[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"similar-robotlearning123--awesome-isaac-gym":3,"tool-robotlearning123--awesome-isaac-gym":65},[4,23,32,40,49,57],{"id":5,"name":6,"github_repo":7,"description_zh":8,"stars":9,"difficulty_score":10,"last_commit_at":11,"category_tags":12,"status":22},2268,"ML-For-Beginners","microsoft\u002FML-For-Beginners","ML-For-Beginners 是由微软推出的一套系统化机器学习入门课程，旨在帮助零基础用户轻松掌握经典机器学习知识。这套课程将学习路径规划为 12 周，包含 26 节精炼课程和 52 道配套测验，内容涵盖从基础概念到实际应用的完整流程，有效解决了初学者面对庞大知识体系时无从下手、缺乏结构化指导的痛点。\n\n无论是希望转型的开发者、需要补充算法背景的研究人员，还是对人工智能充满好奇的普通爱好者，都能从中受益。课程不仅提供了清晰的理论讲解，还强调动手实践，让用户在循序渐进中建立扎实的技能基础。其独特的亮点在于强大的多语言支持，通过自动化机制提供了包括简体中文在内的 50 多种语言版本，极大地降低了全球不同背景用户的学习门槛。此外，项目采用开源协作模式，社区活跃且内容持续更新，确保学习者能获取前沿且准确的技术资讯。如果你正寻找一条清晰、友好且专业的机器学习入门之路，ML-For-Beginners 将是理想的起点。",85092,2,"2026-04-10T11:13:16",[13,14,15,16,17,18,19,20,21],"图像","数据工具","视频","插件","Agent","其他","语言模型","开发框架","音频","ready",{"id":24,"name":25,"github_repo":26,"description_zh":27,"stars":28,"difficulty_score":29,"last_commit_at":30,"category_tags":31,"status":22},5784,"funNLP","fighting41love\u002FfunNLP","funNLP 是一个专为中文自然语言处理（NLP）打造的超级资源库，被誉为\"NLP 民工的乐园”。它并非单一的软件工具，而是一个汇集了海量开源项目、数据集、预训练模型和实用代码的综合性平台。\n\n面对中文 NLP 领域资源分散、入门门槛高以及特定场景数据匮乏的痛点，funNLP 提供了“一站式”解决方案。这里不仅涵盖了分词、命名实体识别、情感分析、文本摘要等基础任务的标准工具，还独特地收录了丰富的垂直领域资源，如法律、医疗、金融行业的专用词库与数据集，甚至包含古诗词生成、歌词创作等趣味应用。其核心亮点在于极高的全面性与实用性，从基础的字典词典到前沿的 BERT、GPT-2 模型代码，再到高质量的标注数据和竞赛方案，应有尽有。\n\n无论是刚刚踏入 NLP 领域的学生、需要快速验证想法的算法工程师，还是从事人工智能研究的学者，都能在这里找到急需的“武器弹药”。对于开发者而言，它能大幅减少寻找数据和复现模型的时间；对于研究者，它提供了丰富的基准测试资源和前沿技术参考。funNLP 以开放共享的精神，极大地降低了中文自然语言处理的开发与研究成本，是中文 AI 社区不可或缺的宝藏仓库。",79857,1,"2026-04-08T20:11:31",[19,14,18],{"id":33,"name":34,"github_repo":35,"description_zh":36,"stars":37,"difficulty_score":29,"last_commit_at":38,"category_tags":39,"status":22},5773,"cs-video-courses","Developer-Y\u002Fcs-video-courses","cs-video-courses 是一个精心整理的计算机科学视频课程清单，旨在为自学者提供系统化的学习路径。它汇集了全球知名高校（如加州大学伯克利分校、新南威尔士大学等）的完整课程录像，涵盖从编程基础、数据结构与算法，到操作系统、分布式系统、数据库等核心领域，并深入延伸至人工智能、机器学习、量子计算及区块链等前沿方向。\n\n面对网络上零散且质量参差不齐的教学资源，cs-video-courses 解决了学习者难以找到成体系、高难度大学级别课程的痛点。该项目严格筛选内容，仅收录真正的大学层级课程，排除了碎片化的简短教程或商业广告，确保用户能接触到严谨的学术内容。\n\n这份清单特别适合希望夯实计算机基础的开发者、需要补充特定领域知识的研究人员，以及渴望像在校生一样系统学习计算机科学的自学者。其独特的技术亮点在于分类极其详尽，不仅包含传统的软件工程与网络安全，还细分了生成式 AI、大语言模型、计算生物学等新兴学科，并直接链接至官方视频播放列表，让用户能一站式获取高质量的教育资源，免费享受世界顶尖大学的课堂体验。",79792,"2026-04-08T22:03:59",[18,13,14,20],{"id":41,"name":42,"github_repo":43,"description_zh":44,"stars":45,"difficulty_score":46,"last_commit_at":47,"category_tags":48,"status":22},3128,"ragflow","infiniflow\u002Fragflow","RAGFlow 是一款领先的开源检索增强生成（RAG）引擎，旨在为大语言模型构建更精准、可靠的上下文层。它巧妙地将前沿的 RAG 技术与智能体（Agent）能力相结合，不仅支持从各类文档中高效提取知识，还能让模型基于这些知识进行逻辑推理和任务执行。\n\n在大模型应用中，幻觉问题和知识滞后是常见痛点。RAGFlow 通过深度解析复杂文档结构（如表格、图表及混合排版），显著提升了信息检索的准确度，从而有效减少模型“胡编乱造”的现象，确保回答既有据可依又具备时效性。其内置的智能体机制更进一步，使系统不仅能回答问题，还能自主规划步骤解决复杂问题。\n\n这款工具特别适合开发者、企业技术团队以及 AI 研究人员使用。无论是希望快速搭建私有知识库问答系统，还是致力于探索大模型在垂直领域落地的创新者，都能从中受益。RAGFlow 提供了可视化的工作流编排界面和灵活的 API 接口，既降低了非算法背景用户的上手门槛，也满足了专业开发者对系统深度定制的需求。作为基于 Apache 2.0 协议开源的项目，它正成为连接通用大模型与行业专有知识之间的重要桥梁。",77062,3,"2026-04-04T04:44:48",[17,13,20,19,18],{"id":50,"name":51,"github_repo":52,"description_zh":53,"stars":54,"difficulty_score":46,"last_commit_at":55,"category_tags":56,"status":22},519,"PaddleOCR","PaddlePaddle\u002FPaddleOCR","PaddleOCR 是一款基于百度飞桨框架开发的高性能开源光学字符识别工具包。它的核心能力是将图片、PDF 等文档中的文字提取出来，转换成计算机可读取的结构化数据，让机器真正“看懂”图文内容。\n\n面对海量纸质或电子文档，PaddleOCR 解决了人工录入效率低、数字化成本高的问题。尤其在人工智能领域，它扮演着连接图像与大型语言模型（LLM）的桥梁角色，能将视觉信息直接转化为文本输入，助力智能问答、文档分析等应用场景落地。\n\nPaddleOCR 适合开发者、算法研究人员以及有文档自动化需求的普通用户。其技术优势十分明显：不仅支持全球 100 多种语言的识别，还能在 Windows、Linux、macOS 等多个系统上运行，并灵活适配 CPU、GPU、NPU 等各类硬件。作为一个轻量级且社区活跃的开源项目，PaddleOCR 既能满足快速集成的需求，也能支撑前沿的视觉语言研究，是处理文字识别任务的理想选择。",75585,"2026-04-14T22:04:17",[19,13,20,18],{"id":58,"name":59,"github_repo":60,"description_zh":61,"stars":62,"difficulty_score":29,"last_commit_at":63,"category_tags":64,"status":22},3215,"awesome-machine-learning","josephmisiti\u002Fawesome-machine-learning","awesome-machine-learning 是一份精心整理的机器学习资源清单，汇集了全球优秀的机器学习框架、库和软件工具。面对机器学习领域技术迭代快、资源分散且难以甄选的痛点，这份清单按编程语言（如 Python、C++、Go 等）和应用场景（如计算机视觉、自然语言处理、深度学习等）进行了系统化分类，帮助使用者快速定位高质量项目。\n\n它特别适合开发者、数据科学家及研究人员使用。无论是初学者寻找入门库，还是资深工程师对比不同语言的技术选型，都能从中获得极具价值的参考。此外，清单还延伸提供了免费书籍、在线课程、行业会议、技术博客及线下聚会等丰富资源，构建了从学习到实践的全链路支持体系。\n\n其独特亮点在于严格的维护标准：明确标记已停止维护或长期未更新的项目，确保推荐内容的时效性与可靠性。作为机器学习领域的“导航图”，awesome-machine-learning 以开源协作的方式持续更新，旨在降低技术探索门槛，让每一位从业者都能高效地站在巨人的肩膀上创新。",72149,"2026-04-03T21:50:24",[20,18],{"id":66,"github_repo":67,"name":68,"description_en":69,"description_zh":70,"ai_summary_zh":71,"readme_en":72,"readme_zh":73,"quickstart_zh":74,"use_case_zh":75,"hero_image_url":76,"owner_login":77,"owner_name":78,"owner_avatar_url":79,"owner_bio":80,"owner_company":81,"owner_location":82,"owner_email":83,"owner_twitter":83,"owner_website":83,"owner_url":84,"languages":85,"stars":90,"forks":91,"last_commit_at":92,"license":83,"difficulty_score":93,"env_os":94,"env_gpu":95,"env_ram":96,"env_deps":97,"category_tags":105,"github_topics":106,"view_count":10,"oss_zip_url":83,"oss_zip_packed_at":83,"status":22,"created_at":112,"updated_at":113,"faqs":114,"releases":125},7659,"robotlearning123\u002Fawesome-isaac-gym","awesome-isaac-gym","A curated list of awesome NVIDIA Issac Gym frameworks, papers, software, and resources","awesome-isaac-gym 是一个精心整理的资源合集，专为 NVIDIA Isaac Gym 生态打造。Isaac Gym 是英伟达推出的高性能机器人学习平台，其核心优势在于利用 GPU 进行大规模并行物理仿真，能同时在单个显卡上运行数千个机器人环境，从而极大加速强化学习训练过程。\n\n这份清单主要解决了研究人员和开发者在入门及使用过程中“资源分散、难以查找”的痛点。它系统性地汇集了官方文档、安装指南、基础教程、前沿学术论文（涵盖机械臂操作、足式机器人运动控制等方向）以及社区开发的工具库和实战案例。从最新的 Isaac Lab 框架到经典的 Sim-to-Real（仿真到现实）迁移研究，用户都能在此找到对应的链接与代码实现。\n\n该资源非常适合机器人领域的科研人员、算法工程师以及高校学生使用。无论是希望快速搭建实验环境的新手，还是追踪最新技术进展的资深专家，都能通过 awesome-isaac-gym 高效获取所需信息。其独特的价值在于不仅提供了基础学习资料，更持续更新包括 HumanoidVerse 在内的多模拟器框架和顶级会议论文复现项目，帮助用户紧跟基于 GPU 加速的机器人","awesome-isaac-gym 是一个精心整理的资源合集，专为 NVIDIA Isaac Gym 生态打造。Isaac Gym 是英伟达推出的高性能机器人学习平台，其核心优势在于利用 GPU 进行大规模并行物理仿真，能同时在单个显卡上运行数千个机器人环境，从而极大加速强化学习训练过程。\n\n这份清单主要解决了研究人员和开发者在入门及使用过程中“资源分散、难以查找”的痛点。它系统性地汇集了官方文档、安装指南、基础教程、前沿学术论文（涵盖机械臂操作、足式机器人运动控制等方向）以及社区开发的工具库和实战案例。从最新的 Isaac Lab 框架到经典的 Sim-to-Real（仿真到现实）迁移研究，用户都能在此找到对应的链接与代码实现。\n\n该资源非常适合机器人领域的科研人员、算法工程师以及高校学生使用。无论是希望快速搭建实验环境的新手，还是追踪最新技术进展的资深专家，都能通过 awesome-isaac-gym 高效获取所需信息。其独特的价值在于不仅提供了基础学习资料，更持续更新包括 HumanoidVerse 在内的多模拟器框架和顶级会议论文复现项目，帮助用户紧跟基于 GPU 加速的机器人学习技术前沿，降低探索成本，提升研发效率。","# Awesome NVIDIA Isaac Gym 🤖\n\n[![Awesome](https:\u002F\u002Fawesome.re\u002Fbadge.svg)](https:\u002F\u002Fawesome.re)\n\nA curated collection of resources related to **NVIDIA Isaac Gym**, a high-performance GPU-based physics simulation environment for robot learning.\n\n## 🎯 Quick Links\n\n- [Official Website](https:\u002F\u002Fdeveloper.nvidia.com\u002Fisaac-gym)\n\n- [Documentation](https:\u002F\u002Fdocs.nvidia.com\u002Fisaac\u002Fisaac\u002Fdoc\u002Findex.html)\n\n- [Community Forum](https:\u002F\u002Fforums.developer.nvidia.com\u002Fc\u002Fagx-autonomous-machines\u002Fisaac\u002Fisaac-gym\u002F322)\n\n- [Latest Release Info](#-latest-releases)\n\n---\n\n## 📋 Contents\n\n- [Latest Releases](#-latest-releases)\n\n- [Getting Started](#-getting-started)\n\n- [Official Resources](#-official-resources)\n\n- [Learning Materials](#-learning-materials)\n\n  - [Tutorials](#tutorials)\n\n  - [Workshops](#workshops)\n\n  - [Video Guides](#video-guides)\n\n- [Research Papers](#-research-papers)\n\n  - [Core Papers](#core-papers)\n\n  - [Robot Manipulation](#robot-manipulation)\n\n  - [Locomotion & Control](#locomotion--control)\n\n  - [Simulation & Learning](#simulation--learning)\n\n- [Tools & Libraries](#-tools--libraries)\n\n  - [RL Frameworks](#rl-frameworks)\n\n  - [Community Projects](#community-projects)\n\n- [Applications & Examples](#-applications--examples)\n\n- [Community Resources](#-community-resources)\n\n---\n\n## 🚀 Latest Releases\n\n- **March 2024**: HumanoidVerse - A Multi-simulator Framework for Humanoid Robot Learning ([GitHub](https:\u002F\u002Fgithub.com\u002FLeCAR-Lab\u002FHumanoidVerse))\n- **February 2024**: Isaac Lab - A unified and modular framework for robot learning ([Website](https:\u002F\u002Fisaac-sim.github.io\u002FIsaacLab\u002Fmain\u002Findex.html))\n- **February 2024**: PhysX 5 SDK release ([GitHub](https:\u002F\u002Fgithub.com\u002FNVIDIA-Omniverse\u002FPhysX))\n- **February 2022**: Isaac Gym Preview 4 (1.3.0)\n- **October 2021**: Isaac Gym Preview 3\n- **June 2021**: [NVIDIA Isaac Sim on Omniverse Open Beta](https:\u002F\u002Fdeveloper.nvidia.com\u002Fblog\u002Fnvidia-isaac-sim-on-omniverse-now-available-in-open-beta\u002F)\n\n- **March 23, 2022:** GTC 2022 Session — [Isaac Gym: The Next Generation — High-performance Reinforcement Learning in Omniverse](https:\u002F\u002Fwww.nvidia.com\u002Fgtc\u002Fsession-catalog\u002F?search=Isaac#\u002Fsession\u002F1638331324610001KvlV).