diff --git a/README.md b/README.md
index ebe147b..d49990c 100644
--- a/README.md
+++ b/README.md
@@ -3,10 +3,10 @@
 > 🤖 Physical AI (Robotics & Embodied AI) 분야의 오픈소스 모델, 데이터셋, 시뮬레이터를 체계적으로 정리한 큐레이션 리스트.
 > A curated list of open-source models, datasets, and simulators for Physical AI (Robotics & Embodied AI).
 
-[![Models](https://img.shields.io/badge/Models-14-blue)](https://pytorchkorea.github.io/Awesome-Physical-AI)
+[![Models](https://img.shields.io/badge/Models-15-blue)](https://pytorchkorea.github.io/Awesome-Physical-AI)
 [![Datasets](https://img.shields.io/badge/Datasets-11-green)](https://pytorchkorea.github.io/Awesome-Physical-AI)
 [![Simulators](https://img.shields.io/badge/Simulators-9-purple)](https://pytorchkorea.github.io/Awesome-Physical-AI)
-[![Organizations](https://img.shields.io/badge/Organizations-30-orange)](https://pytorchkorea.github.io/Awesome-Physical-AI)
+[![Organizations](https://img.shields.io/badge/Organizations-31-orange)](https://pytorchkorea.github.io/Awesome-Physical-AI)
 [![Updated](https://img.shields.io/badge/Updated-2026-06-29-lightgrey)](https://github.com/PyTorchKorea/Awesome-Physical-AI)
 [![Dashboard](https://img.shields.io/badge/🌐_Dashboard-Live-brightgreen)](https://pytorchkorea.github.io/Awesome-Physical-AI)
 
@@ -44,6 +44,7 @@
 | [GR-1](https://github.com/bytedance/GR-1) | BAAI / Beijing Academy of AI | 2024 | manipulation | manipulator | VLA, IL | 312 | [📄](https://arxiv.org/abs/2312.13139)  |
 | [CrossFormer](https://github.com/rail-berkeley/crossformer) | UC Berkeley / others | 2024 | manipulation | manipulator, mobile | IL, VLA | 283 | [📄](https://arxiv.org/abs/2408.11812) [🤗](https://huggingface.co/rail-berkeley/crossformer) |
 | [CogACT](https://github.com/microsoft/CogACT) | Tsinghua University / Microsoft Research Asia | 2024 | manipulation | manipulator | VLA, diffusion | 0 | [📄](https://cogact.github.io/CogACT_paper.pdf) [🤗](https://huggingface.co/CogACT) |
+| [RoboMamba](https://github.com/lmzpai/roboMamba) | Peking University / AI2Robotics | 2024 | manipulation | manipulator | VLA, IL | 0 | [📄](https://arxiv.org/abs/2406.04339)  |
 
 ---
 
diff --git a/data/models.yaml b/data/models.yaml
index a6139e8..ab1d2ea 100644
--- a/data/models.yaml
+++ b/data/models.yaml
@@ -452,3 +452,41 @@
   - VLA
   - diffusion
   - foundation-model
+- id: robomamba
+  name: RoboMamba
+  org: Peking University / AI2Robotics
+  year: 2024
+  description_en: RoboMamba is an efficient end-to-end Vision-Language-Action model
+    that leverages the Mamba state space model for robotic reasoning and manipulation
+    with linear inference complexity. It integrates a vision encoder with Mamba, aligning
+    visual tokens with language embeddings, and uses a lightweight policy head for
+    SE(3) pose prediction. RoboMamba achieves 3x faster inference than existing VLA
+    models while maintaining competitive reasoning and manipulation performance.
+  description_ko: RoboMamba는 Mamba 상태 공간 모델을 활용하여 선형 추론 복잡도로 로봇 추론과 조작을 수행하는 효율적인
+    엔드투엔드 VLA 모델입니다. 비전 인코더와 Mamba를 통합하여 시각 토큰을 언어 임베딩과 정렬하고, 경량 정책 헤드로 SE(3) 포즈 예측을
+    수행합니다. 기존 VLA 모델 대비 3배 빠른 추론 속도를 달성하면서도 경쟁력 있는 추론 및 조작 성능을 보여줍니다.
+  github_url: https://github.com/lmzpai/roboMamba
+  paper_url: https://arxiv.org/abs/2406.04339
+  hf_url: ''
+  project_url: https://sites.google.com/view/robomamba-web
+  categories:
+  - manipulation
+  hardware:
+  - manipulator
+  learning:
+  - VLA
+  - IL
+  framework:
+  - pytorch
+  communication:
+  - other
+  stats:
+    github_stars: 0
+    github_forks: 0
+    hf_downloads: 0
+    last_updated: '2026-06-29'
+  added_date: '2026-06-29'
+  tags:
+  - VLA
+  - mamba
+  - state-space-model
diff --git a/docs/data.json b/docs/data.json
index b9876f4..4b5fbcc 100644
--- a/docs/data.json
+++ b/docs/data.json
@@ -1,10 +1,10 @@
 {
   "metadata": {
     "last_updated": "2026-06-29",
-    "total_models": 14,
+    "total_models": 15,
     "total_datasets": 11,
     "total_tools": 9,
-    "total_orgs": 30
+    "total_orgs": 31
   },
   "models": [
     {
@@ -554,6 +554,46 @@
         "diffusion",
         "foundation-model"
       ]
+    },
+    {
+      "id": "robomamba",
+      "name": "RoboMamba",
+      "org": "Peking University / AI2Robotics",
+      "year": 2024,
+      "description_en": "RoboMamba is an efficient end-to-end Vision-Language-Action model that leverages the Mamba state space model for robotic reasoning and manipulation with linear inference complexity. It integrates a vision encoder with Mamba, aligning visual tokens with language embeddings, and uses a lightweight policy head for SE(3) pose prediction. RoboMamba achieves 3x faster inference than existing VLA models while maintaining competitive reasoning and manipulation performance.",
+      "description_ko": "RoboMamba는 Mamba 상태 공간 모델을 활용하여 선형 추론 복잡도로 로봇 추론과 조작을 수행하는 효율적인 엔드투엔드 VLA 모델입니다. 비전 인코더와 Mamba를 통합하여 시각 토큰을 언어 임베딩과 정렬하고, 경량 정책 헤드로 SE(3) 포즈 예측을 수행합니다. 기존 VLA 모델 대비 3배 빠른 추론 속도를 달성하면서도 경쟁력 있는 추론 및 조작 성능을 보여줍니다.",
+      "github_url": "https://github.com/lmzpai/roboMamba",
+      "paper_url": "https://arxiv.org/abs/2406.04339",
+      "hf_url": "",
+      "project_url": "https://sites.google.com/view/robomamba-web",
+      "categories": [
+        "manipulation"
+      ],
+      "hardware": [
+        "manipulator"
+      ],
+      "learning": [
+        "VLA",
+        "IL"
+      ],
+      "framework": [
+        "pytorch"
+      ],
+      "communication": [
+        "other"
+      ],
+      "stats": {
+        "github_stars": 0,
+        "github_forks": 0,
+        "hf_downloads": 0,
+        "last_updated": "2026-06-29"
+      },
+      "added_date": "2026-06-29",
+      "tags": [
+        "VLA",
+        "mamba",
+        "state-space-model"
+      ]
     }
   ],
   "datasets": [
diff --git a/docs/index.html b/docs/index.html
index 02fd39b..d410c63 100644
--- a/docs/index.html
+++ b/docs/index.html
@@ -28,7 +28,7 @@
   <script src="https://cdn.jsdelivr.net/npm/chart.js@4.4.0/dist/chart.umd.min.js"></script>
 
   <!-- Embedded data — auto-replaced by scripts/generate_site.py -->
-  <script id="embedded-data">window.__PHYSICAL_AI_DATA__ = {"metadata": {"last_updated": "2026-06-29", "total_models": 14, "total_datasets": 11, "total_tools": 9, "total_orgs": 30}, "models": [{"id": "lerobot", "name": "LeRobot", "org": "Hugging Face", "year": 2024, "description_en": "End-to-end learning for real-world robotics. Provides datasets, pretrained models, and training code for manipulation tasks.", "description_ko": "실제 로봇을 위한 엔드-투-엔드 학습 프레임워크. 조작 태스크를 위한 데이터셋, 사전학습 모델, 학습 코드를 제공합니다.", "github_url": "https://github.com/huggingface/lerobot", "paper_url": "https://arxiv.org/abs/2408.01730", "hf_url": "https://huggingface.co/lerobot", "project_url": "", "categories": ["manipulation"], "hardware": ["manipulator"], "learning": ["IL", "diffusion", "VLA"], "framework": ["pytorch"], "communication": [], "stats": {"github_stars": 24952, "github_forks": 4804, "hf_downloads": 0, "last_updated": "2026-06-14"}, "added_date": "2024-06-01", "tags": ["imitation-learning", "diffusion-policy", "act"]}, {"id": "pi0", "name": "π0 (pi-zero)", "org": "Physical Intelligence", "year": 2024, "description_en": "A generalist robot foundation model based on a flow-matching policy. Supports dexterous manipulation and whole-body control.", "description_ko": "플로우 매칭 정책 기반의 범용 로봇 파운데이션 모델. 정교한 조작과 전신 제어를 지원합니다.", "github_url": "https://github.com/Physical-Intelligence/openpi", "paper_url": "https://arxiv.org/abs/2410.24164", "hf_url": "https://huggingface.co/physical-intelligence/pi0", "project_url": "https://www.physicalintelligence.company/blog/pi0", "categories": ["manipulation", "whole-body"], "hardware": ["manipulator", "humanoid"], "learning": ["VLA", "diffusion"], "framework": ["jax", "pytorch"], "communication": [], "stats": {"github_stars": 12331, "github_forks": 2066, "hf_downloads": 0, "last_updated": "2026-06-14"}, "added_date": "2024-10-31", "tags": ["foundation-model", "flow-matching", "dexterous", "open-weights"]}, {"id": "openvla", "name": "OpenVLA", "org": "Stanford / UC Berkeley", "year": 2024, "description_en": "An open-source Vision-Language-Action model fine-tuned from Prismatic VLM on the Open X-Embodiment dataset.", "description_ko": "Open X-Embodiment 데이터셋으로 Prismatic VLM을 파인튜닝한 오픈소스 Vision-Language-Action 모델.", "github_url": "https://github.com/openvla/openvla", "paper_url": "https://arxiv.org/abs/2406.09246", "hf_url": "https://huggingface.co/openvla/openvla-7b", "project_url": "https://openvla.github.io", "categories": ["manipulation"], "hardware": ["manipulator"], "learning": ["VLA"], "framework": ["pytorch"], "communication": [], "stats": {"github_stars": 6421, "github_forks": 756, "hf_downloads": 1198411, "last_updated": "2026-06-14"}, "added_date": "2024-06-13", "tags": ["VLA", "7B", "prismatic"]}, {"id": "octo", "name": "Octo", "org": "UC Berkeley / Stanford / CMU / others", "year": 2023, "description_en": "A transformer-based robot policy pretrained on 800k+ robot trajectories. Supports diverse robot morphologies via tokenized observations.", "description_ko": "80만 개 이상의 로봇 궤적으로 사전학습된 트랜스포머 기반 로봇 정책. 