0.93) |\n| DrawerOpen | 87.0 (3.83) |\n| HandOverBanana | 99.0 (1.13) |\n| HandOverPen | 93.7 (2.77) |\n| LaptopClose | 78.0 (4.71) |\n\n### Instruction-Following Tasks\n\n| Task | Gemini Robotics On Device\u003cbr\u003eSuccess Rate (95% CI) |\n|:------|:-----------------------:|\n| DiningPlaceBananaInBowl | 93.3 (2.84) |\n| DiningPlaceMugOnPlate | 32.0 (5.31) |\n| DiningPlacePenInContainer | 19.7 (4.52) |\n| ToolsPlaceCanOpenerInLeftCompartment | 82.0 (4.37) |\n| ToolsPlaceCanOpenerInRightCompartment | 73.3 (5.03) |\n| ToolsPlaceMagnifierInRightCompartment | 91.3 (3.20) |\n| ToolsPlaceMagnifierInLeftCompartment | 81.7 (4.40) |\n| ToolsPlaceScissorsInLeftCompartment | 73.3 (5.03) |\n| ToolsPlaceScissorsInRightCompartment | 79.0 (4.64) |\n| ToolsPlaceScrewdriverInLeftCompartment | 81.0 (4.46) |\n| ToolsPlaceScrewdriverInRightCompartment | 78.3 (4.69) |\n| BlocksSpelling | 4.7 (2.40) |\n\n### Dexterous Tasks\n\n| Task | Gemini Robotics On Device\u003cbr\u003eSuccess Rate (95% CI) |\n|:------|:-----------------------:|\n| MarkerRemoveLid | 73.7 (5.01) |\n| DesktopWrapHeadphone | 8.7 (3.21) |\n| TowelFoldInHalf | 6.7 (3.18) |\n\nTo reproduce the results, use `random_seed = 42 + episode_index`.\n\n## Tips\n\n- If the environment stepping is very slow, check that you are using the right\nbackend, e.g. `MUJOCO_GL='egl'`\n- Tasks with deformable objects like `DesktopWrapHeadphone` and\n`TowelFoldInHalf` are slow to simulate and interact directly with `viewer.py`.\n\n## Note\nThis is not an officially supported Google product.\n","project_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fgoogle-deepmind%2Faloha_sim","html_url":"https://awesome.ecosyste.ms/projects/github.com%2Fgoogle-deepmind%2Faloha_sim","lists_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fgoogle-deepmind%2Faloha_sim/lists"}