\n\n- **Isaac Gym Overview:** [Isaac Gym Session](https:\u002F\u002Fwww.nvidia.com\u002Fen-us\u002Fon-demand\u002Fsession\u002Fgtcsiliconvalley2019-s9918\u002F).\n- **GTC Spring 2021:** [Isaac Gym: End-to-End GPU-Accelerated Reinforcement Learning](https:\u002F\u002Fwww.nvidia.com\u002Fen-us\u002Fon-demand\u002Fsession\u002Fgtcspring21-s32037\u002F).\n\n---\n\n## 🎓 Getting Started\n\n1. **Installation & Setup**\n\n   - [Official Isaac Gym Download](https:\u002F\u002Fdeveloper.nvidia.com\u002Fisaac-gym)\n\n   - [Quick Start Guide](https:\u002F\u002Fdocs.nvidia.com\u002Fisaac\u002Fisaac\u002Fdoc\u002Fsetup.html)\n\n   - [Environment Setup](https:\u002F\u002Fdocs.nvidia.com\u002Fisaac\u002Fisaac\u002Fdoc\u002Fsetup.html#environment-setup)\n\n2. **Basic Concepts**\n\n   - [Introduction to Isaac Gym](https:\u002F\u002Fdeveloper.nvidia.com\u002Fblog\u002Fintroducing-isaac-gym-rl-for-robotics\u002F)\n\n   - [transic](https:\u002F\u002Fgithub.com\u002Ftransic-robot\u002Ftransic): Official Implementation of \"TRANSIC: Sim-to-Real Policy Transfer by Learning from Online Correction\" CoRL 2024.\n\n   - [Robot Synesthesia](https:\u002F\u002Fgithub.com\u002FYingYuan0414\u002Fin-hand-rotation): Official Implementation of \"Robot Synesthesia: In-Hand Manipulation with Visuotactile Sensing\" ICRA 2024.\n\n   - [RLAfford](https:\u002F\u002Fgithub.com\u002Fhyperplane-lab\u002FRLAfford): Official Implementation of \"RLAfford: End-to-end Affordance Learning with Reinforcement Learning\" ICRA 2023.\n\n   - [Core Components Overview](https:\u002F\u002Fdocs.nvidia.com\u002Fisaac\u002Fisaac\u002Fdoc\u002Findex.html)\n\n   - [Basic Tutorials](https:\u002F\u002Fwww.youtube.com\u002Fplaylist?list=PL3jK4xNnlCVf1SzxjCm7ZxDBNl9QYyV8X)\n\n\n\n## 📚 Official Resources\n\n### Core Documentation\n\n- [Isaac SDK Documentation](https:\u002F\u002Fdocs.nvidia.com\u002Fisaac\u002Fisaac\u002Fdoc\u002Findex.html)\n- [OmniIsaacGymEnvs Repository](https:\u002F\u002Fgithub.com\u002FNVIDIA-Omniverse\u002FOmniIsaacGymEnvs)\n- [Official Blog Posts](https:\u002F\u002Fdeveloper.nvidia.com\u002Fblog\u002Ftag\u002Fisaac\u002F)\n\n### Learning Resources\n\n- [Video Tutorials](https:\u002F\u002Fwww.youtube.com\u002Fplaylist?list=PL3jK4xNnlCVf1SzxjCm7ZxDBNl9QYyV8X)\n- [Developer Blog](https:\u002F\u002Fdeveloper.nvidia.com\u002Fblog\u002Ftag\u002Fisaac\u002F)\n- [NVIDIA Omniverse Channel](https:\u002F\u002Fwww.youtube.com\u002Fc\u002FNVIDIAOmniverse)\n\n## 📖 Learning Materials\n\n### Tutorials\n\nComprehensive tutorial series from RSS 2021 Workshop:\n1. [Introduction & Getting Started](https:\u002F\u002Fyoutu.be\u002FnleDq-oJjGk)\n2. [Environments, Training & Tips](https:\u002F\u002Fyoutu.be\u002F1RSugmJ4_gs)\n3. Academic Labs Series:\n   - [University of Toronto](https:\u002F\u002Fyoutu.be\u002FnXM5_mwUFOI)\n   - [IMLab](https:\u002F\u002Fyoutu.be\u002FVrTVUpDM7K8)\n   - [Stanford University](https:\u002F\u002Fyoutu.be\u002FRhjRrUK2abs)\n   - [Soft-Body Simulation](https:\u002F\u002Fyoutu.be\u002Fi4fGVc6lImo)\n   - [ETH Zurich](https:\u002F\u002Fyoutu.be\u002FAfi17BnSuBM)\n4. [New Frontiers in GPU Accelerated RL](https:\u002F\u002Fyoutu.be\u002FWhaybakLTXE)\n5. [lycheeai-hub](https:\u002F\u002Flycheeai-hub.com\u002F)\n\n### Video Guides\n- [Robot Import Guide](https:\u002F\u002Fyoutu.be\u002FpxPFr58gHmQ)\n- [Simulator Basics](https:\u002F\u002Fwww.youtube.com\u002Fwatch?v=b12M_kCW82o)\n- [Advanced Features](https:\u002F\u002Fyoutu.be\u002FXcvMCs9NJfM)\n- [Community Demonstrations](https:\u002F\u002Fyoutu.be\u002FvpHR0qiH-GY)\n\n- **[From Point Clouds to Material Graphs: Explore the Latest in Omniverse Create 2021.3](https:\u002F\u002Fyoutu.be\u002Ft9nVWhnOgbE)**\n- **[Robot Autonomy with the Digital Twin in Isaac Sim](https:\u002F\u002Fyoutu.be\u002FvOEdzxR-_Iw)**\n- **[Can We Simulate a Real Robot?](https:\u002F\u002Fyoutu.be\u002FphTnbmXM06g)** — A journey through finding a high-quality physics simulator for a robot quadruped.\n- **[Teaching Robots to Walk with Reinforcement Learning](https:\u002F\u002Fyoutu.be\u002F6qbW7Ki9NUc)** — Robot simulation adventure, covering reinforcement learning with the Bittle robot.\n- **[Robot Dog Learns to Walk — Bittle Reinforcement Learning Part 3](https:\u002F\u002Fyoutu.be\u002FA0tPe7-R8z0)** — Further progress in training robot quadrupeds to walk.\n\n\n- **[Isaac Sim GTC 2021 — Sim-to-Real](https:\u002F\u002Fwww.nvidia.com\u002Fen-us\u002Fon-demand\u002Fsession\u002Fgtcspring21-s31824\u002F):** Session on sim-to-real transfer using Isaac Sim.\n- **[Isaac Sim Video Tutorials](https:\u002F\u002Fwww.youtube.com\u002Fplaylist?list=PL3jK4xNnlCVf1SzxjCm7ZxDBNl9QYyV8X):** Official video tutorials.\n- **[Training Your JetBot in NVIDIA Isaac Sim](https:\u002F\u002Fdeveloper.nvidia.com\u002Fblog\u002Ftraining-your-jetbot-in-isaac-sim\u002F):** Guide on training JetBot using Isaac Sim.\n- **[Training Your NVIDIA JetBot to Avoid Collisions Using NVIDIA Isaac Sim](https:\u002F\u002Fdeveloper.nvidia.com\u002Fblog\u002Ftraining-your-nvidia-jetbot-to-avoid-collisions-using-nvidia-isaac-sim\u002F):** Blog post on collision avoidance training.\n- **[Introducing NVIDIA Isaac Gym: End-to-End Reinforcement Learning for Robotics](https:\u002F\u002Fdeveloper.nvidia.com\u002Fblog\u002Fintroducing-isaac-gym-rl-for-robotics\u002F):** Introduction to Isaac Gym.\n- **[Accelerating Robotics Simulation with NVIDIA Omniverse Isaac Sim](https:\u002F\u002Fdeveloper.nvidia.com\u002Fblog\u002Faccelerating-robotics-simulation-with-nvidia-omniverse-isaac-sim\u002F):** Blog post on using Omniverse with Isaac Sim.\n- **[Developing Robotics Applications in Python with NVIDIA Isaac SDK](https:\u002F\u002Fdeveloper.nvidia.com\u002Fblog\u002Fdeveloping-robotics-applications-in-python-with-isaac-sdk\u002F):** Guide on using Isaac SDK with Python.\n- **[Building an Intelligent Robot Dog with the NVIDIA Isaac SDK](https:\u002F\u002Fdeveloper.nvidia.com\u002Fblog\u002Fbuilding-intelligent-robot-dog-with-isaac-sdk\u002F):** Tutorial on building a robot dog.\n- **[NVIDIA Omniverse YouTube Channel](https:\u002F\u002Fwww.youtube.com\u002Fc\u002FNVIDIAOmniverse\u002Fvideos?&ab_channel=NVIDIAOmniverse):** Official channel with various tutorials and demos.