토큰화된 관측을 통해 다양한 로봇 형태를 지원합니다.", "github_url": "https://github.com/octo-models/octo", "paper_url": "https://arxiv.org/abs/2405.12213", "hf_url": "https://huggingface.co/rail-berkeley/octo-base", "project_url": "https://octo-models.github.io", "categories": ["manipulation"], "hardware": ["manipulator", "mobile"], "learning": ["IL", "VLA"], "framework": ["jax"], "communication": [], "stats": {"github_stars": 1669, "github_forks": 268, "hf_downloads": 48, "last_updated": "2026-06-14"}, "added_date": "2023-10-01", "tags": ["generalist", "transformer", "multi-robot"]}, {"id": "act", "name": "ACT (Action Chunking with Transformers)", "org": "Stanford", "year": 2023, "description_en": "Imitation learning with action chunking. Achieves dexterous bimanual manipulation from 50 demonstrations using a CVAE-based transformer.", "description_ko": "액션 청킹을 활용한 모방 학습. CVAE 기반 트랜스포머로 50개의 시연만으로 정교한 양팔 조작을 달성합니다.", "github_url": "https://github.com/tonyzhaozh/act", "paper_url": "https://arxiv.org/abs/2304.13705", "hf_url": "", "project_url": "https://tonyzhaozh.github.io/aloha/", "categories": ["manipulation", "dexterous"], "hardware": ["manipulator"], "learning": ["IL"], "framework": ["pytorch"], "communication": ["ros2"], "stats": {"github_stars": 2001, "github_forks": 388, "hf_downloads": 0, "last_updated": "2026-06-14"}, "added_date": "2023-04-01", "tags": ["ALOHA", "bimanual", "CVAE", "imitation-learning"]}, {"id": "diffusion-policy", "name": "Diffusion Policy", "org": "Columbia University", "year": 2023, "description_en": "Visuomotor policy learning via conditional denoising diffusion. Achieves state-of-the-art performance on 11 robotic manipulation tasks.", "description_ko": "조건부 노이즈 제거 확산을 통한 시각운동 정책 학습. 11개의 로봇 조작 태스크에서 최고 성능을 달성합니다.", "github_url": "https://github.com/real-stanford/diffusion_policy", "paper_url": "https://arxiv.org/abs/2303.04137", "hf_url": "", "project_url": "https://diffusion-policy.cs.columbia.edu", "categories": ["manipulation"], "hardware": ["manipulator"], "learning": ["IL", "diffusion"], "framework": ["pytorch"], "communication": [], "stats": {"github_stars": 4270, "github_forks": 782, "hf_downloads": 0, "last_updated": "2026-06-14"}, "added_date": "2023-03-07", "tags": ["diffusion", "visuomotor", "score-matching"]}, {"id": "mobile-aloha", "name": "Mobile ALOHA", "org": "Stanford", "year": 2024, "description_en": "Whole-body teleoperation and imitation learning for mobile manipulation. Extends ALOHA with a mobile base for household tasks.", "description_ko": "모바일 조작을 위한 전신 원격 조작 및 모방 학습. ALOHA를 모바일 베이스로 확장하여 가정용 태스크를 수행합니다.", "github_url": "https://github.com/MarkFzp/mobile-aloha", "paper_url": "https://arxiv.org/abs/2401.02117", "hf_url": "", "project_url": "https://mobile-aloha.github.io", "categories": ["manipulation", "navigation", "whole-body"], "hardware": ["mobile", "manipulator"], "learning": ["IL"], "framework": ["pytorch"], "communication": ["ros2"], "stats": {"github_stars": 4439, "github_forks": 730, "hf_downloads": 0, "last_updated": "2026-06-14"}, "added_date": "2024-01-04", "tags": ["mobile-manipulation", "household", "teleoperation"]}, {"id": "humanplus", "name": "HumanPlus", "org": "Stanford", "year": 2024, "description_en": "Humanoid shadowing and imitation from human video. A humanoid robot that learns to shadow humans and imitate skills from egocentric video.", "description_ko": "인간 영상으로부터의 휴머노이드 섀도잉 및 모방 학습. 자아중심 비디오에서 기술을 모방하는 휴머노이드 로봇입니다.", "github_url": "https://github.com/MarkFzp/humanplus", "paper_url": "https://arxiv.org/abs/2406.10454", "hf_url": "", "project_url": "https://humanoid-ai.github.io", "categories": ["manipulation", "whole-body"], "hardware": ["humanoid"], "learning": ["IL"], "framework": ["pytorch"], "communication": ["ros2"], "stats": {"github_stars": 849, "github_forks": 122, "hf_downloads": 0, "last_updated": "2026-06-14"}, "added_date": "2024-06-14", "tags": ["humanoid", "shadowing", "egocentric"]}, {"id": "groot", "name": "GR00T N1", "org": "NVIDIA", "year": 2025, "description_en": "NVIDIA's humanoid foundation model with cross-embodiment generalization. Combines vision-language understanding with action generation.", "description_ko": "NVIDIA의 교차 구현 일반화 기능을 갖춘 휴머노이드 파운데이션 모델. 비전-언어 이해와 행동 생성을 결합합니다.", "github_url": "https://github.com/NVIDIA/Isaac-GR00T", "paper_url": "https://arxiv.org/abs/2503.14734", "hf_url": "https://huggingface.co/nvidia/GR00T-N1-2B", "project_url": "https://developer.nvidia.com/isaac/gr00t", "categories": ["manipulation", "whole-body"], "hardware": ["humanoid"], "learning": ["VLA", "IL"], "framework": ["pytorch"], "communication": ["ros2"], "stats": {"github_stars": 7335, "github_forks": 1255, "hf_downloads": 144, "last_updated": "2026-06-14"}, "added_date": "2025-03-19", "tags": ["humanoid", "foundation-model", "NVIDIA", "cross-embodiment"]}, {"id": "roboflamingo", "name": "RoboFlamingo", "org": "ByteDance", "year": 2023, "description_en": "A vision-language manipulation policy that fine-tunes OpenFlamingo for robotic control on the CALVIN benchmark.", "description_ko": "CALVIN 벤치마크에서 로봇 제어를 위해 OpenFlamingo를 파인튜닝한 비전-언어 조작 정책.", "github_url": "https://github.com/RoboFlamingo/RoboFlamingo", "paper_url": "https://arxiv.org/abs/2311.01378", "hf_url": "", "project_url": "", "categories": ["manipulation"], "hardware": ["manipulator"], "learning": ["VLA"], "framework": ["pytorch"], "communication": [], "stats": {"github_stars": 434, "github_forks": 37, "hf_downloads": 0, "last_updated": "2026-06-14"}, "added_date": "2023-11-02", "tags": ["VLA", "CALVIN", "OpenFlamingo"]}, {"id": "gr1", "name": "GR-1", "org": "BAAI / Beijing Academy of AI", "year": 2024, "description_en": "Unleashing Large-Scale Video Generative Pre-Training for Visual Robot Manipulation. GPT-style pre-training on video then fine-tuned for robot control.", "description_ko": "시각 로봇 조작을 위한 대규모 비디오 생성 사전학습. 비디오에서 GPT 스타일 사전학습 후 로봇 제어로 파인튜닝.", "github_url": "https://github.com/bytedance/GR-1", "paper_url": "https://arxiv.org/abs/2312.13139", "hf_url": "", "project_url": "https://gr1-manipulation.github.io", "categories": ["manipulation"], "hardware": ["manipulator"], "learning": ["VLA", "IL"], "framework": ["pytorch"], "communication": [], "stats": {"github_stars": 312, "github_forks": 15, "hf_downloads": 0, "last_updated": "2026-06-14"}, "added_date": "2024-01-01", "tags": ["video-pretraining", "GPT", "manipulation"]}, {"id": "crossformer", "name": "CrossFormer", "org": "UC Berkeley / others", "year": 2024, "description_en": "A cross-embodiment transformer robot policy trained on 900k diverse robot trajectories. Flexible architecture for heterogeneous robots.", "description_ko": "90만 개의 다양한 로봇 궤적으로 학습된 교차 구현 트랜스포머 로봇 정책. 이기종 로봇을 위한 유연한 아키텍처.", "github_url": "https://github.com/rail-berkeley/crossformer", "paper_url": "https://arxiv.org/abs/2408.11812", "hf_url": "https://huggingface.co/rail-berkeley/crossformer", "project_url": "https://crossformer-model.github.io", "categories": ["manipulation"], "hardware": ["manipulator", "mobile"], "learning": ["IL", "VLA"], "framework": ["jax"], "communication": [], "stats": {"github_stars": 283, "github_forks": 37, "hf_downloads": 12, "last_updated": "2026-06-14"}, "added_date": "2024-08-21", "tags": ["cross-embodiment", "generalist", "transformer"]}, {"id": "robovlms", "name": "RoboVLMs", "org": "Multiple", "year": 2025, "description_en": "A systematically optimized VLA  framework for generalist robots. It easily transfers various foundation VLMs into robust control policies with minimal manual design and supports cross-embodiment data integration.", "description_ko": "범용 로봇을 위해 체계적으로 최적화된 VLA 프레임워크. 최소한의 설계로 다양한 파운데이션 VLM을 강건한 로봇 제어 정책으로 전환하며, 다기종 로봇 데이터 통합을 지원합니다.", "github_url": "https://github.com/Robot-VLAs/RoboVLMs", "paper_url": "https://arxiv.org/abs/2412.14058", "hf_url": "https://huggingface.co/robovlms/RoboVLMs", "project_url": "https://robovlms.github.io/", "categories": ["manipulation"], "hardware": ["manipulator"], "learning": ["VLA", "IL"], "framework": ["pytorch"], "communication": [], "stats": {"github_stars": 473, "github_forks": 21, "hf_downloads": 0, "last_updated": "2026-06-14"}, "added_date": "2026-05-17", "tags": []}, {"id": "cogact", "name": "CogACT", "org": "Tsinghua University / Microsoft Research Asia", "year": 2024, "description_en": "CogACT is a componentized Vision-Language-Action architecture that decouples cognition from action. It utilizes powerful Vision-Language Models  to extract cognitive features, which then condition a specialized Diffusion Transformer-based action module to predict continuous, temporally-correlated robotic action sequences.", "description_ko": "CogACT는 기존의 단일 신경망 모델들과 달리 인지와 행동을 명확히 분리한 컴포넌트형 비전-언어-행동 모델입니다. 강력한 비전-언어 모델을 통해 인지적 특징을 추출하고, 이를 조건으로 특화된 Diffusion Transformer 기반의 행동 모델이 복잡하고 연속적인 로봇의 물리적 제어 궤적을 예측합니다.", "github_url": "https://github.com/microsoft/CogACT", "paper_url": "https://cogact.github.io/CogACT_paper.pdf", "hf_url": "https://huggingface.co/CogACT", "project_url": "https://cogact.github.io/", "categories": ["manipulation"], "hardware": ["manipulator"], "learning": ["VLA", "diffusion"], "framework": ["pytorch"], "communication": ["other"], "stats": {"github_stars": 0, "github_forks": 0, "hf_downloads": 0, "last_updated": "2026-06-21"}, "added_date": "2026-06-21", "tags": ["VLA", "diffusion", "foundation-model"]}], "datasets": [{"id": "open-x-embodiment", "name": "Open X-Embodiment (OXE)", "org": "Google DeepMind / RT-X Team", "year": 2023, "description_en": "A large-scale robot learning dataset aggregating 22 robot embodiments, 527 skills, and 1M+ robot trajectories from 21 institutions.", "description_ko": "21개 기관의 22개 로봇 구현체, 527개 기술, 100만 개 이상의 로봇 궤적을 통합한 대규모 로봇 학습 데이터셋.", "github_url": "https://github.com/google-deepmind/open_x_embodiment", "paper_url": "https://arxiv.org/abs/2310.08864", "hf_url": "https://huggingface.co/datasets/jxu124/OpenX-Embodiment", "project_url": "https://robotics-transformer-x.github.io", "categories": ["manipulation"], "hardware": ["manipulator", "mobile"], "source": ["real", "teleoperation"], "modality": ["rgb", "rgbd", "proprioception"], "scale": {"trajectories": 1000000, "hours": 3000, "environments": 527, "robots": 22}, "stats": {"github_stars": 1883, "hf_downloads": 20001, "last_updated": "2026-06-14", "github_forks": 117}, "added_date": "2023-10-13", "tags": ["large-scale", "multi-robot", "RT-X", "generalist"]}, {"id": "droid", "name": "DROID", "org": "UC Berkeley / Stanford / others", "year": 2024, "description_en": "A large-scale in-the-wild robot manipulation dataset with 350 hours of data collected across 564 scenes in 86 buildings using a Franka Panda arm.", "description_ko": "Franka Panda 팔로 86개 건물의 564개 장면에서 수집된 350시간 분량의 대규모 실제 환경 로봇 조작 데이터셋.", "github_url": "https://github.com/droid-dataset/droid", "paper_url": "https://arxiv.org/abs/2403.12945", "hf_url": "https://huggingface.co/datasets/droid-dataset/droid", "project_url": "https://droid-dataset.github.io", "categories": ["manipulation"], "hardware": ["manipulator"], "source": ["real", "teleoperation"], "modality": ["rgb", "rgbd", "proprioception"], "scale": {"trajectories": 76000, "hours": 350, "environments": 564, "robots": 1}, "stats": {"github_stars": 371, "hf_downloads": 0, "last_updated": "2026-06-14", "github_forks": 79}, "added_date": "2024-03-19", "tags": ["in-the-wild", "Franka", "large-scale", "diverse"]}, {"id": "bridgev2", "name": "BridgeData V2", "org": "UC Berkeley", "year": 2023, "description_en": "A large, diverse dataset of robotic manipulation behaviors collected in a tabletop kitchen environment. 60k+ demonstrations across 24 environments.", "description_ko": "테이블탑 주방 환경에서 수집된 크고 다양한 로봇 조작 행동 데이터셋. 24개 환경에 걸쳐 6만 개 이상의 시연.", "github_url": "https://github.com/rail-berkeley/bridge_data_v2", "paper_url": "https://arxiv.org/abs/2308.12952", "hf_url": "https://huggingface.co/datasets/rail-berkeley/bridge_orig", "project_url": "https://rail-berkeley.github.io/bridgedata/", "categories": ["manipulation"], "hardware": ["manipulator"], "source": ["real", "teleoperation"], "modality": ["rgb", "proprioception"], "scale": {"trajectories": 60096, "hours": 0, "environments": 24, "robots": 1}, "stats": {"github_stars": 282, "hf_downloads": 0, "last_updated": "2026-06-14", "github_forks": 36}, "added_date": "2023-08-24", "tags": ["kitchen", "tabletop", "WidowX", "diverse-skills"]}, {"id": "rh20t", "name": "RH20T", "org": "Shanghai AI Lab", "year": 2023, "description_en": "A comprehensive large-scale robotic dataset with 110k+ robot demonstrations of 140+ tasks from 10+ robots, 4 camera views, and multi-sensor data.", "description_ko": "10개 이상의 로봇, 4개 카메라 뷰, 멀티센서 데이터로 140개 이상의 태스크에 대한 11만 개 이상의 로봇 시연을 포함하는 대규모 데이터셋.", "github_url": "https://github.com/rh20t/rh20t_api", "paper_url": "https://arxiv.org/abs/2307.00595", "hf_url": "", "project_url": "https://rh20t.github.io", "categories": ["manipulation", "dexterous"], "hardware": ["manipulator"], "source": ["real", "teleoperation"], "modality": ["rgb", "rgbd", "tactile", "proprioception", "audio"], "scale": {"trajectories": 110000, "hours": 0, "environments": 0, "robots": 10}, "stats": {"github_stars": 108, "hf_downloads": 0, "last_updated": "2026-06-14", "github_forks": 14}, "added_date": "2023-07-02", "tags": ["multi-sensor", "tactile", "multi-robot", "comprehensive"]}, {"id": "libero", "name": "LIBERO", "org": "UMass Amherst / Bosch", "year": 2023, "description_en": "A benchmark for studying knowledge transfer in robot learning with 130 language-conditioned manipulation tasks across 4 task suites.", "description_ko": "4개의 태스크 스위트에 걸쳐 130개의 언어 조건부 조작 태스크로 로봇 학습의 지식 전이를 연구하는 벤치마크.", "github_url": "https://github.com/Lifelong-Robot-Learning/LIBERO", "paper_url": "https://arxiv.org/abs/2306.03310", "hf_url": "https://huggingface.co/datasets/openvla/modified_libero_spatial", "project_url": "https://lifelong-robot-learning.github.io/LIBERO/", "categories": ["manipulation"], "hardware": ["manipulator"], "source": ["simulation"], "modality": ["rgb", "proprioception"], "scale": {"trajectories": 130000, "hours": 0, "environments": 4, "robots": 1}, "stats": {"github_stars": 1941, "hf_downloads": 0, "last_updated": "2026-06-14", "github_forks": 422}, "added_date": "2023-06-05", "tags": ["benchmark", "lifelong-learning", "language-conditioned", "simulation"]}, {"id": "maniskill2", "name": "ManiSkill2", "org": "UC San Diego / Shanghai AI Lab", "year": 2023, "description_en": "A unified benchmark for generalizable manipulation skills with 20 task families and 2000+ diverse objects. GPU-parallelized simulation.", "description_ko": "20개 태스크 패밀리와 2000개 이상의 다양한 객체를 포함하는 일반화 가능한 조작 기술을 위한 통합 벤치마크. GPU 병렬 시뮬레이션.", "github_url": "https://github.com/haosulab/ManiSkill", "paper_url": "https://arxiv.org/abs/2302.04659", "hf_url": "https://huggingface.co/datasets/haosulab/ManiSkill", "project_url": "https://maniskill.ai", "categories": ["manipulation", "dexterous"], "hardware": ["manipulator"], "source": ["simulation"], "modality": ["rgb", "rgbd", "proprioception"], "scale": {"trajectories": 36000, "hours": 0, "environments": 20, "robots": 8}, "stats": {"github_stars": 2996, "hf_downloads": 247, "last_updated": "2026-06-14", "github_forks": 501}, "added_date": "2023-02-09", "tags": ["benchmark", "GPU-parallel", "SAPIEN", "diverse-objects"]}, {"id": "language-table", "name": "Language-Table", "org": "Google", "year": 2023, "description_en": "A dataset and benchmark of 600k+ robot demonstrations with natural language instructions for pushing and rearranging tabletop blocks.", "description_ko": "테이블탑 블록 밀기와 재배치를 위한 자연어 지시와 함께 60만 개 이상의 로봇 시연 데이터셋 및 벤치마크.", "github_url": "https://github.com/google-research/language-table", "paper_url": "https://arxiv.org/abs/2210.01911", "hf_url": "https://huggingface.co/datasets/google/language_table", "project_url": "", "categories": ["manipulation"], "hardware": ["manipulator"], "source": ["real", "simulation", "teleoperation"], "modality": ["rgb", "proprioception"], "scale": {"trajectories": 600000, "hours": 0, "environments": 1, "robots": 1}, "stats": {"github_stars": 360, "hf_downloads": 0, "last_updated": "2026-06-14", "github_forks": 30}, "added_date": "2022-10-04", "tags": ["language-conditioned", "blocks", "tabletop", "large-scale"]}, {"id": "humanoidbench", "name": "HumanoidBench", "org": "CMU / UC San Diego / MIT", "year": 2024, "description_en": "A simulation benchmark for whole-body humanoid locomotion and manipulation with 27 tasks. Evaluates RL and IL algorithms on a full-body robot.", "description_ko": "27개 태스크를 포함하는 전신 휴머노이드 이동 및 조작을 위한 시뮬레이션 벤치마크. 