\n\n\n- [ICRA2021] Sim-to-Real for Robotic Tactile Sensing via Physics-Based Simulation and Learned Latent Projections: [paper](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2103.16747.pdf)\n\n- [2021] DeformerNet: A Deep Learning Approach to 3D Deformable Object Manipulation: [paper](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2107.08067.pdf)\n\n- [RSS2021_VLRR] A Simple Method for Complex In-Hand Manipulation: [paper](https:\u002F\u002Frssvlrr.github.io\u002Fpapers\u002F13_CameraReady_RSS2021_VLRR.pdf), [project](https:\u002F\u002Fsites.google.com\u002Fview\u002Fin-hand-reorientation)\n\n### Locomotion\n- [RSS2024] Agile But Safe: Learning Collision-Free High-Speed Legged Locomotion: [paper](https:\u002F\u002Farxiv.org\u002Fabs\u002F2401.17583), [code](https:\u002F\u002Fgithub.com\u002FLeCAR-Lab\u002FABS)\n\n- [RSS2022] Rapid Locomotion via Reinforcement Learning: [paper](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2205.02824), [openreview](https:\u002F\u002Fopenreview.net\u002Fforum?id=wK2fDDJ5VcF), [**code**](https:\u002F\u002Fgithub.com\u002FImprobable-AI\u002Frapid-locomotion-rl\u002Ftree\u002Fmain)\n\n- [CoRL2021] Learning to Walk in Minutes Using Massively Parallel Deep Reinforcement Learning: [paper](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2109.11978.pdf), [openreview](https:\u002F\u002Fopenreview.net\u002Fforum?id=wK2fDDJ5VcF), [**code**](https:\u002F\u002Fgithub.com\u002Fleggedrobotics\u002Flegged_gym), [project](https:\u002F\u002Fleggedrobotics.github.io\u002Flegged_gym\u002F)\n\n- [ICRA2021] Dynamics Randomization Revisited:A Case Study for Quadrupedal Locomotion: [project](https:\u002F\u002Fwww.pair.toronto.edu\u002Funderstanding-dr\u002F), [paper](https:\u002F\u002Farxiv.org\u002Fabs\u002F2011.02404), [video](https:\u002F\u002Fyoutu.be\u002FckdHWWpfSpk)\n\n- [2021] GLiDE: Generalizable Quadrupedal Locomotion in Diverse Environments with a Centroidal Model: [project](https:\u002F\u002Fwww.pair.toronto.edu\u002Fglide-quadruped\u002F), [paper](https:\u002F\u002Farxiv.org\u002Fabs\u002F2104.09771)\n\n- [CoRL2020] Learning a Contact-Adaptive Controller for Robust, Efficient Legged Locomotion: [paper](https:\u002F\u002Farxiv.org\u002Fabs\u002F2009.10019), [video](https:\u002F\u002Fyoutu.be\u002FJJOmFZKpYTo), [project](https:\u002F\u002Fsites.google.com\u002Fview\u002Flearn-contact-controller\u002Fhome), [blog](https:\u002F\u002Fdeveloper.nvidia.com\u002Fblog\u002Fcontact-adaptive-controller-locomotion\u002F)\n\n- [RAL2021] Learning a State Representation and Navigation in Cluttered and Dynamic Environments: [paper](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2103.04351.pdf)\n\n- **[HumanoidVerse](https:\u002F\u002Fgithub.com\u002FLeCAR-Lab\u002FHumanoidVerse):** A Multi-simulator Framework for Humanoid Robot Learning (2024)\n  - Features multi-simulator support (Isaac Gym, Flex, MuJoCo)\n  - Includes diverse humanoid models and environments\n  - Provides comprehensive benchmarking tools\n  - Enables efficient parallel training across simulators\n\n- **[HIMLoco](https:\u002F\u002Fjunfeng-long.github.io\u002FHIMLoco\u002F):** Hierarchical Imitation Learning for Robust Humanoid Locomotion (2024)\n  - [Paper](https:\u002F\u002Farxiv.org\u002Fabs\u002F2403.14864)\n  - Features hierarchical imitation learning framework\n  - Enables robust humanoid locomotion in challenging environments\n  - Demonstrates successful real-world deployment\n\n- **[ASAP](https:\u002F\u002Fagile.human2humanoid.com\u002F):** Aligning Simulation and Real-World Physics for Learning Agile Humanoid Whole-Body Skills (2025)\n  - [Paper](https:\u002F\u002Farxiv.org\u002Fabs\u002F2502.01143)\n  - Features two-stage framework for sim-to-real transfer\n  - Enables highly agile humanoid motions like kicks and jumps\n  - Successfully deployed on real Unitree G1 humanoid robot\n  - Demonstrates significant improvement over SysID and DR baselines\n\n### Blogs\n\n- **[A Brief Introduction to NVIDIA Omniverse](https:\u002F\u002Fzhuanlan.zhihu.com\u002Fp\u002F462305733)**\n\n---\n\n## 📑 Research Papers\n\n### Core Papers\n- [Isaac Gym: High Performance GPU-Based Physics Simulation](https:\u002F\u002Farxiv.org\u002Fabs\u002F2108.10470) (NeurIPS 2021)\n  - [Project Page](https:\u002F\u002Fsites.google.com\u002Fview\u002Fisaacgym-nvidia)\n  - [OpenReview Discussion](https:\u002F\u002Fopenreview.net\u002Fforum?id=fgFBtYgJQX_)\n\n### Robot Manipulation\n\n- **[RLAfford](https:\u002F\u002Fgithub.com\u002Fhyperplane-lab\u002FRLAfford):** Official implementation of \"RLAfford: End-to-end Affordance Learning with Reinforcement Learning\", ICRA 2023.\n- **[Masked Visual Pre-training for Robotics (MVP)](https:\u002F\u002Fgithub.com\u002Fir413\u002Fmvp):** Repository for the MVP project.\n- **[Factory: Fast Contact for Robotic Assembly](https:\u002F\u002Fsites.google.com\u002Fnvidia.com\u002Ffactory):** RSS 2022.\n  - [Paper](http:\u002F\u002Fdoi.acm.org\u002F10.1145\u002F3450626.3459670)\n  - [Code](https:\u002F\u002Fgithub.com\u002FNVIDIA-Omniverse\u002FIsaacGymEnvs)\n- **[ASE: Adversarial Skill Embeddings](https:\u002F\u002Fnv-tlabs.github.io\u002FASE\u002F):** SIGGRAPH 2022.\n  - [Paper](https:\u002F\u002Farxiv.org\u002Fabs\u002F2205.01906)\n  - [Code](https:\u002F\u002Fgithub.com\u002FNVIDIA-Omniverse\u002FIsaacGymEnvs)\n- **[Data-Driven Operational Space Control (OSCAR)](https:\u002F\u002Fcremebrule.github.io\u002Foscar-web\u002F):** Adaptive and robust robot manipulation.\n  - [Paper](https:\u002F\u002Farxiv.org\u002Fabs\u002F2110.00704)\n  - [Code](https:\u002F\u002Fgithub.com\u002Fnvlabs\u002Foscar)\n- **[DefGraspSim](https:\u002F\u002Fsites.google.com\u002Fnvidia.com\u002Fdefgraspsim):** Simulation-based grasping of deformable objects.\n  - [Paper](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2107.05778.pdf)\n  - [Video](https:\u002F\u002Fyoutu.be\u002FCaj0AtsKKVI)\n  - [Code](https:\u002F\u002Fgithub.com\u002FNVlabs\u002Fdeformable_object_grasping)\n- **[In-Hand Object Pose Tracking](https:\u002F\u002Fsites.google.com\u002Fview\u002Fin-hand-object-pose-tracking\u002F):** ICRA 2021.\n  - [Paper](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2002.12160.pdf)\n- **[STORM: Fast Joint-Space MPC for Reactive Manipulation](https:\u002F\u002Fsites.google.com\u002Fview\u002Fmanipulation-mpc):** CoRL 2021.\n  - [Paper](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2104.13542.pdf)\n  - [Code](https:\u002F\u002Fgithub.com\u002FNVlabs\u002Fstorm)\n- **[Transferring Dexterous Manipulation from GPU Simulation to Real-World TriFinger](https:\u002F\u002Fs2r2-ig.github.io\u002F):**\n  - [Paper](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2108.09779.pdf)\n  - [Code](https:\u002F\u002Fgithub.com\u002Fpairlab\u002Fleibnizgym)\n- **[Causal Reasoning in Simulation for Robot Manipulation Policies](https:\u002F\u002Fsites.google.com\u002Fview\u002Fcrest-causal-struct-xfer-manip):** ICRA 2021.\n  - [Paper](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2103.16772.pdf)\n- **[Reactive Long Horizon Task Execution](https:\u002F\u002Fwww.youtube.com\u002Fplaylist?list=PL-oD0xHUngeLfQmpngYkGFZarstfPOXqX):** IROS 2021.