전신 로봇에서 RL과 IL 알고리즘을 평가합니다.", "github_url": "https://github.com/carlosferrazza/humanoid-bench", "paper_url": "https://arxiv.org/abs/2403.10506", "hf_url": "", "project_url": "https://humanoid-bench.github.io", "categories": ["locomotion", "manipulation", "whole-body"], "hardware": ["humanoid"], "source": ["simulation"], "modality": ["rgb", "proprioception"], "scale": {"trajectories": 0, "hours": 0, "environments": 27, "robots": 1}, "stats": {"github_stars": 773, "hf_downloads": 0, "last_updated": "2026-06-14", "github_forks": 122}, "added_date": "2024-03-15", "tags": ["humanoid", "benchmark", "whole-body", "locomotion"]}, {"id": "metaworld", "name": "Meta-World", "org": "Stanford / Berkeley", "year": 2019, "description_en": "A multi-task RL benchmark with 50 diverse robot manipulation tasks. Widely used for meta-learning and multi-task RL research.", "description_ko": "50개의 다양한 로봇 조작 태스크를 포함하는 멀티태스크 RL 벤치마크. 메타러닝 및 멀티태스크 RL 연구에 널리 사용됩니다.", "github_url": "https://github.com/Farama-Foundation/Metaworld", "paper_url": "https://arxiv.org/abs/1910.10897", "hf_url": "", "project_url": "https://meta-world.github.io", "categories": ["manipulation"], "hardware": ["manipulator"], "source": ["simulation"], "modality": ["proprioception"], "scale": {"trajectories": 0, "hours": 0, "environments": 50, "robots": 1}, "stats": {"github_stars": 1836, "hf_downloads": 0, "last_updated": "2026-06-14", "github_forks": 344}, "added_date": "2019-10-23", "tags": ["benchmark", "meta-learning", "multi-task", "Sawyer"]}, {"id": "roboagent", "name": "RoboAgent Dataset", "org": "CMU", "year": 2023, "description_en": "A dataset for scalable data-efficient learning of robotic manipulation strategies. 7,500 demonstrations covering 38 tasks with semantic augmentation.", "description_ko": "로봇 조작 전략의 확장 가능하고 데이터 효율적인 학습을 위한 데이터셋. 의미론적 증강으로 38개 태스크에 걸쳐 7,500개의 시연.", "github_url": "https://github.com/robopen/roboagent", "paper_url": "https://arxiv.org/abs/2309.01918", "hf_url": "", "project_url": "https://robopen.github.io", "categories": ["manipulation"], "hardware": ["manipulator"], "source": ["real", "teleoperation"], "modality": ["rgb", "proprioception"], "scale": {"trajectories": 7500, "hours": 0, "environments": 38, "robots": 1}, "stats": {"github_stars": 372, "hf_downloads": 0, "last_updated": "2026-06-14", "github_forks": 30}, "added_date": "2023-09-05", "tags": ["semantic-augmentation", "data-efficient", "Franka"]}, {"id": "behavior-1k", "name": "BEHAVIOR-1K", "org": "Stanford Vision and Learning Lab", "year": 2024, "description_en": "BEHAVIOR-1K is a comprehensive simulation benchmark for human-centered robotics. It features 1,000 everyday activities grounded in 50 scenes (houses, gardens, restaurants, offices, etc.) with more than 9,000 objects annotated with rich physical and semantic properties. Built on OmniGibson, it supports physics-based simulation of rigid bodies, deformable bodies, and liquids. The benchmark supports 12 robot models and provides 10,000+ human-collected teleoperation demonstrations across 50 tasks.", "description_ko": "BEHAVIOR-1K는 인간 중심 로봇 공학을 위한 종합적인 시뮬레이션 벤치마크입니다. 주택, 정원, 레스토랑, 사무실 등 50개의 사실적인 장면에 1,000개의 일상 활동이 정의되어 있으며, 풍부한 물리적·의미론적 속성이 주석된 9,000개 이상의 객체를 포함합니다. OmniGibson을 기반으로 강체, 변형체, 액체의 물리 기반 시뮬레이션을 지원합니다. 12종의 로봇 모델을 지원하며, 50개 태스크에 걸쳐 10,000건 이상의 텔레오퍼레이션 시연 데이터를 제공합니다.", "github_url": "https://github.com/StanfordVL/BEHAVIOR-1K", "paper_url": "https://arxiv.org/abs/2403.09227", "hf_url": "https://huggingface.co/behavior-1k", "project_url": "https://behavior.stanford.edu/", "categories": ["manipulation", "navigation"], "hardware": ["manipulator", "mobile"], "source": ["simulation", "teleoperation", "human_demo"], "modality": ["rgb", "depth", "proprioception"], "scale": {"trajectories": 10000, "hours": 0, "environments": 0, "robots": 12}, "stats": {"github_stars": 0, "hf_downloads": 0, "last_updated": "2026-06-29"}, "added_date": "2026-06-29", "tags": []}], "tools": [{"id": "mujoco", "name": "MuJoCo", "org": "DeepMind / Google", "year": 2022, "description_en": "Open-source physics engine for contact-rich robot simulation. Industry standard for locomotion and manipulation research, with high-fidelity rigid body dynamics. Open-sourced by DeepMind in 2022.", "description_ko": "접촉이 많은 로봇 시뮬레이션을 위한 오픈소스 물리 엔진. 고충실도 강체 역학으로 보행 및 조작 연구의 업계 표준. 2022년 DeepMind가 오픈소스화.", "github_url": "https://github.com/google-deepmind/mujoco", "paper_url": "https://arxiv.org/abs/2103.09615", "project_url": "https://mujoco.org", "type": "physics_engine", "gpu_accelerated": false, "ros_support": false, "language": ["c", "python", "c++"], "stats": {"github_stars": 13859, "last_updated": "2026-06-14"}, "added_date": "2022-10-18", "tags": ["physics-engine", "contact-rich", "locomotion", "manipulation", "standard"]}, {"id": "isaac-lab", "name": "Isaac Lab", "org": "NVIDIA", "year": 2023, "description_en": "GPU-accelerated robot learning framework built on NVIDIA Isaac Sim. Supports RL, imitation learning, and sim-to-real transfer across manipulators, humanoids, and quadrupeds. Thousands of parallel environments on a single GPU.", "description_ko": "NVIDIA Isaac Sim 기반 GPU 가속 로봇 학습 프레임워크. RL, 모방 학습, sim-to-real 전이를 지원하며 단일 GPU에서 수천 개의 병렬 환경 실행 가능.", "github_url": "https://github.com/isaac-sim/IsaacLab", "paper_url": "https://arxiv.org/abs/2301.04195", "project_url": "https://isaac-sim.github.io/IsaacLab", "type": "rl_framework", "gpu_accelerated": true, "ros_support": true, "language": ["python"], "stats": {"github_stars": 7438, "last_updated": "2026-06-14"}, "added_date": "2023-05-01", "tags": ["simulation", "GPU", "RL", "sim-to-real", "NVIDIA"]}, {"id": "genesis", "name": "Genesis", "org": "Genesis Authors (MIT/CMU/Stanford)", "year": 2024, "description_en": "Generative universal physics engine for robotics and embodied AI. Achieves 43M FPS on a single RTX 4090. Supports rigid/soft/fluid simulation and generative scene creation via natural language.", "description_ko": "로보틱스와 구현된 AI를 위한 생성형 범용 물리 엔진. RTX 4090 단일 GPU에서 4300만 FPS 달성. 강체/연체/유체 시뮬레이션 및 자연어 기반 씬 생성 지원.", "github_url": "https://github.com/Genesis-Embodied-AI/Genesis", "paper_url": "https://arxiv.org/abs/2501.00599", "project_url": "https://genesis-world.readthedocs.io", "type": "physics_engine", "gpu_accelerated": true, "ros_support": false, "language": ["python"], "stats": {"github_stars": 29334, "last_updated": "2026-06-14"}, "added_date": "2024-12-01", "tags": ["simulation", "physics-engine", "generative", "ultra-fast"]}, {"id": "simplerenv", "name": "SimplerEnv", "org": "UC Berkeley / Google", "year": 2024, "description_en": "Simulation benchmark for evaluating real-robot policies in controlled environments. Built on MuJoCo/SAPIEN to reproduce real-world robot setups and provide reproducible evaluation of VLA/manipulation policies.", "description_ko": "제어된 환경에서 실제 로봇 정책을 평가하기 위한 시뮬레이션 벤치마크. MuJoCo/SAPIEN 기반으로 실제 로봇 설정을 재현하고 VLA/조작 정책의 재현 가능한 평가 제공.", "github_url": "https://github.com/simpler-env/SimplerEnv", "paper_url": "https://arxiv.org/abs/2405.05941", "project_url": "https://simpler-env.github.io", "type": "benchmark", "gpu_accelerated": false, "ros_support": false, "language": ["python"], "stats": {"github_stars": 1087, "last_updated": "2026-06-14"}, "added_date": "2024-05-09", "tags": ["benchmark", "evaluation", "VLA", "manipulation", "sim-to-real"]}, {"id": "robosuite", "name": "Robosuite", "org": "Stanford / UT Austin", "year": 2020, "description_en": "Modular robotic simulation framework built on MuJoCo. Provides standardized environments for manipulation research with diverse robot models, grippers, and tasks. Widely used benchmark for IL and RL research.", "description_ko": "MuJoCo 기반 모듈형 로봇 시뮬레이션 프레임워크. 다양한 로봇 모델, 그리퍼, 태스크로 조작 연구를 위한 표준화된 환경 제공. IL 및 RL 연구에 널리 사용되는 벤치마크.", "github_url": "https://github.com/ARISE-Initiative/robosuite", "paper_url": "https://arxiv.org/abs/2009.12293", "project_url": "https://robosuite.ai", "type": "benchmark", "gpu_accelerated": false, "ros_support": false, "language": ["python"], "stats": {"github_stars": 2462, "last_updated": "2026-06-14"}, "added_date": "2020-09-25", "tags": ["benchmark", "manipulation", "MuJoCo", "modular"]}, {"id": "robocasa", "name": "RoboCasa", "org": "UT Austin", "year": 2024, "description_en": "Large-scale simulation framework for training generalist robot agents in household environments. Extends Robosuite with diverse kitchen/home scenes, 100+ object categories, and language-conditioned task generation.", "description_ko": "가정 환경에서 범용 로봇 에이전트를 훈련하기 위한 대규모 시뮬레이션 프레임워크. Robosuite를 확장하여 다양한 주방/가정 씬, 100개 이상 객체 카테고리, 언어 조건부 태스크 생성 제공.", "github_url": "https://github.com/robocasa/robocasa", "paper_url": "https://arxiv.org/abs/2406.02523", "project_url": "https://robocasa.ai", "type": "full_stack", "gpu_accelerated": false, "ros_support": false, "language": ["python"], "stats": {"github_stars": 1461, "last_updated": "2026-06-14"}, "added_date": "2024-06-04", "tags": ["household", "generalist", "language-conditioned", "manipulation"]}, {"id": "brax", "name": "Brax", "org": "Google", "year": 2021, "description_en": "JAX-based GPU/TPU-accelerated differentiable physics engine for locomotion and continuous control. Supports massively parallel RL training, achieving billions of environment steps per second on accelerator hardware.", "description_ko": "보행 및 연속 제어를 위한 JAX 기반 GPU/TPU 가속 미분 가능 물리 엔진. 가속기 하드웨어에서 초당 수십억 개의 환경 스텝으로 대규모 병렬 RL 학습 지원.", "github_url": "https://github.com/google/brax", "paper_url": "https://arxiv.org/abs/2106.13281", "project_url": "", "type": "physics_engine", "gpu_accelerated": true, "ros_support": false, "language": ["python"], "stats": {"github_stars": 3191, "last_updated": "2026-06-14"}, "added_date": "2021-06-25", "tags": ["JAX", "differentiable", "locomotion", "GPU", "TPU"]}, {"id": "dm-control", "name": "dm_control", "org": "DeepMind", "year": 2018, "description_en": "DeepMind's continuous control suite and Python bindings for MuJoCo. Provides physics-based simulation environments and the DMControl Suite benchmark for locomotion and manipulation RL research.", "description_ko": "DeepMind의 연속 제어 스위트 및 MuJoCo용 Python 바인딩. 보행 및 조작 RL 연구를 위한 물리 기반 시뮬레이션 환경과 DMControl Suite 벤치마크 제공.", "github_url": "https://github.com/google-deepmind/dm_control", "paper_url": "https://arxiv.org/abs/2006.12983", "project_url": "", "type": "benchmark", "gpu_accelerated": false, "ros_support": false, "language": ["python"], "stats": {"github_stars": 4616, "last_updated": "2026-06-14"}, "added_date": "2018-01-01", "tags": ["MuJoCo", "benchmark", "locomotion", "RL"]}, {"id": "sapien", "name": "SAPIEN", "org": "UC San Diego / others", "year": 2020, "description_en": "Realistic and physics-rich simulation environment for articulated object manipulation. Powers the ManiSkill benchmark. Features ray-tracing rendering, GPU parallelism, and a large dataset of articulated household objects.", "description_ko": "분절 객체 조작을 위한 현실적이고 물리적으로 풍부한 시뮬레이션 환경. ManiSkill 벤치마크의 기반. 레이트레이싱 렌더링, GPU 병렬화, 대규모 분절 가정용 객체 데이터셋 제공.", "github_url": "https://github.com/haosulab/SAPIEN", "paper_url": "https://arxiv.org/abs/2003.08515", "project_url": "https://sapien.ucsd.edu", "type": "physics_engine", "gpu_accelerated": true, "ros_support": false, "language": ["python", "c++"], "stats": {"github_stars": 780, "last_updated": "2026-06-14"}, "added_date": "2020-03-18", "tags": ["articulated-objects", "manipulation", "rendering", "ManiSkill"]}]};</script>
+  <script id="embedded-data">window.__PHYSICAL_AI_DATA__ = {"metadata": {"last_updated": "2026-06-29", "total_models": 15, "total_datasets": 11, "total_tools": 9, "total_orgs": 31}, "models": [{"id": "lerobot", "name": "LeRobot", "org": "Hugging Face", "year": 2024, "description_en": "End-to-end learning for real-world robotics. Provides datasets, pretrained models, and training code for manipulation tasks.", "description_ko": "실제 로봇을 위한 엔드-투-엔드 학습 프레임워크. 조작 태스크를 위한 데이터셋, 사전학습 모델, 학습 코드를 제공합니다.", "github_url": "https://github.com/huggingface/lerobot", "paper_url": "https://arxiv.org/abs/2408.01730", "hf_url": "https://huggingface.co/lerobot", "project_url": "", "categories": ["manipulation"], "hardware": ["manipulator"], "learning": ["IL", "diffusion", "VLA"], "framework": ["pytorch"], "communication": [], "stats": {"github_stars": 24952, "github_forks": 4804, "hf_downloads": 0, "last_updated": "2026-06-14"}, "added_date": "2024-06-01", "tags": ["imitation-learning", "diffusion-policy", "act"]}, {"id": "pi0", "name": "π0 (pi-zero)", "org": "Physical Intelligence", "year": 2024, "description_en": "A generalist robot foundation model based on a flow-matching policy. Supports dexterous manipulation and whole-body control.", "description_ko": "플로우 매칭 정책 기반의 범용 로봇 파운데이션 모델. 정교한 조작과 전신 제어를 지원합니다.", "github_url": "https://github.com/Physical-Intelligence/openpi", "paper_url": "https://arxiv.org/abs/2410.24164", "hf_url": "https://huggingface.co/physical-intelligence/pi0", "project_url": "https://www.physicalintelligence.company/blog/pi0", "categories": ["manipulation", "whole-body"], "hardware": ["manipulator", "humanoid"], "learning": ["VLA", "diffusion"], "framework": ["jax", "pytorch"], "communication": [], "stats": {"github_stars": 12331, "github_forks": 2066, "hf_downloads": 0, "last_updated": "2026-06-14"}, "added_date": "2024-10-31", "tags": ["foundation-model", "flow-matching", "dexterous", "open-weights"]}, {"id": "openvla", "name": "OpenVLA", "org": "Stanford / UC Berkeley", "year": 2024, "description_en": "An open-source Vision-Language-Action model fine-tuned from Prismatic VLM on the Open X-Embodiment dataset.", "description_ko": "Open X-Embodiment 데이터셋으로 Prismatic VLM을 파인튜닝한 오픈소스 Vision-Language-Action 모델.", "github_url": "https://github.com/openvla/openvla", "paper_url": "https://arxiv.org/abs/2406.09246", "hf_url": "https://huggingface.co/openvla/openvla-7b", "project_url": "https://openvla.github.io", "categories": ["manipulation"], "hardware": ["manipulator"], "learning": ["VLA"], "framework": ["pytorch"], "communication": [], "stats": {"github_stars": 6421, "github_forks": 756, "hf_downloads": 1198411, "last_updated": "2026-06-14"}, "added_date": "2024-06-13", "tags": ["VLA", "7B", "prismatic"]}, {"id": "octo", "name": "Octo", "org": "UC Berkeley / Stanford / CMU / others", "year": 2023, "description_en": "A transformer-based robot policy pretrained on 800k+ robot trajectories. Supports diverse robot morphologies via tokenized observations.", "description_ko": "80만 개 이상의 로봇 궤적으로 사전학습된 트랜스포머 기반 로봇 정책. 토큰화된 관측을 통해 다양한 로봇 형태를 지원합니다.", "github_url": "https://github.com/octo-models/octo", "paper_url": "https://arxiv.org/abs/2405.12213", "hf_url": "https://huggingface.co/rail-berkeley/octo-base", "project_url": "https://octo-models.github.io", "categories": ["manipulation"], "hardware": ["manipulator", "mobile"], "learning": ["IL", "VLA"], "framework": ["jax"], "communication": [], "stats": {"github_stars": 1669, "github_forks": 268, "hf_downloads": 48, "last_updated": "2026-06-14"}, "added_date": "2023-10-01", "tags": ["generalist", "transformer", "multi-robot"]}, {"id": "act", "name": "ACT (Action Chunking with Transformers)", "org": "Stanford", "year": 2023, "description_en": "Imitation learning with action chunking. Achieves dexterous bimanual manipulation from 50 demonstrations using a CVAE-based transformer.", "description_ko": "액션 청킹을 활용한 모방 학습. CVAE 기반 트랜스포머로 50개의 시연만으로 정교한 양팔 조작을 달성합니다.", "github_url": "https://github.com/tonyzhaozh/act", "paper_url": "https://arxiv.org/abs/2304.13705", "hf_url": "", "project_url": "https://tonyzhaozh.github.io/aloha/", "categories": ["manipulation", "dexterous"], "hardware": ["manipulator"], "learning": ["IL"], "framework": ["pytorch"], "communication": ["ros2"], "stats": {"github_stars": 2001, "github_forks": 388, "hf_downloads": 0, "last_updated": "2026-06-14"}, "added_date": "2023-04-01", "tags": ["ALOHA", "bimanual", "CVAE", "imitation-learning"]}, {"id": "diffusion-policy", "name": "Diffusion Policy", "org": "Columbia University", "year": 2023, "description_en": "Visuomotor policy learning via conditional denoising diffusion. Achieves state-of-the-art performance on 11 robotic manipulation tasks.", "description_ko": "조건부 노이즈 제거 확산을 통한 시각운동 정책 학습. 11개의 로봇 조작 태스크에서 최고 성능을 달성합니다.", "github_url": "https://github.com/real-stanford/diffusion_policy", "paper_url": "https://arxiv.org/abs/2303.04137", "hf_url": "", "project_url": "https://diffusion-policy.cs.columbia.edu", "categories": ["manipulation"], "hardware": ["manipulator"], "learning": ["IL", "diffusion"], "framework": ["pytorch"], "communication": [], "stats": {"github_stars": 4270, "github_forks": 782, "hf_downloads": 0, "last_updated": "2026-06-14"}, "added_date": "2023-03-07", "tags": ["diffusion", "visuomotor", "score-matching"]}, {"id": "mobile-aloha", "name": "Mobile ALOHA", "org": "Stanford", "year": 2024, "description_en": "Whole-body teleoperation and imitation learning for mobile manipulation. Extends ALOHA with a mobile base for household tasks.", "description_ko": "모바일 조작을 위한 전신 원격 조작 및 모방 학습. ALOHA를 모바일 베이스로 확장하여 가정용 태스크를 수행합니다.", "github_url": "https://github.com/MarkFzp/mobile-aloha", "paper_url": "https://arxiv.org/abs/2401.02117", "hf_url": "", "project_url": "https://mobile-aloha.github.io", "categories": ["manipulation", "navigation", "whole-body"], "hardware": ["mobile", "manipulator"], "learning": ["IL"], "framework": ["pytorch"], "communication": ["ros2"], "stats": {"github_stars": 4439, "github_forks": 730, "hf_downloads": 0, "last_updated": "2026-06-14"}, "added_date": "2024-01-04", "tags": ["mobile-manipulation", "household", "teleoperation"]}, {"id": "humanplus", "name": "HumanPlus", "org": "Stanford", "year": 2024, "description_en": "Humanoid shadowing and imitation from human video. A humanoid robot that learns to shadow humans and imitate skills from egocentric video.", "description_ko": "인간 영상으로부터의 휴머노이드 섀도잉 및 모방 학습. 