\n  - [Paper](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2011.08694.pdf)\n- **[RoboDuet](https:\u002F\u002Flocomanip-duet.github.io\u002F):** Learning a Cooperative Policy for Whole-body Legged Loco-Manipulation (2024)\n  - [Paper](https:\u002F\u002Farxiv.org\u002Fabs\u002F2403.17367)\n  - [Code](https:\u002F\u002Fgithub.com\u002FRobo-Duet\u002FRoboDuet)\n  - Features two-policy framework for locomotion and manipulation\n  - Achieves 23% improvement in success rate for mobile manipulation tasks\n  - Zero-shot transfer from Go1+ARX5 to Go2+ARX5\n\n- **[Kitchen Worlds](https:\u002F\u002Fgithub.com\u002FLearning-and-Intelligent-Systems\u002Fkitchen-worlds):** Long-horizon Task-and-Motion-Planning (TAMP) in Kitchen Scenes (2024)\n  - [Paper](https:\u002F\u002Farxiv.org\u002Fabs\u002F2410.02193)\n  - [Code](https:\u002F\u002Fgithub.com\u002FLearning-and-Intelligent-Systems\u002Fkitchen-worlds)\n  - Library of TAMP problems in kitchen and household scenes\n  - Integrates VLM (e.g. GPT-4v) guidance for planning\n  - Includes procedural scene and trajectory generation\n\n### Localization & Control\n\n- **[HumanoidVerse](https:\u002F\u002Fgithub.com\u002FLeCAR-Lab\u002FHumanoidVerse):** A Multi-simulator Framework for Humanoid Robot Learning (2024)\n  - Features multi-simulator support (Isaac Gym, Flex, MuJoCo)\n  - Includes diverse humanoid models and environments\n  - Provides comprehensive benchmarking tools\n  - Enables efficient parallel training across simulators\n\n- **[HIMLoco](https:\u002F\u002Fjunfeng-long.github.io\u002FHIMLoco\u002F):** Hierarchical Imitation Learning for Robust Humanoid Locomotion (2024)\n  - [Paper](https:\u002F\u002Farxiv.org\u002Fabs\u002F2403.14864)\n  - Features hierarchical imitation learning framework\n  - Enables robust humanoid locomotion in challenging environments\n  - Demonstrates successful real-world deployment\n\n- **[ASAP](https:\u002F\u002Fagile.human2humanoid.com\u002F):** Aligning Simulation and Real-World Physics for Learning Agile Humanoid Whole-Body Skills (2025)\n  - [Paper](https:\u002F\u002Farxiv.org\u002Fabs\u002F2502.01143)\n  - Features two-stage framework for sim-to-real transfer\n  - Enables highly agile humanoid motions like kicks and jumps\n  - Successfully deployed on real Unitree G1 humanoid robot\n  - Demonstrates significant improvement over SysID and DR baselines\n\n- **[Learning to Walk in Minutes Using Massively Parallel Deep RL](https:\u002F\u002Fleggedrobotics.github.io\u002Flegged_gym\u002F):** CoRL 2021.\n  - [Paper](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2109.11978.pdf)\n  - [Code](https:\u002F\u002Fgithub.com\u002Fleggedrobotics\u002Flegged_gym)\n\n- **[Dynamics Randomization Revisited](https:\u002F\u002Fwww.pair.toronto.edu\u002Funderstanding-dr\u002F):** A case study for quadrupedal locomotion.\n  - [Paper](https:\u002F\u002Farxiv.org\u002Fabs\u002F2011.02404)\n  - [Video](https:\u002F\u002Fyoutu.be\u002FckdHWWpfSpk)\n\n- **[GLiDE: Generalizable Quadrupedal Locomotion](https:\u002F\u002Fwww.pair.toronto.edu\u002Fglide-quadruped\u002F):**\n  - [Paper](https:\u002F\u002Farxiv.org\u002Fabs\u002F2104.09771)\n\n- **[Learning a Contact-Adaptive Controller](https:\u002F\u002Fsites.google.com\u002Fview\u002Flearn-contact-controller\u002Fhome):** For robust, efficient legged locomotion.\n  - [Paper](https:\u002F\u002Farxiv.org\u002Fabs\u002F2009.10019)\n  - [Video](https:\u002F\u002Fyoutu.be\u002FJJOmFZKpYTo)\n  - [Blog](https:\u002F\u002Fdeveloper.nvidia.com\u002Fblog\u002Fcontact-adaptive-controller-locomotion\u002F)\n\n- **[Learning a State Representation and Navigation](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2103.04351.pdf):** In cluttered and dynamic environments.\n\n### Others\n\n- **[BayesSimIG](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2107.04527.pdf):** Scalable parameter inference for adaptive domain randomization with Isaac Gym.\n  - [Code](https:\u002F\u002Fgithub.com\u002FNVlabs\u002Fbayes-sim-ig)\n- **[Isaac Gym: High Performance GPU-Based Physics Simulation](https:\u002F\u002Fsites.google.com\u002Fview\u002Fisaacgym-nvidia):** NeurIPS 2021.\n  - [Paper](https:\u002F\u002Farxiv.org\u002Fabs\u002F2108.10470)\n  - [OpenReview](https:\u002F\u002Fopenreview.net\u002Fforum?id=fgFBtYgJQX_)\n- **[Learning to Swim](https:\u002F\u002Farxiv.org\u002Fabs\u002F2410.00120v1):** Reinforcement learning for 6-DOF control of thruster-driven AUVs.\n- **[MarineGym: Accelerated Training for Underwater Vehicles with High-Fidelity RL Simulation](https:\u002F\u002Farxiv.org\u002Fabs\u002F2410.14117):** Based on Issac Sim\n- **[space_robotics_bench](https:\u002F\u002Fgithub.com\u002FAndrejOrsula\u002Fspace_robotics_bench)** Space Robotics Bench\n- **[Humanoid-Gym](https:\u002F\u002Fgithub.com\u002Froboterax\u002Fhumanoid-gym):** Humanoid-Gym: Reinforcement Learning for Humanoid Robot with Zero-Shot Sim2Real Transfer\n\n\n## 🛠 Tools & Libraries\n\n### RL Frameworks\n\n- [RL Games](https:\u002F\u002Fgithub.com\u002FDenys88\u002Frl_games) - Compatible RL algorithms\n- [ElegantRL](https:\u002F\u002Fgithub.com\u002FAI4Finance-Foundation\u002FElegantRL)\n- [skrl](https:\u002F\u002Fgithub.com\u002FToni-SM\u002Fskrl) - Modular RL library\n- [Minimal Stable PPO](https:\u002F\u002Fgithub.com\u002FToruOwO\u002Fminimal-stable-PPO)\n\n\n- [skrl](https:\u002F\u002Fgithub.com\u002FToni-SM\u002Fskrl), [paper](https:\u002F\u002Farxiv.org\u002Fabs\u002F2202.03825)\n  \n- [RSL RL](https:\u002F\u002Fgithub.com\u002Fleggedrobotics\u002Frsl_rl)\n\n## Related GitHub Repos\n\n- [IsaacGymEnvs](https:\u002F\u002Fgithub.com\u002FNVIDIA-Omniverse\u002FIsaacGymEnvs)\n\n- [isaacgym_hammering](https:\u002F\u002Fgithub.com\u002FLiCHOTHU\u002Fisaacgym_hammering)\n\n- [isaacgym-utils](https:\u002F\u002Fgithub.com\u002Fiamlab-cmu\u002Fisaacgym-utils): Developed by the CMU Intelligent Autonomous Manipulation Lab\n\n- [isaacgym_sandbox](https:\u002F\u002Fgithub.com\u002Fkploeger\u002Fisaacgym_sandbox)\n\n- [thormang3-gogoro-PPO](https:\u002F\u002Fgithub.com\u002Fguichristmann\u002Fthormang3-gogoro-PPO): Steering-based control of a two-wheeled vehicle using RL-PPO and NVIDIA Isaac Gym\n\n- [dvrk_IssacGym](https:\u002F\u002Fgithub.com\u002Fbaotruyenthach\u002Fdvrk_IssacGym), [link](https:\u002F\u002Fgithub.com\u002Fbaotruyenthach\u002Fdvrk_grasp_pipeline_isaacgym)\n\n\n### Community Projects\n\n- **[IsaacGymEnvs](https:\u002F\u002Fgithub.com\u002FNVIDIA-Omniverse\u002FIsaacGymEnvs):** Official Isaac Gym RL environments.\n- **[isaacgym_hammering](https:\u002F\u002Fgithub.com\u002FLiCHOTHU\u002Fisaacgym_hammering):** Hammering task implementation.\n- **[isaacgym-utils](https:\u002F\u002Fgithub.com\u002Fiamlab-cmu\u002Fisaacgym-utils):** Utilities by CMU's Intelligent Autonomous Manipulation Lab.\n- **[isaacgym_sandbox](https:\u002F\u002Fgithub.com\u002Fkploeger\u002Fisaacgym_sandbox):** Sandbox for Isaac Gym experiments.\n- **[thormang3-gogoro-PPO](https:\u002F\u002Fgithub.com\u002Fguichristmann\u002Fthormang3-gogoro-PPO):** Two-wheeled vehicle control using PPO.\n- **[Bez_IsaacGym](https:\u002F\u002Fgithub.com\u002Futra-robosoccer\u002FBez_IsaacGym):** Environments for humanoid robot Bez.\n- **[DexterousHands](https:\u002F\u002Fgithub.com\u002FPKU-MARL\u002FDexterousHands):** Dual dexterous hand manipulation tasks.\n- **[legged_gym_isaac](https:\u002F\u002Fgithub.com\u002Fchengxuxin\u002Flegged_gym_isaac):** Legged robots in Isaac Gym.\n- **[shifu](https:\u002F\u002Fgithub.com\u002F42jaylonw\u002Fshifu):** Environment builder for any robot.\n- **[Rofunc](https:\u002F\u002Fgithub.com\u002FSkylark0924\u002FRofunc):** Python package for robot learning from demonstration.\n- **[Dofbot Reacher](https:\u002F\u002Fgithub.com\u002Fj3soon\u002FOmniIsaacGymEnvs-DofbotReacher):** Sim2Real environment for Dofbot.\n- **[UR10 Reacher](https:\u002F\u002Fgithub.com\u002Fj3soon\u002FOmniIsaacGymEnvs-UR10Reacher):** Sim2Real environment for UR10.\n- **[TimeChamber](https:\u002F\u002Fgithub.com\u002Finspirai\u002FTimeChamber):** Massively parallel self-play framework.\n- **[RL-MPC-Locomotion](https:\u002F\u002Fgithub.com\u002Fsilvery107\u002Frl-mpc-locomotion):** Deep RL for quadruped locomotion.\n- **[Isaac_Underwater](https:\u002F\u002Fgithub.com\u002Fleonlime\u002Fisaac_underwater):** Water and underwater tests using NVIDIA Isaac Sim.\n- **[VRKitchen2.0-IndoorKit](https:\u002F\u002Fgithub.com\u002Fyizhouzhao\u002FVRKitchen2.0-IndoorKit):** Omniverse IndoorKit Extension.\n- **[agibot_x1_train](https:\u002F\u002Fgithub.com\u002FAgibotTech\u002Fagibot_x1_train):** The reinforcement learning training code for AgiBot X1.\n- \n\n---\n\n## Conference Sessions and Talks\n\n- **[Isaac Gym and Omniverse: High Performance Reinforcement Learning Evolved [A31118]](https:\u002F\u002Fevents.rainfocus.com\u002Fwidget\u002Fnvidia\u002Fnvidiagtc\u002Fsessioncatalog?search=A31118)**\n- **[Learning Challenging Tasks for Quadrupedal Robots: From Simulation to Reality [A31308]](https:\u002F\u002Fevents.rainfocus.com\u002Fwidget\u002Fnvidia\u002Fnvidiagtc\u002Fsessioncatalog?search=A31308)**\n- **[Sim-to-Real in Isaac Sim](https:\u002F\u002Fwww.nvidia.com\u002Fen-us\u002Fon-demand\u002Fsession\u002Fgtcspring21-s31824\u002F)**\n- **[Isaac Gym: End-to-End GPU-Accelerated Reinforcement Learning](https:\u002F\u002Fwww.nvidia.com\u002Fen-us\u002Fon-demand\u002Fsession\u002Fgtcspring21-s32037\u002F)**\n- **[Bridging Sim2Real Gap: Simulation Tuning for Training Deep Learning Robotic Perception Models](https:\u002F\u002Fwww.nvidia.com\u002Fen-us\u002Fon-demand\u002Fsession\u002Fgtcspring21-s31649\u002F)**\n- **[Reinforcement Learning and Intralogistics](https:\u002F\u002Fwww.nvidia.com\u002Fen-us\u002Fon-demand\u002Fsession\u002Fgtcspring21-e31467\u002F)**\n- **[Building Robotics Applications Using NVIDIA Isaac SDK](https:\u002F\u002Fwww.nvidia.com\u002Fen-us\u002Fon-demand\u002Fsession\u002Fgtcfall20-a21856\u002F)**\n- **[NVIDIA Isaac Sim — Amazing Robot Models and Tasks](https:\u002F\u002Fwww.nvidia.com\u002Fen-us\u002Fon-demand\u002Fsession\u002Fgtcsj20-d2s43\u002F)**\n- **[Omniverse View 2021.2 — Application Tour](https:\u002F\u002Fwww.nvidia.com\u002Fen-us\u002Fon-demand\u002Fsession\u002Fomniverse2020-om1315\u002F)**\n- **[ISAAC SIM Introduction and Live Demo](https:\u002F\u002Fwww.nvidia.com\u002Fen-us\u002Fon-demand\u002Fsession\u002Fomniverse2020-om1314\u002F)**\n- **[NVIDIA On-Demand ISAAC SIM Sessions](https:\u002F\u002Fwww.nvidia.com\u002Fen-us\u002Fon-demand\u002Fsearch\u002F?facet.mimetype[]=event%20session&layout=list&page=1&q=isaac%20sim&sort=relevance)**\n\n---\n\n## 🤖 Automated Research Updates\n\n- Source: arXiv queries for \"isaac gym\", \"omni isaac\", and \"isaac lab\".\n- Schedule: **Daily at 09:00 UTC** or manual via Actions \"Research Bot\".\n- Output: updates content between the 'research-bot:start' and 'research-bot:end' markers in this README.\n- Safety: opens a Draft PR labeled \"needs-approval\"; nothing merges automatically.\n- Config: edit `.research-bot.yaml` to adjust queries and limits.\n\n## 🌟 Contributing\n\nContributions are welcome! Please read our [contribution guidelines](CONTRIBUTING.md) before submitting a pull request.\n\n## 📄 License\n\nThis repository is licensed under the MIT License - see the [LICENSE](LICENSE) file for details.\n\n## 🙏 Acknowledgments\n\nSpecial thanks to all contributors and the NVIDIA Isaac team for making these resources available to the robotics community.\n\n## 🧠 Latest Research (auto-updated)\n\n\u003C!-- research-bot:start -->\n- [HumanoidVerse: A Versatile Humanoid for Vision-Language Guided\n  Multi-Object Rearrangement](https:\u002F\u002Farxiv.org\u002Fabs\u002F2508.16943) — Haozhuo Zhang, Jingkai Sun, Michele Caprio, et al. (2025-08-23) [pdf](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2508.16943.pdf)\n\u003C!-- research-bot:end -->\n","# 令人惊叹的 NVIDIA Isaac Gym 🤖\n\n[![Awesome](https:\u002F\u002Fawesome.re\u002Fbadge.svg)](https:\u002F\u002Fawesome.re)\n\n这是一份精心整理的资源合集，专注于 **NVIDIA Isaac Gym**——一个基于 GPU 的高性能物理仿真环境，专为机器人学习而设计。\n\n## 🎯 快速链接\n\n- [官方网站](https:\u002F\u002Fdeveloper.nvidia.com\u002Fisaac-gym)\n\n- [文档](https:\u002F\u002Fdocs.nvidia.com\u002Fisaac\u002Fisaac\u002Fdoc\u002Findex.html)\n\n- [社区论坛](https:\u002F\u002Fforums.developer.nvidia.com\u002Fc\u002Fagx-autonomous-machines\u002Fisaac\u002Fisaac-gym\u002F322)\n\n- [最新发布信息](#-latest-releases)\n\n---\n\n## 📋 目录\n\n- [最新发布](#-latest-releases)\n\n- [入门指南](#-getting-started)\n\n- [官方资源](#-official-resources)\n\n- [学习资料](#-learning-materials)\n\n  - [教程](#tutorials)\n\n  - [研讨会](#workshops)\n\n  - [视频指南](#video-guides)\n\n- [研究论文](#-research-papers)\n\n  - [核心论文](#core-papers)\n\n  - [机器人操作](#robot-manipulation)\n\n  - [运动与控制](#locomotion--control)\n\n  - [仿真与学习](#simulation--learning)\n\n- [工具与库](#-tools--libraries)\n\n  - [强化学习框架](#rl-frameworks)\n\n  - [社区项目](#community-projects)\n\n- [应用与示例](#-applications--examples)\n\n- [社区资源](#-community-resources)\n\n---\n\n## 🚀 最新发布\n\n- **2024年3月**: HumanoidVerse - 用于人形机器人学习的多仿真器框架 ([GitHub](https:\u002F\u002Fgithub.com\u002FLeCAR-Lab\u002FHumanoidVerse))\n- **2024年2月**: Isaac Lab - 一个统一且模块化的机器人学习框架 ([官网](https:\u002F\u002Fisaac-sim.github.io\u002FIsaacLab\u002Fmain\u002Findex.html))\n- **2024年2月**: PhysX 5 SDK 发布 ([GitHub](https:\u002F\u002Fgithub.com\u002FNVIDIA-Omniverse\u002FPhysX))\n- **2022年2月**: Isaac Gym 预览版 4 (1.3.0)\n- **2021年10月**: Isaac Gym 预览版 3\n- **2021年6月**: [NVIDIA Isaac Sim 在 Omniverse 上的公开测试版](https:\u002F\u002Fdeveloper.nvidia.com\u002Fblog\u002Fnvidia-isaac-sim-on-omniverse-now-available-in-open-beta\u002F)\n\n- **2022年3月23日:** GTC 2022 会议 — [Isaac Gym: 下一代 — Omniverse 中的高性能强化学习](https:\u002F\u002Fwww.nvidia.com\u002Fgtc\u002Fsession-catalog\u002F?search=Isaac#\u002Fsession\u002F1638331324610001KvlV)。\n\n- **Isaac Gym 概述:** [Isaac Gym 会话](https:\u002F\u002Fwww.nvidia.com\u002Fen-us\u002Fon-demand\u002Fsession\u002Fgtcsiliconvalley2019-s9918\u002F)。\n- **GTC 春季 2021:** [Isaac Gym: 端到端 GPU 加速的强化学习](https:\u002F\u002Fwww.nvidia.com\u002Fen-us\u002Fon-demand\u002Fsession\u002Fgtcspring21-s32037\u002F)。\n\n---\n\n## 🎓 入门指南\n\n1. **安装与设置**\n\n   - [官方 Isaac Gym 下载](https:\u002F\u002Fdeveloper.nvidia.com\u002Fisaac-gym)\n\n   - [快速入门指南](https:\u002F\u002Fdocs.nvidia.com\u002Fisaac\u002Fisaac\u002Fdoc\u002Fsetup.html)\n\n   - [环境设置](https:\u002F\u002Fdocs.nvidia.com\u002Fisaac\u002Fisaac\u002Fdoc\u002Fsetup.html#environment-setup)\n\n2. **基本概念**\n\n   - [Isaac Gym 简介](https:\u002F\u002Fdeveloper.nvidia.com\u002Fblog\u002Fintroducing-isaac-gym-rl-for-robotics\u002F)\n\n   - [transic](https:\u002F\u002Fgithub.com\u002Ftransic-robot\u002Ftransic): “TRANSIC：通过在线校正进行从仿真到现实的策略迁移” CoRL 2024 的官方实现。\n\n   - [机器人共感觉](https:\u002F\u002Fgithub.com\u002FYingYuan0414\u002Fin-hand-rotation): “机器人共感觉：基于视觉触觉感知的手中操作” ICRA 2024 的官方实现。\n\n   - [RLAfford](https:\u002F\u002Fgithub.com\u002Fhyperplane-lab\u002FRLAfford): “RLAfford：端到端的强化学习式效用学习” ICRA 2023 的官方实现。\n\n   - [核心组件概述](https:\u002F\u002Fdocs.nvidia.com\u002Fisaac\u002Fisaac\u002Fdoc\u002Findex.html)\n\n   - [基础教程](https:\u002F\u002Fwww.youtube.com\u002Fplaylist?list=PL3jK4xNnlCVf1SzxjCm7ZxDBNl9QYyV8X)\n\n\n\n## 📚 官方资源\n\n### 核心文档\n\n- [Isaac SDK 文档](https:\u002F\u002Fdocs.nvidia.com\u002Fisaac\u002Fisaac\u002Fdoc\u002Findex.html)\n- [OmniIsaacGymEnvs 仓库](https:\u002F\u002Fgithub.com\u002FNVIDIA-Omniverse\u002FOmniIsaacGymEnvs)\n- [官方博客文章](https:\u002F\u002Fdeveloper.nvidia.com\u002Fblog\u002Ftag\u002Fisaac\u002F)\n\n### 学习资源\n\n- [视频教程](https:\u002F\u002Fwww.youtube.com\u002Fplaylist?list=PL3jK4xNnlCVf1SzxjCm7ZxDBNl9QYyV8X)\n- [开发者博客](https:\u002F\u002Fdeveloper.nvidia.com\u002Fblog\u002Ftag\u002Fisaac\u002F)\n- [NVIDIA Omniverse 频道](https:\u002F\u002Fwww.youtube.com\u002Fc\u002FNVIDIAOmniverse)\n\n## 📖 学习资料\n\n### 教程\n\n来自 RSS 2021 研讨会的综合教程系列：\n1. [简介与入门](https:\u002F\u002Fyoutu.be\u002FnleDq-oJjGk)\n2. [环境、训练与技巧](https:\u002F\u002Fyoutu.be\u002F1RSugmJ4_gs)\n3. 