자아중심 비디오에서 기술을 모방하는 휴머노이드 로봇입니다.", "github_url": "https://github.com/MarkFzp/humanplus", "paper_url": "https://arxiv.org/abs/2406.10454", "hf_url": "", "project_url": "https://humanoid-ai.github.io", "categories": ["manipulation", "whole-body"], "hardware": ["humanoid"], "learning": ["IL"], "framework": ["pytorch"], "communication": ["ros2"], "stats": {"github_stars": 849, "github_forks": 122, "hf_downloads": 0, "last_updated": "2026-06-14"}, "added_date": "2024-06-14", "tags": ["humanoid", "shadowing", "egocentric"]}, {"id": "groot", "name": "GR00T N1", "org": "NVIDIA", "year": 2025, "description_en": "NVIDIA's humanoid foundation model with cross-embodiment generalization. Combines vision-language understanding with action generation.", "description_ko": "NVIDIA의 교차 구현 일반화 기능을 갖춘 휴머노이드 파운데이션 모델. 비전-언어 이해와 행동 생성을 결합합니다.", "github_url": "https://github.com/NVIDIA/Isaac-GR00T", "paper_url": "https://arxiv.org/abs/2503.14734", "hf_url": "https://huggingface.co/nvidia/GR00T-N1-2B", "project_url": "https://developer.nvidia.com/isaac/gr00t", "categories": ["manipulation", "whole-body"], "hardware": ["humanoid"], "learning": ["VLA", "IL"], "framework": ["pytorch"], "communication": ["ros2"], "stats": {"github_stars": 7335, "github_forks": 1255, "hf_downloads": 144, "last_updated": "2026-06-14"}, "added_date": "2025-03-19", "tags": ["humanoid", "foundation-model", "NVIDIA", "cross-embodiment"]}, {"id": "roboflamingo", "name": "RoboFlamingo", "org": "ByteDance", "year": 2023, "description_en": "A vision-language manipulation policy that fine-tunes OpenFlamingo for robotic control on the CALVIN benchmark.", "description_ko": "CALVIN 벤치마크에서 로봇 제어를 위해 OpenFlamingo를 파인튜닝한 비전-언어 조작 정책.", "github_url": "https://github.com/RoboFlamingo/RoboFlamingo", "paper_url": "https://arxiv.org/abs/2311.01378", "hf_url": "", "project_url": "", "categories": ["manipulation"], "hardware": ["manipulator"], "learning": ["VLA"], "framework": ["pytorch"], "communication": [], "stats": {"github_stars": 434, "github_forks": 37, "hf_downloads": 0, "last_updated": "2026-06-14"}, "added_date": "2023-11-02", "tags": ["VLA", "CALVIN", "OpenFlamingo"]}, {"id": "gr1", "name": "GR-1", "org": "BAAI / Beijing Academy of AI", "year": 2024, "description_en": "Unleashing Large-Scale Video Generative Pre-Training for Visual Robot Manipulation. GPT-style pre-training on video then fine-tuned for robot control.", "description_ko": "시각 로봇 조작을 위한 대규모 비디오 생성 사전학습. 비디오에서 GPT 스타일 사전학습 후 로봇 제어로 파인튜닝.", "github_url": "https://github.com/bytedance/GR-1", "paper_url": "https://arxiv.org/abs/2312.13139", "hf_url": "", "project_url": "https://gr1-manipulation.github.io", "categories": ["manipulation"], "hardware": ["manipulator"], "learning": ["VLA", "IL"], "framework": ["pytorch"], "communication": [], "stats": {"github_stars": 312, "github_forks": 15, "hf_downloads": 0, "last_updated": "2026-06-14"}, "added_date": "2024-01-01", "tags": ["video-pretraining", "GPT", "manipulation"]}, {"id": "crossformer", "name": "CrossFormer", "org": "UC Berkeley / others", "year": 2024, "description_en": "A cross-embodiment transformer robot policy trained on 900k diverse robot trajectories. Flexible architecture for heterogeneous robots.", "description_ko": "90만 개의 다양한 로봇 궤적으로 학습된 교차 구현 트랜스포머 로봇 정책. 이기종 로봇을 위한 유연한 아키텍처.", "github_url": "https://github.com/rail-berkeley/crossformer", "paper_url": "https://arxiv.org/abs/2408.11812", "hf_url": "https://huggingface.co/rail-berkeley/crossformer", "project_url": "https://crossformer-model.github.io", "categories": ["manipulation"], "hardware": ["manipulator", "mobile"], "learning": ["IL", "VLA"], "framework": ["jax"], "communication": [], "stats": {"github_stars": 283, "github_forks": 37, "hf_downloads": 12, "last_updated": "2026-06-14"}, "added_date": "2024-08-21", "tags": ["cross-embodiment", "generalist", "transformer"]}, {"id": "robovlms", "name": "RoboVLMs", "org": "Multiple", "year": 2025, "description_en": "A systematically optimized VLA  framework for generalist robots. It easily transfers various foundation VLMs into robust control policies with minimal manual design and supports cross-embodiment data integration.", "description_ko": "범용 로봇을 위해 체계적으로 최적화된 VLA 프레임워크. 최소한의 설계로 다양한 파운데이션 VLM을 강건한 로봇 제어 정책으로 전환하며, 다기종 로봇 데이터 통합을 지원합니다.", "github_url": "https://github.com/Robot-VLAs/RoboVLMs", "paper_url": "https://arxiv.org/abs/2412.14058", "hf_url": "https://huggingface.co/robovlms/RoboVLMs", "project_url": "https://robovlms.github.io/", "categories": ["manipulation"], "hardware": ["manipulator"], "learning": ["VLA", "IL"], "framework": ["pytorch"], "communication": [], "stats": {"github_stars": 473, "github_forks": 21, "hf_downloads": 0, "last_updated": "2026-06-14"}, "added_date": "2026-05-17", "tags": []}, {"id": "cogact", "name": "CogACT", "org": "Tsinghua University / Microsoft Research Asia", "year": 2024, "description_en": "CogACT is a componentized Vision-Language-Action architecture that decouples cognition from action. It utilizes powerful Vision-Language Models  to extract cognitive features, which then condition a specialized Diffusion Transformer-based action module to predict continuous, temporally-correlated robotic action sequences.", "description_ko": "CogACT는 기존의 단일 신경망 모델들과 달리 인지와 행동을 명확히 분리한 컴포넌트형 비전-언어-행동 모델입니다. 강력한 비전-언어 모델을 통해 인지적 특징을 추출하고, 이를 조건으로 특화된 Diffusion Transformer 기반의 행동 모델이 복잡하고 연속적인 로봇의 물리적 제어 궤적을 예측합니다.", "github_url": "https://github.com/microsoft/CogACT", "paper_url": "https://cogact.github.io/CogACT_paper.pdf", "hf_url": "https://huggingface.co/CogACT", "project_url": "https://cogact.github.io/", "categories": ["manipulation"], "hardware": ["manipulator"], "learning": ["VLA", "diffusion"], "framework": ["pytorch"], "communication": ["other"], "stats": {"github_stars": 0, "github_forks": 0, "hf_downloads": 0, "last_updated": "2026-06-21"}, "added_date": "2026-06-21", "tags": ["VLA", "diffusion", "foundation-model"]}, {"id": "robomamba", "name": "RoboMamba", "org": "Peking University / AI2Robotics", "year": 2024, "description_en": "RoboMamba is an efficient end-to-end Vision-Language-Action model that leverages the Mamba state space model for robotic reasoning and manipulation with linear inference complexity. It integrates a vision encoder with Mamba, aligning visual tokens with language embeddings, and uses a lightweight policy head for SE(3) pose prediction. RoboMamba achieves 3x faster inference than existing VLA models while maintaining competitive reasoning and manipulation performance.", "description_ko": "RoboMamba는 Mamba 상태 공간 모델을 활용하여 선형 추론 복잡도로 로봇 추론과 조작을 수행하는 효율적인 엔드투엔드 VLA 모델입니다. 비전 인코더와 Mamba를 통합하여 시각 토큰을 언어 임베딩과 정렬하고, 경량 정책 헤드로 SE(3) 포즈 예측을 수행합니다. 기존 VLA 모델 대비 3배 빠른 추론 속도를 달성하면서도 경쟁력 있는 추론 및 조작 성능을 보여줍니다.", "github_url": "https://github.com/lmzpai/roboMamba", "paper_url": "https://arxiv.org/abs/2406.04339", "hf_url": "", "project_url": "https://sites.google.com/view/robomamba-web", "categories": ["manipulation"], "hardware": ["manipulator"], "learning": ["VLA", "IL"], "framework": ["pytorch"], "communication": ["other"], "stats": {"github_stars": 0, "github_forks": 0, "hf_downloads": 0, "last_updated": "2026-06-29"}, "added_date": "2026-06-29", "tags": ["VLA", "mamba", "state-space-model"]}], "datasets": [{"id": "open-x-embodiment", "name": "Open X-Embodiment (OXE)", "org": "Google DeepMind / RT-X Team", "year": 2023, "description_en": "A large-scale robot learning dataset aggregating 22 robot embodiments, 527 skills, and 1M+ robot trajectories from 21 institutions.", "description_ko": "21개 기관의 22개 로봇 구현체, 527개 기술, 100만 개 이상의 로봇 궤적을 통합한 대규모 로봇 학습 데이터셋.", "github_url": "https://github.com/google-deepmind/open_x_embodiment", "paper_url": "https://arxiv.org/abs/2310.08864", "hf_url": "https://huggingface.co/datasets/jxu124/OpenX-Embodiment", "project_url": "https://robotics-transformer-x.github.io", "categories": ["manipulation"], "hardware": ["manipulator", "mobile"], "source": ["real", "teleoperation"], "modality": ["rgb", "rgbd", "proprioception"], "scale": {"trajectories": 1000000, "hours": 3000, "environments": 527, "robots": 22}, "stats": {"github_stars": 1883, "hf_downloads": 20001, "last_updated": "2026-06-14", "github_forks": 117}, "added_date": "2023-10-13", "tags": ["large-scale", "multi-robot", "RT-X", "generalist"]}, {"id": "droid", "name": "DROID", "org": "UC Berkeley / Stanford / others", "year": 2024, "description_en": "A large-scale in-the-wild robot manipulation dataset with 350 hours of data collected across 564 scenes in 86 buildings using a Franka Panda arm.", "description_ko": "Franka Panda 팔로 86개 건물의 564개 장면에서 수집된 350시간 분량의 대규모 실제 환경 로봇 조작 데이터셋.", "github_url": "https://github.com/droid-dataset/droid", "paper_url": "https://arxiv.org/abs/2403.12945", "hf_url": "https://huggingface.co/datasets/droid-dataset/droid", "project_url": "https://droid-dataset.github.io", "categories": ["manipulation"], "hardware": ["manipulator"], "source": ["real", "teleoperation"], "modality": ["rgb", "rgbd", "proprioception"], "scale": {"trajectories": 76000, "hours": 350, "environments": 564, "robots": 1}, "stats": {"github_stars": 371, "hf_downloads": 0, "last_updated": "2026-06-14", "github_forks": 79}, "added_date": "2024-03-19", "tags": ["in-the-wild", "Franka", "large-scale", "diverse"]}, {"id": "bridgev2", "name": "BridgeData V2", "org": "UC Berkeley", "year": 2023, "description_en": "A large, diverse dataset of robotic manipulation behaviors collected in a tabletop kitchen environment. 60k+ demonstrations across 24 environments.", "description_ko": "테이블탑 주방 환경에서 수집된 크고 다양한 로봇 조작 행동 데이터셋. 24개 환경에 걸쳐 6만 개 이상의 시연.", "github_url": "https://github.com/rail-berkeley/bridge_data_v2", "paper_url": "https://arxiv.org/abs/2308.12952", "hf_url": "https://huggingface.co/datasets/rail-berkeley/bridge_orig", "project_url": "https://rail-berkeley.github.io/bridgedata/", "categories": ["manipulation"], "hardware": ["manipulator"], "source": ["real", "teleoperation"], "modality": ["rgb", "proprioception"], "scale": {"trajectories": 60096, "hours": 0, "environments": 24, "robots": 1}, "stats": {"github_stars": 282, "hf_downloads": 0, "last_updated": "2026-06-14", "github_forks": 36}, "added_date": "2023-08-24", "tags": ["kitchen", "tabletop", "WidowX", "diverse-skills"]}, {"id": "rh20t", "name": "RH20T", "org": "Shanghai AI Lab", "year": 2023, "description_en": "A comprehensive large-scale robotic dataset with 110k+ robot demonstrations of 140+ tasks from 10+ robots, 4 camera views, and multi-sensor data.", "description_ko": "10개 이상의 로봇, 4개 카메라 뷰, 멀티센서 데이터로 140개 이상의 태스크에 대한 11만 개 이상의 로봇 시연을 포함하는 대규모 데이터셋.", "github_url": "https://github.com/rh20t/rh20t_api", "paper_url": "https://arxiv.org/abs/2307.00595", "hf_url": "", "project_url": "https://rh20t.github.io", "categories": ["manipulation", "dexterous"], "hardware": ["manipulator"], "source": ["real", "teleoperation"], "modality": ["rgb", "rgbd", "tactile", "proprioception", "audio"], "scale": {"trajectories": 110000, "hours": 0, "environments": 0, "robots": 10}, "stats": {"github_stars": 108, "hf_downloads": 0, "last_updated": "2026-06-14", "github_forks": 14}, "added_date": "2023-07-02", "tags": ["multi-sensor", "tactile", "multi-robot", "comprehensive"]}, {"id": "libero", "name": "LIBERO", "org": "UMass Amherst / Bosch", "year": 2023, "description_en": "A benchmark for studying knowledge transfer in robot learning with 130 language-conditioned manipulation tasks across 4 task suites.", "description_ko": "4개의 태스크 스위트에 걸쳐 130개의 언어 조건부 조작 태스크로 로봇 학습의 지식 전이를 연구하는 벤치마크.", "github_url": "https://github.com/Lifelong-Robot-Learning/LIBERO", "paper_url": "https://arxiv.org/abs/2306.03310", "hf_url": "https://huggingface.co/datasets/openvla/modified_libero_spatial", "project_url": "https://lifelong-robot-learning.github.io/LIBERO/", "categories": ["manipulation"], "hardware": ["manipulator"], "source": ["simulation"], "modality": ["rgb", "proprioception"], "scale": {"trajectories": 130000, "hours": 0, "environments": 4, "robots": 1}, "stats": {"github_stars": 1941, "hf_downloads": 0, "last_updated": "2026-06-14", "github_forks": 422}, "added_date": "2023-06-05", "tags": ["benchmark", "lifelong-learning", "language-conditioned", "simulation"]}, {"id": "maniskill2", "name": "ManiSkill2", "org": "UC San Diego / Shanghai AI Lab", "year": 2023, "description_en": "A unified benchmark for generalizable manipulation skills with 20 task families and 2000+ diverse objects. GPU-parallelized simulation.", "description_ko": "20개 태스크 패밀리와 2000개 이상의 다양한 객체를 포함하는 일반화 가능한 조작 기술을 위한 통합 벤치마크. GPU 병렬 시뮬레이션.", "github_url": "https://github.com/haosulab/ManiSkill", "paper_url": "https://arxiv.org/abs/2302.04659", "hf_url": "https://huggingface.co/datasets/haosulab/ManiSkill", "project_url": "https://maniskill.ai", "categories": ["manipulation", "dexterous"], "hardware": ["manipulator"], "source": ["simulation"], "modality": ["rgb", "rgbd", "proprioception"], "scale": {"trajectories": 36000, "hours": 0, "environments": 20, "robots": 8}, "stats": {"github_stars": 2996, "hf_downloads": 247, "last_updated": "2026-06-14", "github_forks": 501}, "added_date": "2023-02-09", "tags": ["benchmark", "GPU-parallel", "SAPIEN", "diverse-objects"]}, {"id": "language-table", "name": "Language-Table", "org": "Google", "year": 2023, "description_en": "A dataset and benchmark of 600k+ robot demonstrations with natural language instructions for pushing and rearranging tabletop blocks.", "description_ko": "테이블탑 블록 밀기와 재배치를 위한 자연어 지시와 함께 60만 개 이상의 로봇 시연 데이터셋 및 벤치마크.", "github_url": "https://github.com/google-research/language-table", "paper_url": "https://arxiv.org/abs/2210.01911", "hf_url": "https://huggingface.co/datasets/google/language_table", "project_url": "", "categories": ["manipulation"], "hardware": ["manipulator"], "source": ["real", "simulation", "teleoperation"], "modality": ["rgb", "proprioception"], "scale": {"trajectories": 600000, "hours": 0, "environments": 1, "robots": 1}, "stats": {"github_stars": 360, "hf_downloads": 0, "last_updated": "2026-06-14", "github_forks": 30}, "added_date": "2022-10-04", "tags": ["language-conditioned", "blocks", "tabletop", "large-scale"]}, {"id": "humanoidbench", "name": "HumanoidBench", "org": "CMU / UC San Diego / MIT", "year": 2024, "description_en": "A simulation benchmark for whole-body humanoid locomotion and manipulation with 27 tasks. Evaluates RL and IL algorithms on a full-body robot.", "description_ko": "27개 태스크를 포함하는 전신 휴머노이드 이동 및 조작을 위한 시뮬레이션 벤치마크. 전신 로봇에서 RL과 IL 알고리즘을 평가합니다.", "github_url": "https://github.com/carlosferrazza/humanoid-bench", "paper_url": "https://arxiv.org/abs/2403.10506", "hf_url": "", "project_url": "https://humanoid-bench.github.io", "categories": ["locomotion", "manipulation", "whole-body"], "hardware": ["humanoid"], "source": ["simulation"], "modality": ["rgb", "proprioception"], "scale": {"trajectories": 0, "hours": 0, "environments": 27, "robots": 1}, "stats": {"github_stars": 773, "hf_downloads": 0, "last_updated": "2026-06-14", "github_forks": 122}, "added_date": "2024-03-15", "tags": ["humanoid", "benchmark", "whole-body", "locomotion"]}, {"id": "metaworld", "name": "Meta-World", "org": "Stanford / Berkeley", "year": 2019, "description_en": "A multi-task RL benchmark with 50 diverse robot manipulation tasks. Widely used for meta-learning and multi-task RL research.", "description_ko": "50개의 다양한 로봇 조작 태스크를 포함하는 멀티태스크 RL 벤치마크. 메타러닝 및 멀티태스크 RL 연구에 널리 사용됩니다.", "github_url": "https://github.com/Farama-Foundation/Metaworld", "paper_url": "https://arxiv.org/abs/1910.10897", "hf_url": "", "project_url": "https://meta-world.github.io", "categories": ["manipulation"], "hardware": ["manipulator"], "source": ["simulation"], "modality": ["proprioception"], "scale": {"trajectories": 0, "hours": 0, "environments": 50, "robots": 1}, "stats": {"github_stars": 1836, "hf_downloads": 0, "last_updated": "2026-06-14", "github_forks": 344}, "added_date": "2019-10-23", "tags": ["benchmark", "meta-learning", "multi-task", "Sawyer"]}, {"id": "roboagent", "name": "RoboAgent Dataset", "org": "CMU", "year": 2023, "description_en": "A dataset for scalable data-efficient learning of robotic manipulation strategies. 7,500 demonstrations covering 38 tasks with semantic augmentation.", "description_ko": "로봇 조작 전략의 확장 가능하고 데이터 효율적인 학습을 위한 데이터셋. 의미론적 증강으로 38개 태스크에 걸쳐 7,500개의 시연.", "github_url": "https://github.com/robopen/roboagent", "paper_url": "https://arxiv.org/abs/2309.01918", "hf_url": "", "project_url": "https://robopen.github.io", "categories": ["manipulation"], "hardware": ["manipulator"], "source": ["real", "teleoperation"], "modality": ["rgb", "proprioception"], "scale": {"trajectories": 7500, "hours": 0, "environments": 38, "robots": 1}, "stats": {"github_stars": 372, "hf_downloads": 0, "last_updated": "2026-06-14", "github_forks": 30}, "added_date": "2023-09-05", "tags": ["semantic-augmentation", "data-efficient", "Franka"]}, {"id": "behavior-1k", "name": "BEHAVIOR-1K", "org": "Stanford Vision and Learning Lab", "year": 2024, "description_en": "BEHAVIOR-1K is a comprehensive simulation benchmark for human-centered robotics. It features 1,000 everyday activities grounded in 50 scenes (houses, gardens, restaurants, offices, etc.) with more than 9,000 objects annotated with rich physical and semantic properties. Built on OmniGibson, it supports physics-based simulation of rigid bodies, deformable bodies, and liquids. The benchmark supports 12 robot models and provides 10,000+ human-collected teleoperation demonstrations across 50 tasks.", "description_ko": "BEHAVIOR-1K는 인간 중심 로봇 공학을 위한 종합적인 시뮬레이션 벤치마크입니다. 