学术实验室系列：\n   - [多伦多大学](https:\u002F\u002Fyoutu.be\u002FnXM5_mwUFOI)\n   - [IMLab](https:\u002F\u002Fyoutu.be\u002FVrTVUpDM7K8)\n   - [斯坦福大学](https:\u002F\u002Fyoutu.be\u002FRhjRrUK2abs)\n   - [软体仿真](https:\u002F\u002Fyoutu.be\u002Fi4fGVc6lImo)\n   - [苏黎世联邦理工学院](https:\u002F\u002Fyoutu.be\u002FAfi17BnSuBM)\n4. [GPU 加速强化学习的新前沿](https:\u002F\u002Fyoutu.be\u002FWhaybakLTXE)\n5. [lycheeai-hub](https:\u002F\u002Flycheeai-hub.com\u002F)\n\n### 视频指南\n- [机器人导入指南](https:\u002F\u002Fyoutu.be\u002FpxPFr58gHmQ)\n- [模拟器基础](https:\u002F\u002Fwww.youtube.com\u002Fwatch?v=b12M_kCW82o)\n- [高级功能](https:\u002F\u002Fyoutu.be\u002FXcvMCs9NJfM)\n- [社区演示](https:\u002F\u002Fyoutu.be\u002FvpHR0qiH-GY)\n\n- **[从点云到材质图：探索 Omniverse Create 2021.3 的最新进展](https:\u002F\u002Fyoutu.be\u002Ft9nVWhnOgbE)**\n- **[使用 Isaac Sim 中的数字孪生实现机器人自主性](https:\u002F\u002Fyoutu.be\u002FvOEdzxR-_Iw)**\n- **[我们能模拟一台真实的机器人吗？](https:\u002F\u002Fyoutu.be\u002FphTnbmXM06g)** — 一段寻找高质量物理模拟器来模拟四足机器人的旅程。\n- **[通过强化学习教机器人行走](https:\u002F\u002Fyoutu.be\u002F6qbW7Ki9NUc)** — 一场关于机器人模拟的冒险，内容涵盖使用 Bittle 机器人进行强化学习。\n- **[机器狗学会走路 — Bittle 强化学习第 3 部分](https:\u002F\u002Fyoutu.be\u002FA0tPe7-R8z0)** — 在训练四足机器人行走方面取得进一步进展。\n\n\n- **[Isaac Sim GTC 2021 — 从仿真到现实](https:\u002F\u002Fwww.nvidia.com\u002Fen-us\u002Fon-demand\u002Fsession\u002Fgtcspring21-s31824\u002F):** 使用 Isaac Sim 进行仿真到现实迁移的专题讨论。\n- **[Isaac Sim 视频教程](https:\u002F\u002Fwww.youtube.com\u002Fplaylist?list=PL3jK4xNnlCVf1SzxjCm7ZxDBNl9QYyV8X):** 官方视频教程。\n- **[在 NVIDIA Isaac Sim 中训练 JetBot](https:\u002F\u002Fdeveloper.nvidia.com\u002Fblog\u002Ftraining-your-jetbot-in-isaac-sim\u002F):** 关于使用 Isaac Sim 训练 JetBot 的指南。\n- **[使用 NVIDIA Isaac Sim 训练 NVIDIA JetBot 避免碰撞](https:\u002F\u002Fdeveloper.nvidia.com\u002Fblog\u002Ftraining-your-nvidia-jetbot-to-avoid-collisions-using-nvidia-isaac-sim\u002F):** 关于碰撞避免训练的博客文章。\n- **[介绍 NVIDIA Isaac Gym：面向机器人技术的端到端强化学习](https:\u002F\u002Fdeveloper.nvidia.com\u002Fblog\u002Fintroducing-isaac-gym-rl-for-robotics\u002F):** Isaac Gym 的介绍。\n- **[借助 NVIDIA Omniverse Isaac Sim 加速机器人仿真](https:\u002F\u002Fdeveloper.nvidia.com\u002Fblog\u002Faccelerating-robotics-simulation-with-nvidia-omniverse-isaac-sim\u002F):** 关于将 Omniverse 与 Isaac Sim 结合使用的博客文章。\n- **[使用 NVIDIA Isaac SDK 用 Python 开发机器人应用](https:\u002F\u002Fdeveloper.nvidia.com\u002Fblog\u002Fdeveloping-robotics-applications-in-python-with-isaac-sdk\u002F):** 关于使用 Isaac SDK 与 Python 结合的指南。\n- **[使用 NVIDIA Isaac SDK 构建智能机器狗](https:\u002F\u002Fdeveloper.nvidia.com\u002Fblog\u002Fbuilding-intelligent-robot-dog-with-isaac-sdk\u002F):** 构建机器狗的教程。\n- **[NVIDIA Omniverse YouTube 频道](https:\u002F\u002Fwww.youtube.com\u002Fc\u002FNVIDIAOmniverse\u002Fvideos?&ab_channel=NVIDIAOmniverse):** 官方频道，提供各类教程和演示。\n\n\n- [ICRA2021] 基于物理的仿真和学习的潜在投影用于机器人触觉感知的仿真到现实迁移：[论文](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2103.16747.pdf)\n\n- [2021] DeformerNet：一种用于 3D 可变形物体操作的深度学习方法：[论文](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2107.08067.pdf)\n\n- [RSS2021_VLRR] 一种用于复杂手中操纵的简单方法：[论文](https:\u002F\u002Frssvlrr.github.io\u002Fpapers\u002F13_CameraReady_RSS2021_VLRR.pdf)，[项目](https:\u002F\u002Fsites.google.com\u002Fview\u002Fin-hand-reorientation)\n\n### 行走运动\n- [RSS2024] 敏捷而安全：学习无碰撞的高速腿部行走运动：[论文](https:\u002F\u002Farxiv.org\u002Fabs\u002F2401.17583)，[代码](https:\u002F\u002Fgithub.com\u002FLeCAR-Lab\u002FABS)\n\n- [RSS2022] 通过强化学习实现快速行走：[论文](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2205.02824)，[OpenReview](https:\u002F\u002Fopenreview.net\u002Fforum?id=wK2fDDJ5VcF)，[**代码**](https:\u002F\u002Fgithub.com\u002FImprobable-AI\u002Frapid-locomotion-rl\u002Ftree\u002Fmain)\n\n- [CoRL2021] 使用大规模并行深度强化学习在几分钟内学会行走：[论文](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2109.11978.pdf)，[OpenReview](https:\u002F\u002Fopenreview.net\u002Fforum?id=wK2fDDJ5VcF)，[**代码**](https:\u002F\u002Fgithub.com\u002Fleggedrobotics\u002Flegged_gym)，[项目](https:\u002F\u002Fleggedrobotics.github.io\u002Flegged_gym\u002F)\n\n- [ICRA2021] 再探动力学随机化：以四足行走为例的研究：[项目](https:\u002F\u002Fwww.pair.toronto.edu\u002Funderstanding-dr\u002F)，[论文](https:\u002F\u002Farxiv.org\u002Fabs\u002F2011.02404)，[视频](https:\u002F\u002Fyoutu.be\u002FckdHWWpfSpk)\n\n- [2021] GLiDE：基于质心模型，在不同环境中实现可推广的四足行走：[项目](https:\u002F\u002Fwww.pair.toronto.edu\u002Fglide-quadruped\u002F)，[论文](https:\u002F\u002Farxiv.org\u002Fabs\u002F2104.09771)\n\n- [CoRL2020] 学习接触自适应控制器，实现稳健高效的腿部行走：[论文](https:\u002F\u002Farxiv.org\u002Fabs\u002F2009.10019)，[视频](https:\u002F\u002Fyoutu.be\u002FJJOmFZKpYTo)，[项目](https:\u002F\u002Fsites.google.com\u002Fview\u002Flearn-contact-controller\u002Fhome)，[博客](https:\u002F\u002Fdeveloper.nvidia.com\u002Fblog\u002Fcontact-adaptive-controller-locomotion\u002F)\n\n- [RAL2021] 学习状态表示及在杂乱动态环境中的导航：[论文](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2103.04351.pdf)\n\n- **[HumanoidVerse](https:\u002F\u002Fgithub.com\u002FLeCAR-Lab\u002FHumanoidVerse):** 用于人形机器人学习的多模拟器框架（2024年）\n  - 支持多种模拟器（Isaac Gym、Flex、MuJoCo）\n  - 包含多样化的人形机器人模型和环境\n  - 提供全面的基准测试工具\n  - 能够在不同模拟器上高效地进行并行训练\n\n- **[HIMLoco](https:\u002F\u002Fjunfeng-long.github.io\u002FHIMLoco\u002F):** 用于稳健人形机器人行走的层次模仿学习（2024年）\n  - [论文](https:\u002F\u002Farxiv.org\u002Fabs\u002F2403.14864)\n  - 具有层次模仿学习框架\n  - 能够在复杂环境中实现稳健的人形机器人行走\n  - 已成功应用于实际场景\n\n- **[ASAP](https:\u002F\u002Fagile.human2humanoid.com\u002F):** 对齐仿真与真实物理，用于学习敏捷的人形全身技能（2025年）\n  - [论文](https:\u002F\u002Farxiv.org\u002Fabs\u002F2502.01143)\n  - 采用两阶段框架实现仿真到现实的迁移\n  - 能够实现高度敏捷的人形动作，如踢腿和跳跃\n  - 已成功部署在 Unitree G1 真实人形机器人上\n  - 相较于 SysID 和 DR 基线有显著提升\n\n### 博客\n\n- **[NVIDIA Omniverse 简介](https:\u002F\u002Fzhuanlan.zhihu.com\u002Fp\u002F462305733)**\n\n---\n\n## 📑 研究论文\n\n### 核心论文\n- [Isaac Gym：高性能 GPU 加速物理仿真](https:\u002F\u002Farxiv.org\u002Fabs\u002F2108.10470)（NeurIPS 2021）\n  - [项目页面](https:\u002F\u002Fsites.google.com\u002Fview\u002Fisaacgym-nvidia)\n  - [OpenReview 讨论](https:\u002F\u002Fopenreview.net\u002Fforum?id=fgFBtYgJQX_)\n\n### 机器人操控\n\n- **[RLAfford](https:\u002F\u002Fgithub.com\u002Fhyperplane-lab\u002FRLAfford)：** “RLAfford：基于强化学习的端到端效用学习”官方实现，ICRA 2023。\n- **[面向机器人的掩码视觉预训练（MVP）](https:\u002F\u002Fgithub.com\u002Fir413\u002Fmvp)：** MVP项目的代码库。\n- **[Factory：用于机器人装配的快速接触建模](https:\u002F\u002Fsites.google.com\u002Fnvidia.com\u002Ffactory)：** RSS 2022。\n  - [论文](http:\u002F\u002Fdoi.acm.org\u002F10.1145\u002F3450626.3459670)\n  - [代码](https:\u002F\u002Fgithub.com\u002FNVIDIA-Omniverse\u002FIsaacGymEnvs)\n- **[ASE：对抗性技能嵌入](https:\u002F\u002Fnv-tlabs.github.io\u002FASE\u002F)：** SIGGRAPH 2022。\n  - [论文](https:\u002F\u002Farxiv.org\u002Fabs\u002F2205.01906)\n  - [代码](https:\u002F\u002Fgithub.com\u002FNVIDIA-Omniverse\u002FIsaacGymEnvs)\n- **[数据驱动的操作空间控制（OSCAR）](https:\u002F\u002Fcremebrule.github.io\u002Foscar-web\u002F)：** 自适应且鲁棒的机器人操作。\n  - [论文](https:\u002F\u002Farxiv.org\u002Fabs\u002F2110.00704)\n  - [代码](https:\u002F\u002Fgithub.com\u002Fnvlabs\u002Foscar)\n- **[DefGraspSim：基于仿真的人工抓取可变形物体](https:\u002F\u002Fsites.google.com\u002Fnvidia.com\u002Fdefgraspsim)：** 基于仿真的可变形物体抓取。\n  - [论文](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2107.05778.pdf)\n  - [视频](https:\u002F\u002Fyoutu.be\u002FCaj0AtsKKVI)\n  - [代码](https:\u002F\u002Fgithub.com\u002FNVlabs\u002Fdeformable_object_grasping)\n- **[手持物体位姿跟踪](https:\u002F\u002Fsites.google.com\u002Fview\u002Fin-hand-object-pose-tracking\u002F)：** ICRA 2021。\n  - [论文](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2002.12160.pdf)\n- **[STORM：用于反应式操作的快速关节空间模型预测控制](https:\u002F\u002Fsites.google.com\u002Fview\u002Fmanipulation-mpc)：** CoRL 2021。\n  - [论文](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2104.13542.pdf)\n  - [代码](https:\u002F\u002Fgithub.com\u002FNVlabs\u002Fstorm)\n- **[将灵巧操作从GPU仿真迁移到真实世界的TriFinger机器人](https:\u002F\u002Fs2r2-ig.github.io\u002F)：**\n  - [论文](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2108.09779.pdf)\n  - [代码](https:\u002F\u002Fgithub.com\u002Fpairlab\u002Fleibnizgym)\n- **[用于机器人操作策略的仿真因果推理](https:\u002F\u002Fsites.google.com\u002Fview\u002Fcrest-causal-struct-xfer-manip)：** ICRA 2021。\n  - [论文](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2103.16772.pdf)\n- **[反应式长 horizon 任务执行](https:\u002F\u002Fwww.youtube.com\u002Fplaylist?list=PL-oD0xHUngeLfQmpngYkGFZarstfPOXqX)：** IROS 2021。\n  - [论文](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2011.08694.pdf)\n- **[RoboDuet](https:\u002F\u002Flocomanip-duet.github.io\u002F)：** 学习全身足式运动与操作协作策略（2024年）\n  - [论文](https:\u002F\u002Farxiv.org\u002Fabs\u002F2403.17367)\n  - [代码](https:\u002F\u002Fgithub.com\u002FRobo-Duet\u002FRoboDuet)\n  - 采用运动与操作双策略框架\n  - 在移动操作任务中成功率提升23%\n  - 无需微调即可从Go1+ARX5迁移至Go2+ARX5\n\n- **[Kitchen Worlds](https:\u002F\u002Fgithub.com\u002FLearning-and-Intelligent-Systems\u002Fkitchen-worlds)：** 厨房场景中的长 horizon 任务与运动规划（TAMP）（2024年）\n  - [论文](https:\u002F\u002Farxiv.org\u002Fabs\u002F2410.02193)\n  - [代码](https:\u002F\u002Fgithub.com\u002FLearning-and-Intelligent-Systems\u002Fkitchen-worlds)\n  - 提供厨房及家居场景中的TAMP问题库\n  - 集成VLM（如GPT-4v）指导进行规划\n  - 包括程序化场景与轨迹生成功能\n\n### 定位与控制\n\n- **[HumanoidVerse](https:\u002F\u002Fgithub.com\u002FLeCAR-Lab\u002FHumanoidVerse)：** 用于类人机器人学习的多仿真框架（2024年）\n  - 支持多种仿真器（Isaac Gym、Flex、MuJoCo）\n  - 包含多样化的类人机器人模型和环境\n  - 提供全面的基准测试工具\n  - 实现跨仿真器的高效并行训练\n\n- **[HIMLoco](https:\u002F\u002Fjunfeng-long.github.io\u002FHIMLoco\u002F)：** 用于鲁棒类人机器人步态的分层模仿学习（2024年）\n  - [论文](https:\u002F\u002Farxiv.org\u002Fabs\u002F2403.14864)\n  - 采用分层模仿学习框架\n  - 实现复杂环境下类人机器人的稳健行走\n  - 已成功应用于实际场景\n\n- **[ASAP](https:\u002F\u002Fagile.human2humanoid.com\u002F)：** 对齐仿真与真实物理以学习敏捷的类人机器人全身技能（2025年）\n  - [论文](https:\u002F\u002Farxiv.org\u002Fabs\u002F2502.01143)\n  - 采用两阶段的仿真到现实迁移框架\n  - 可实现高敏捷性的类人机器人动作，如踢腿和跳跃\n  - 已成功部署在真实的Unitree G1类人机器人上\n  - 相较于SysID和DR基线有显著提升\n\n- **[利用大规模并行深度强化学习在几分钟内学会行走](https:\u002F\u002Fleggedrobotics.github.io\u002Flegged_gym\u002F)：** CoRL 2021。\n  - [论文](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2109.11978.pdf)\n  - [代码](https:\u002F\u002Fgithub.com\u002Fleggedrobotics\u002Flegged_gym)\n\n- **[动力学随机化再探](https:\u002F\u002Fwww.pair.toronto.edu\u002Funderstanding-dr\u002F)：** 四足机器人步态案例研究。\n  - [论文](https:\u002F\u002Farxiv.org\u002Fabs\u002F2011.02404)\n  - [视频](https:\u002F\u002Fyoutu.be\u002FckdHWWpfSpk)\n\n- **[GLiDE：通用四足机器人步态](https:\u002F\u002Fwww.pair.toronto.edu\u002Fglide-quadruped\u002F)：**\n  - [论文](https:\u002F\u002Farxiv.org\u002Fabs\u002F2104.09771)\n\n- **[学习接触自适应控制器](https:\u002F\u002Fsites.google.com\u002Fview\u002Flearn-contact-controller\u002Fhome)：** 用于鲁棒高效的足式行走。\n  - [论文](https:\u002F\u002Farxiv.org\u002Fabs\u002F2009.10019)\n  - [视频](https:\u002F\u002Fyoutu.be\u002FJJOmFZKpYTo)\n  - [博客](https:\u002F\u002Fdeveloper.nvidia.com\u002Fblog\u002Fcontact-adaptive-controller-locomotion\u002F)\n\n- **[学习状态表示与导航](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2103.04351.pdf)：** 适用于杂乱且动态的环境。\n\n### 其他\n\n- **[BayesSimIG](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2107.04527.pdf)：** 基于Isaac Gym的自适应领域随机化中的可扩展参数推断。\n  - [代码](https:\u002F\u002Fgithub.com\u002FNVlabs\u002Fbayes-sim-ig)\n- **[Isaac Gym：高性能GPU加速物理仿真](https:\u002F\u002Fsites.google.com\u002Fview\u002Fisaacgym-nvidia)：** NeurIPS 2021。\n  - [论文](https:\u002F\u002Farxiv.org\u002Fabs\u002F2108.10470)\n  - [OpenReview](https:\u002F\u002Fopenreview.net\u002Fforum?id=fgFBtYgJQX_)\n- **[学习游泳](https:\u002F\u002Farxiv.org\u002Fabs\u002F2410.00120v1)：** 针对推进器驱动AUV的六自由度强化学习控制。\n- **[MarineGym：基于高保真强化学习仿真的水下航行器加速训练](https:\u002F\u002Farxiv.org\u002Fabs\u002F2410.14117)：** 基于Issac Sim。\n- **[space_robotics_bench](https:\u002F\u002Fgithub.com\u002FAndrejOrsula\u002Fspace_robotics_bench)** 空间机器人基准测试平台。\n- **[Humanoid-Gym](https:\u002F\u002Fgithub.com\u002Froboterax\u002Fhumanoid-gym)：** 类人机器人强化学习，支持零样本仿真到现实迁移。\n\n\n## 🛠 工具与库\n\n### 强化学习框架\n\n- [RL Games](https:\u002F\u002Fgithub.com\u002FDenys88\u002Frl_games) - 兼容的强化学习算法\n- [ElegantRL](https:\u002F\u002Fgithub.com\u002FAI4Finance-Foundation\u002FElegantRL)\n- [skrl](https:\u002F\u002Fgithub.com\u002FToni-SM\u002Fskrl) - 模块化强化学习库\n- [Minimal Stable PPO](https:\u002F\u002Fgithub.com\u002FToruOwO\u002Fminimal-stable-PPO)\n\n\n- [skrl](https:\u002F\u002Fgithub.com\u002FToni-SM\u002Fskrl)，[论文](https:\u002F\u002Farxiv.org\u002Fabs\u002F2202.03825)\n  \n- [RSL RL](https:\u002F\u002Fgithub.com\u002Fleggedrobotics\u002Frsl_rl)\n\n## 相关 GitHub 仓库\n\n- [IsaacGymEnvs](https:\u002F\u002Fgithub.com\u002FNVIDIA-Omniverse\u002FIsaacGymEnvs)\n\n- [isaacgym_hammering](https:\u002F\u002Fgithub.com\u002FLiCHOTHU\u002Fisaacgym_hammering)\n\n- [isaacgym-utils](https:\u002F\u002Fgithub.com\u002Fiamlab-cmu\u002Fisaacgym-utils)：由 CMU 智能自主操作实验室开发\n\n- [isaacgym_sandbox](https:\u002F\u002Fgithub.com\u002Fkploeger\u002Fisaacgym_sandbox)\n\n- [thormang3-gogoro-PPO](https:\u002F\u002Fgithub.com\u002Fguichristmann\u002Fthormang3-gogoro-PPO)：使用 RL-PPO 和 NVIDIA Isaac Gym 对两轮车辆进行基于转向的控制\n\n- [dvrk_IssacGym](https:\u002F\u002Fgithub.com\u002Fbaotruyenthach\u002Fdvrk_IssacGym)，[链接](https:\u002F\u002Fgithub.com\u002Fbaotruyenthach\u002Fdvrk_grasp_pipeline_isaacgym)\n\n\n### 社区项目\n\n- **[IsaacGymEnvs](https:\u002F\u002Fgithub.com\u002FNVIDIA-Omniverse\u002FIsaacGymEnvs)：** 官方 Isaac Gym 强化学习环境。\n- **[isaacgym_hammering](https:\u002F\u002Fgithub.com\u002FLiCHOTHU\u002Fisaacgym_hammering)：** 锤击任务实现。\n- **[isaacgym-utils](https:\u002F\u002Fgithub.com\u002Fiamlab-cmu\u002Fisaacgym-utils)：** CMU 智能自主操作实验室提供的工具集。\n- **[isaacgym_sandbox](https:\u002F\u002Fgithub.com\u002Fkploeger\u002Fisaacgym_sandbox)：** Isaac Gym 实验沙盒。\n- **[thormang3-gogoro-PPO](https:\u002F\u002Fgithub.com\u002Fguichristmann\u002Fthormang3-gogoro-PPO)：** 使用 PPO 控制两轮车辆。\n- **[Bez_IsaacGym](https:\u002F\u002Fgithub.com\u002Futra-robosoccer\u002FBez_IsaacGym)：** 用于人形机器人 Bez 的环境。\n- **[DexterousHands](https:\u002F\u002Fgithub.com\u002FPKU-MARL\u002FDexterousHands)：** 双手灵巧操作任务。\n- **[legged_gym_isaac](https:\u002F\u002Fgithub.com\u002Fchengxuxin\u002Flegged_gym_isaac)：** 在 Isaac Gym 中模拟的足式机器人。\n- **[shifu](https:\u002F\u002Fgithub.com\u002F42jaylonw\u002Fshifu)：** 适用于任何机器人的环境构建工具。\n- **[Rofunc](https:\u002F\u002Fgithub.com\u002FSkylark0924\u002FRofunc)：** 用于机器人示教学习的 Python 包。\n- **[Dofbot Reacher](https:\u002F\u002Fgithub.com\u002Fj3soon\u002FOmniIsaacGymEnvs-DofbotReacher)：** Dofbot 的 Sim2Real 环境。\n- **[UR10 Reacher](https:\u002F\u002Fgithub.com\u002Fj3soon\u002FOmniIsaacGymEnvs-UR10Reacher)：** UR10 的 Sim2Real 环境。\n- **[TimeChamber](https:\u002F\u002Fgithub.com\u002Finspirai\u002FTimeChamber)：** 大规模并行自我博弈框架。\n- **[RL-MPC-Locomotion](https:\u002F\u002Fgithub.com\u002Fsilvery107\u002Frl-mpc-locomotion)：** 四足机器人运动的深度强化学习。\n- **[Isaac_Underwater](https:\u002F\u002Fgithub.com\u002Fleonlime\u002Fisaac_underwater)：** 使用 NVIDIA Isaac Sim 进行水上和水下测试。\n- **[VRKitchen2.0-IndoorKit](https:\u002F\u002Fgithub.com\u002Fyizhouzhao\u002FVRKitchen2.0-IndoorKit)：** Omniverse IndoorKit 扩展。\n- **[agibot_x1_train](https:\u002F\u002Fgithub.com\u002FAgibotTech\u002Fagibot_x1_train)：** AgiBot X1 的强化学习训练代码。