주택, 정원, 레스토랑, 사무실 등 50개의 사실적인 장면에 1,000개의 일상 활동이 정의되어 있으며, 풍부한 물리적·의미론적 속성이 주석된 9,000개 이상의 객체를 포함합니다. OmniGibson을 기반으로 강체, 변형체, 액체의 물리 기반 시뮬레이션을 지원합니다. 12종의 로봇 모델을 지원하며, 50개 태스크에 걸쳐 10,000건 이상의 텔레오퍼레이션 시연 데이터를 제공합니다.", "github_url": "https://github.com/StanfordVL/BEHAVIOR-1K", "paper_url": "https://arxiv.org/abs/2403.09227", "hf_url": "https://huggingface.co/behavior-1k", "project_url": "https://behavior.stanford.edu/", "categories": ["manipulation", "navigation"], "hardware": ["manipulator", "mobile"], "source": ["simulation", "teleoperation", "human_demo"], "modality": ["rgb", "depth", "proprioception"], "scale": {"trajectories": 10000, "hours": 0, "environments": 0, "robots": 12}, "stats": {"github_stars": 0, "hf_downloads": 0, "last_updated": "2026-06-29"}, "added_date": "2026-06-29", "tags": []}], "tools": [{"id": "mujoco", "name": "MuJoCo", "org": "DeepMind / Google", "year": 2022, "description_en": "Open-source physics engine for contact-rich robot simulation. Industry standard for locomotion and manipulation research, with high-fidelity rigid body dynamics. Open-sourced by DeepMind in 2022.", "description_ko": "접촉이 많은 로봇 시뮬레이션을 위한 오픈소스 물리 엔진. 고충실도 강체 역학으로 보행 및 조작 연구의 업계 표준. 2022년 DeepMind가 오픈소스화.", "github_url": "https://github.com/google-deepmind/mujoco", "paper_url": "https://arxiv.org/abs/2103.09615", "project_url": "https://mujoco.org", "type": "physics_engine", "gpu_accelerated": false, "ros_support": false, "language": ["c", "python", "c++"], "stats": {"github_stars": 13859, "last_updated": "2026-06-14"}, "added_date": "2022-10-18", "tags": ["physics-engine", "contact-rich", "locomotion", "manipulation", "standard"]}, {"id": "isaac-lab", "name": "Isaac Lab", "org": "NVIDIA", "year": 2023, "description_en": "GPU-accelerated robot learning framework built on NVIDIA Isaac Sim. Supports RL, imitation learning, and sim-to-real transfer across manipulators, humanoids, and quadrupeds. Thousands of parallel environments on a single GPU.", "description_ko": "NVIDIA Isaac Sim 기반 GPU 가속 로봇 학습 프레임워크. RL, 모방 학습, sim-to-real 전이를 지원하며 단일 GPU에서 수천 개의 병렬 환경 실행 가능.", "github_url": "https://github.com/isaac-sim/IsaacLab", "paper_url": "https://arxiv.org/abs/2301.04195", "project_url": "https://isaac-sim.github.io/IsaacLab", "type": "rl_framework", "gpu_accelerated": true, "ros_support": true, "language": ["python"], "stats": {"github_stars": 7438, "last_updated": "2026-06-14"}, "added_date": "2023-05-01", "tags": ["simulation", "GPU", "RL", "sim-to-real", "NVIDIA"]}, {"id": "genesis", "name": "Genesis", "org": "Genesis Authors (MIT/CMU/Stanford)", "year": 2024, "description_en": "Generative universal physics engine for robotics and embodied AI. Achieves 43M FPS on a single RTX 4090. Supports rigid/soft/fluid simulation and generative scene creation via natural language.", "description_ko": "로보틱스와 구현된 AI를 위한 생성형 범용 물리 엔진. RTX 4090 단일 GPU에서 4300만 FPS 달성. 강체/연체/유체 시뮬레이션 및 자연어 기반 씬 생성 지원.", "github_url": "https://github.com/Genesis-Embodied-AI/Genesis", "paper_url": "https://arxiv.org/abs/2501.00599", "project_url": "https://genesis-world.readthedocs.io", "type": "physics_engine", "gpu_accelerated": true, "ros_support": false, "language": ["python"], "stats": {"github_stars": 29334, "last_updated": "2026-06-14"}, "added_date": "2024-12-01", "tags": ["simulation", "physics-engine", "generative", "ultra-fast"]}, {"id": "simplerenv", "name": "SimplerEnv", "org": "UC Berkeley / Google", "year": 2024, "description_en": "Simulation benchmark for evaluating real-robot policies in controlled environments. Built on MuJoCo/SAPIEN to reproduce real-world robot setups and provide reproducible evaluation of VLA/manipulation policies.", "description_ko": "제어된 환경에서 실제 로봇 정책을 평가하기 위한 시뮬레이션 벤치마크. MuJoCo/SAPIEN 기반으로 실제 로봇 설정을 재현하고 VLA/조작 정책의 재현 가능한 평가 제공.", "github_url": "https://github.com/simpler-env/SimplerEnv", "paper_url": "https://arxiv.org/abs/2405.05941", "project_url": "https://simpler-env.github.io", "type": "benchmark", "gpu_accelerated": false, "ros_support": false, "language": ["python"], "stats": {"github_stars": 1087, "last_updated": "2026-06-14"}, "added_date": "2024-05-09", "tags": ["benchmark", "evaluation", "VLA", "manipulation", "sim-to-real"]}, {"id": "robosuite", "name": "Robosuite", "org": "Stanford / UT Austin", "year": 2020, "description_en": "Modular robotic simulation framework built on MuJoCo. Provides standardized environments for manipulation research with diverse robot models, grippers, and tasks. Widely used benchmark for IL and RL research.", "description_ko": "MuJoCo 기반 모듈형 로봇 시뮬레이션 프레임워크. 다양한 로봇 모델, 그리퍼, 태스크로 조작 연구를 위한 표준화된 환경 제공. IL 및 RL 연구에 널리 사용되는 벤치마크.", "github_url": "https://github.com/ARISE-Initiative/robosuite", "paper_url": "https://arxiv.org/abs/2009.12293", "project_url": "https://robosuite.ai", "type": "benchmark", "gpu_accelerated": false, "ros_support": false, "language": ["python"], "stats": {"github_stars": 2462, "last_updated": "2026-06-14"}, "added_date": "2020-09-25", "tags": ["benchmark", "manipulation", "MuJoCo", "modular"]}, {"id": "robocasa", "name": "RoboCasa", "org": "UT Austin", "year": 2024, "description_en": "Large-scale simulation framework for training generalist robot agents in household environments. Extends Robosuite with diverse kitchen/home scenes, 100+ object categories, and language-conditioned task generation.", "description_ko": "가정 환경에서 범용 로봇 에이전트를 훈련하기 위한 대규모 시뮬레이션 프레임워크. Robosuite를 확장하여 다양한 주방/가정 씬, 100개 이상 객체 카테고리, 언어 조건부 태스크 생성 제공.", "github_url": "https://github.com/robocasa/robocasa", "paper_url": "https://arxiv.org/abs/2406.02523", "project_url": "https://robocasa.ai", "type": "full_stack", "gpu_accelerated": false, "ros_support": false, "language": ["python"], "stats": {"github_stars": 1461, "last_updated": "2026-06-14"}, "added_date": "2024-06-04", "tags": ["household", "generalist", "language-conditioned", "manipulation"]}, {"id": "brax", "name": "Brax", "org": "Google", "year": 2021, "description_en": "JAX-based GPU/TPU-accelerated differentiable physics engine for locomotion and continuous control. Supports massively parallel RL training, achieving billions of environment steps per second on accelerator hardware.", "description_ko": "보행 및 연속 제어를 위한 JAX 기반 GPU/TPU 가속 미분 가능 물리 엔진. 가속기 하드웨어에서 초당 수십억 개의 환경 스텝으로 대규모 병렬 RL 학습 지원.", "github_url": "https://github.com/google/brax", "paper_url": "https://arxiv.org/abs/2106.13281", "project_url": "", "type": "physics_engine", "gpu_accelerated": true, "ros_support": false, "language": ["python"], "stats": {"github_stars": 3191, "last_updated": "2026-06-14"}, "added_date": "2021-06-25", "tags": ["JAX", "differentiable", "locomotion", "GPU", "TPU"]}, {"id": "dm-control", "name": "dm_control", "org": "DeepMind", "year": 2018, "description_en": "DeepMind's continuous control suite and Python bindings for MuJoCo. Provides physics-based simulation environments and the DMControl Suite benchmark for locomotion and manipulation RL research.", "description_ko": "DeepMind의 연속 제어 스위트 및 MuJoCo용 Python 바인딩. 보행 및 조작 RL 연구를 위한 물리 기반 시뮬레이션 환경과 DMControl Suite 벤치마크 제공.", "github_url": "https://github.com/google-deepmind/dm_control", "paper_url": "https://arxiv.org/abs/2006.12983", "project_url": "", "type": "benchmark", "gpu_accelerated": false, "ros_support": false, "language": ["python"], "stats": {"github_stars": 4616, "last_updated": "2026-06-14"}, "added_date": "2018-01-01", "tags": ["MuJoCo", "benchmark", "locomotion", "RL"]}, {"id": "sapien", "name": "SAPIEN", "org": "UC San Diego / others", "year": 2020, "description_en": "Realistic and physics-rich simulation environment for articulated object manipulation. Powers the ManiSkill benchmark. Features ray-tracing rendering, GPU parallelism, and a large dataset of articulated household objects.", "description_ko": "분절 객체 조작을 위한 현실적이고 물리적으로 풍부한 시뮬레이션 환경. ManiSkill 벤치마크의 기반. 레이트레이싱 렌더링, GPU 병렬화, 대규모 분절 가정용 객체 데이터셋 제공.", "github_url": "https://github.com/haosulab/SAPIEN", "paper_url": "https://arxiv.org/abs/2003.08515", "project_url": "https://sapien.ucsd.edu", "type": "physics_engine", "gpu_accelerated": true, "ros_support": false, "language": ["python", "c++"], "stats": {"github_stars": 780, "last_updated": "2026-06-14"}, "added_date": "2020-03-18", "tags": ["articulated-objects", "manipulation", "rendering", "ManiSkill"]}]};</script>
 
   <style>
     [x-cloak] { display: none !important; }