\n- \n\n---\n\n## 会议环节与演讲\n\n- **[Isaac Gym 和 Omniverse：高性能强化学习的演进 [A31118]](https:\u002F\u002Fevents.rainfocus.com\u002Fwidget\u002Fnvidia\u002Fnvidiagtc\u002Fsessioncatalog?search=A31118)**\n- **[四足机器人挑战性任务的学习：从仿真到现实 [A31308]](https:\u002F\u002Fevents.rainfocus.com\u002Fwidget\u002Fnvidia\u002Fnvidiagtc\u002Fsessioncatalog?search=A31308)**\n- **[Isaac Sim 中的 Sim-to-Real](https:\u002F\u002Fwww.nvidia.com\u002Fen-us\u002Fon-demand\u002Fsession\u002Fgtcspring21-s31824\u002F)**\n- **[Isaac Gym：端到端 GPU 加速的强化学习](https:\u002F\u002Fwww.nvidia.com\u002Fen-us\u002Fon-demand\u002Fsession\u002Fgtcspring21-s32037\u002F)**\n- **[弥合 Sim2Real 差距：用于训练深度学习机器人感知模型的仿真调优](https:\u002F\u002Fwww.nvidia.com\u002Fen-us\u002Fon-demand\u002Fsession\u002Fgtcspring21-s31649\u002F)**\n- **[强化学习与内部物流](https:\u002F\u002Fwww.nvidia.com\u002Fen-us\u002Fon-demand\u002Fsession\u002Fgtcspring21-e31467\u002F)**\n- **[使用 NVIDIA Isaac SDK 构建机器人应用](https:\u002F\u002Fwww.nvidia.com\u002Fen-us\u002Fon-demand\u002Fsession\u002Fgtcfall20-a21856\u002F)**\n- **[NVIDIA Isaac Sim — 令人惊叹的机器人模型和任务](https:\u002F\u002Fwww.nvidia.com\u002Fen-us\u002Fon-demand\u002Fsession\u002Fgtcsj20-d2s43\u002F)**\n- **[Omniverse View 2021.2 — 应用程序巡览](https:\u002F\u002Fwww.nvidia.com\u002Fen-us\u002Fon-demand\u002Fsession\u002Fomniverse2020-om1315\u002F)**\n- **[ISAAC SIM 介绍与现场演示](https:\u002F\u002Fwww.nvidia.com\u002Fen-us\u002Fon-demand\u002Fsession\u002Fomniverse2020-om1314\u002F)**\n- **[NVIDIA 按需 ISAAC SIM 会话](https:\u002F\u002Fwww.nvidia.com\u002Fen-us\u002Fon-demand\u002Fsearch\u002F?facet.mimetype[]=event%20session&layout=list&page=1&q=isaac%20sim&sort=relevance)**\n\n---\n\n## 🤖 自动化研究更新\n\n- 数据来源：arXiv 上关于 \"isaac gym\"、\"omni isaac\" 和 \"isaac lab\" 的查询结果。\n- 更新时间：**每天 UTC 时间 09:00**，或通过 Actions \"Research Bot\" 手动触发。\n- 输出内容：位于本 README 中 'research-bot:start' 和 'research-bot:end' 标记之间的更新内容。\n- 安全机制：系统会打开一个标记为 \"needs-approval\" 的草稿 PR；不会自动合并任何更改。\n- 配置：编辑 `.research-bot.yaml` 文件以调整查询条件和限制。\n\n## 🌟 贡献说明\n\n欢迎各位贡献！请在提交拉取请求之前阅读我们的 [贡献指南](CONTRIBUTING.md)。\n\n## 📄 许可证\n\n本仓库采用 MIT 许可证授权——详情请参阅 [LICENSE](LICENSE) 文件。\n\n## 🙏 致谢\n\n特别感谢所有贡献者以及 NVIDIA Isaac 团队，正是他们让这些资源得以向机器人社区开放。\n\n## 🧠 最新研究（自动更新）\n\n\u003C!-- research-bot:start -->\n- [HumanoidVerse：一种用于视觉-语言引导的多物体重排任务的多功能人形机器人](https:\u002F\u002Farxiv.org\u002Fabs\u002F2508.16943) — Haozhuo Zhang, Jingkai Sun, Michele Caprio 等 (2025年8月23日) [pdf](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2508.16943.pdf)\n\u003C!-- research-bot:end -->","# NVIDIA Isaac Gym 快速上手指南\n\nNVIDIA Isaac Gym 是一个基于 GPU 的高性能物理仿真环境，专为机器人强化学习（RL）设计，支持大规模并行训练。本指南帮助中国开发者快速搭建环境并运行第一个示例。\n\n## 1. 环境准备\n\n在开始之前，请确保您的开发机器满足以下硬件和软件要求。\n\n### 系统要求\n*   **操作系统**: Ubuntu 18.04 或 20.04 (推荐 20.04)\n*   **GPU**: NVIDIA RTX 系列或更高版本 (显存建议 8GB 以上)，需支持 CUDA\n*   **CPU**: 多核处理器 (核心数越多，并行仿真效率越高)\n*   **内存**: 建议 32GB 或以上\n\n### 前置依赖\n请安装以下基础依赖库：\n\n```bash\nsudo apt-get update\nsudo apt-get install -y libomp5 libgl1-mesa-glx libegl1-mesa libxrandr2 libxss1 libxcursor1 libxcomposite1 libasound2 libxi6 libxtst6\n```\n\n**注意**：\n*   确保已安装与您的 GPU 匹配的 **NVIDIA Driver**。\n*   推荐安装 **CUDA Toolkit 11.3** 或更高版本（Isaac Gym Preview 4 通常兼容 CUDA 11.3+）。\n*   建议创建独立的 Python 虚拟环境（如使用 `conda` 或 `venv`），Python 版本推荐 **3.7** 或 **3.8**。\n\n## 2. 安装步骤\n\n由于 NVIDIA 官方下载需要注册账号，请按以下步骤操作：\n\n### 第一步：下载安装包\n1. 访问 [NVIDIA Isaac Gym 下载页面](https:\u002F\u002Fdeveloper.nvidia.com\u002Fisaac-gym)。\n2. 登录 NVIDIA 开发者账号。\n3. 下载 **Isaac Gym Preview 4** (当前最稳定版本) 的压缩包 (`.tar.gz`)。\n   * *提示：国内用户若下载缓慢，可尝试使用网络加速工具或寻找社区维护的镜像资源。*\n\n### 第二步：解压与安装\n将下载的压缩包移动到目标目录并解压：\n\n```bash\n# 替换为您的实际下载路径\nmv isaacgym_preview_4.tar.gz ~\u002Ftools\u002F\ncd ~\u002Ftools\u002F\n\n# 解压文件\ntar -xzf isaacgym_preview_4.tar.gz\n```\n\n### 第三步：配置 Python 环境\n进入解压后的 `python` 目录并安装依赖：\n\n```bash\ncd isaacgym\u002Fpython\n\n# 创建并激活 conda 环境 (可选但推荐)\nconda create -n isaac python=3.8\nconda activate isaac\n\n# 安装 PyTorch (需匹配您的 CUDA 版本，以下为 CUDA 11.3 示例)\npip install torch==1.10.0+cu113 torchvision==0.11.1+cu113 torchaudio==0.10.0 -f https:\u002F\u002Fdownload.pytorch.org\u002Fwhl\u002Fcu113\u002Ftorch_stable.html\n\n# 安装 Isaac Gym 核心包\npip install -e .\n```\n\n## 3. 基本使用\n\n安装完成后，您可以运行官方提供的示例来验证环境。Isaac Gym 的核心优势在于能够同时在 GPU 上运行数千个仿真环境。\n\n### 运行第一个示例：机械臂抓取\n以下命令将启动一个包含 1024 个并行环境的 Ant 机器人行走训练示例（这是最经典的基准测试）：\n\n```bash\ncd ~\u002Ftools\u002Fisaacgym\u002Fpython\u002Fexamples\n\n# 运行 Ant 机器人强化学习示例\n# --num_envs 设置并行环境数量，根据显存大小调整 (RTX 3090\u002F4090 可尝试 4096+)\npython 1080_balls_of_solitude.py --num_envs 1024 --headless\n```\n\n*   `--headless`: 无头模式运行，不显示图形界面，适合服务器训练。\n*   若要查看图形化仿真过程，请去掉 `--headless` 参数（需本地连接显示器或通过 X11 转发）。\n\n### 代码结构简述\n在自定义项目中，您通常需要继承 `VecTask` 类。一个简单的训练循环逻辑如下：\n\n```python\nfrom isaacgym import gymapi, gymtorch\nimport torch\n\n# 初始化 Gym\ngym = gymapi.acquire_gym()\n\n# 创建仿真环境、加载资产、设置传感器等...\n# (具体参考 examples 目录下的源码)\n\n# 训练循环示例\nfor iteration in range(num_iterations):\n    # 1. 计算动作 (Policy Forward)\n    actions = policy.compute(observations)\n    \n    # 2. 执行动作并步进仿真 (Step Simulation)\n    gym.step_simulation(sim_ptr)\n    \n    # 3. 获取新的观测值和奖励\n    observations, rewards, dones = get_sim_state()\n    \n    # 4. 更新强化学习模型 (PPO\u002FSAC 等)\n    agent.update(observations, actions, rewards, dones)\n```\n\n### 下一步建议\n*   **文档查阅**: 详细 API 请参考 `docs` 文件夹或 [在线文档](https:\u002F\u002Fdocs.nvidia.com\u002Fisaac\u002Fisaac\u002Fdoc\u002Findex.html)。\n*   **进阶框架**: 对于更复杂的项目，推荐结合使用 [Isaac Lab](https:\u002F\u002Fisaac-sim.github.io\u002FIsaacLab\u002F) 或 [OmniIsaacGymEnvs](https:\u002F\u002Fgithub.com\u002FNVIDIA-Omniverse\u002FOmniIsaacGymEnvs)，它们提供了更模块化的 PPO 实现和环境库。\n*   **社区资源**: 关注 [NVIDIA 开发者论坛](https:\u002F\u002Fforums.developer.nvidia.com\u002Fc\u002Fagx-autonomous-machines\u002Fisaac\u002Fisaac-gym\u002F322) 获取最新补丁和技术讨论。","某机器人实验室团队正致力于训练一个人形机器人在复杂地形中实现稳定行走，需要快速复现前沿算法并搭建高效的强化学习仿真环境。\n\n### 没有 awesome-isaac-gym 时\n- **资源分散难寻**：团队成员需在 GitHub、arXiv 和各类论坛中大海捞针，花费数周时间才凑齐零散的 Isaac Gym 教程、核心论文和代码库。\n- **环境配置踩坑**：缺乏统一的安装指引和版本对照表，成员在配置 GPU 加速环境和 PhysX 依赖时频繁报错，导致项目启动严重滞后。\n- **复现成本高昂**：找不到经过验证的基准项目（如 HumanoidVerse 或 Isaac Lab），从零编写仿真逻辑不仅耗时，还容易因物理参数设置不当导致训练失败。\n- **技术迭代脱节**：难以及时获取最新的社区工具和研究进展，团队使用的框架版本陈旧，无法利用最新的并行训练特性提升效率。\n\n### 使用 awesome-isaac-gym 后\n- **一站式资源聚合**：直接通过分类清晰的清单获取官方文档、精选论文及开源代码，将前期调研时间从数周缩短至几天。\n- **平滑上手体验**：依据\"Getting Started\"中的权威指南和常见问题解答，迅速完成环境部署，避免了兼容性陷阱，让团队当天即可运行首个 Demo。\n- **高效基准复用**：直接集成列表中推荐的模块化框架（如 Isaac Lab）和成熟案例，基于现有高质量代码进行二次开发，大幅降低算法验证门槛。\n- **紧跟前沿动态**：实时追踪\"Latest Releases\"板块中更新的工具链（如 PhysX 5 SDK），确保仿真环境始终处于性能最优状态，加速模型收敛。\n\nawesome-isaac-gym 通过将碎片化的生态资源系统化，让人形机器人研发团队从繁琐的环境搭建中解放出来，专注于核心算法的创新与落地。","https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Frobotlearning123_awesome-isaac-gym_0a186b38.png","robotlearning123","Cong","https:\u002F\u002Foss.gittoolsai.com\u002Favatars\u002Frobotlearning123_cc02546e.jpg","Building AGI","Harvard University","Cambridge",null,"https:\u002F\u002Fgithub.com\u002Frobotlearning123",[86],{"name":87,"color":88,"percentage":89},"Python","#3572A5",100,1171,77,"2026-04-14T19:27:05",4,"Linux","必需 NVIDIA GPU (支持 CUDA)，具体型号和显存未说明 (基于 PhysX 5 和 GPU 加速特性，通常推荐 RTX 30\u002F40 系列或更高)","未说明",{"notes":98,"python":96,"dependencies":99},"该仓库是资源列表而非单一软件包。核心工具 NVIDIA Isaac Gym 主要支持 Linux 系统，需通过 NVIDIA 官网单独下载安装。运行环境强依赖 NVIDIA GPU 进行物理仿真加速，并通常需要配合 NVIDIA Omniverse 使用。具体版本要求（如 Python、CUDA）需参考官方安装指南链接。",[100,101,102,103,104],"NVIDIA Isaac Gym","NVIDIA Omniverse","PhysX 5 SDK","PyTorch (隐含于 RL 框架)","Python (隐含)",[18],[107,108,109,110,111],"isaac-gym","robot-learning","robotics","reinforcement-learning","openai-gym","2026-03-27T02:49:30.150509","2026-04-15T13:17:39.868128",[115,120],{"id":116,"question_zh":117,"answer_zh":118,"source_url":119},34286,"Isaac Gym 支持水下机器人仿真吗？","目前尚未完全支持水下仿真，但简单的流体环境是可以运行的。维护者已更新了相关资源链接供参考：\n1. https:\u002F\u002Fgithub.com\u002Fwangcongrobot\u002Fawesome-isaac-gym\u002Fcommit\u002F9dd4ab654a5cb826882add548d8ff053f6e48b8b\n2. https:\u002F\u002Fgithub.com\u002Fwangcongrobot\u002Fawesome-isaac-gym\u002Fcommit\u002Fbba1bb15d593820e6d700b106ab3479486e8faf3\n关于 GPU 并行化的 Mujoco (MJX) 是否能处理流体环境，目前尚无明确结论，建议关注官方动态。","https:\u002F\u002Fgithub.com\u002Frobotlearning123\u002Fawesome-isaac-gym\u002Fissues\u002F3",{"id":121,"question_zh":122,"answer_zh":123,"source_url":124},34287,"在 Isaac Gym version 4 中遇到 'create_force_sensor' 属性错误怎么办？","该问题属于特定版本的 API 兼容性或 Bug 问题，建议直接前往 NVIDIA 官方论坛发帖求助，以获取更专业的技术支持和解决方案。","https:\u002F\u002Fgithub.com\u002Frobotlearning123\u002Fawesome-isaac-gym\u002Fissues\u002F2",[]]