1. [主页](/) [行业](https://developer.nvidia.cn/industries/overview/) Isaac # NVIDIA Isaac 准备开始你的 AI 机器人开发了吗？NVIDIA Isaac™ 是理想的起点。这一开放式机器人开发平台由[仿真](https://www.nvidia.cn/use-cases/robotics-simulation/)和[机器人学习](https://www.nvidia.cn/use-cases/robot-learning/)框架、基于 NVIDIA® CUDA® 加速的库、AI 模型以及参考工作流组成，可用于创建自主移动机器人 (AMR)、机械臂、操作机器人以及[类人机器人](https://www.nvidia.cn/use-cases/humanoid-robots/)。 * * * ## NVIDIA Isaac 库和 AI 模型 NVIDIA 机器人开发全栈 CUDA 加速库和优化的 AI 模型让您能够以更好、更高效的方式开发、训练、仿真、部署、运营和优化机器人系统。 ## 用于操作的 NVIDIA Isaac ![借助 NVIDIA cuMotion 进行运动规划](https://developer.download.nvidia.com/images/isaac/cuMotion.gif) ### 运动规划 NVIDIA cuMotion 是一个 NVIDIA CUDA 加速库，通过同时运行多个轨迹优化来返回最佳解决方案，帮助大规模解决机器人运动规划问题。 [阅读指南](https://nvidia-isaac.github.io/cumotion/index.html) [从 GitHub 下载库](https://github.com/nvidia-isaac/cumotion/releases) ![姿态估计和追踪](https://developer.download.nvidia.com/images/isaac/FoundationPose.gif) ### 姿态估计和追踪 FoundationPose 是一个基础模型，用于对新物体进行 6D 姿态估计和追踪。它可以跟踪和估计不可见物体的姿态，并且可以处理具有挑战性的物体属性 (无纹理、光滑、微小) 以及快速运动或严重遮挡的场景。 [阅读指南](https://nvidia-isaac-ros.github.io/repositories_and_packages/isaac_ros_pose_estimation/index.html) 从 NGC 下载模型 ![使用 Foudnations 进行深度估计 Stereo](https://developer.download.nvidia.com/images/isaac/isaac-manipulator-features-foundationStereo.gif) ### 深度估计 FoundationStereo 是一个基础模型，旨在实现立体匹配的强大零样本泛化。 阅读指南 [从 NGC 下载模型](https://catalog.ngc.nvidia.com/orgs/nvidia/teams/tao/models/foundationstereo) ![使用 SyntheticDETR 进行物体检测](https://developer.download.nvidia.com/images/isaac/SyntheticaDETR.gif) ### 物体检测 SyntheticaDETR 是用于室内环境物体检测的预训练模型。它可用作 FoundationPose 等姿态估计器的前端，因此可以在姿态估计之前使用 2D 边界框定位物体。 [阅读指南](https://nvidia-isaac-ros.github.io/repositories_and_packages/isaac_ros_object_detection/index.html) [从 NGC 下载模型](https://catalog.ngc.nvidia.com/orgs/nvidia/teams/isaac/models/synthetica_detr) ### Isaac TeleOp 通过现实世界和模拟中的远程操作收集高质量的人类演示。 阅读指南 ## NVIDIA Isaac 助力出行 ![NVIDIA OSMO 云原生工作流编排平台](https://developer.download.nvidia.com/images/isaac/real-time-3-d-occupancy-grid.jpg) ### 实时 3D 占用率网格 借助 NVIDIA nvblox CUDA 加速的 3D 重建库，机器人能够识别 5 米以外 3D 空间中的障碍物，并生成 2D 成本图。获得结果的速度比以 CPU 为中心的方法快 100 倍。 [阅读指南](https://nvidia-isaac-ros.github.io/concepts/scene_reconstruction/nvblox/index.html) ![NVIDIA Isaac Manipulator 或支持 AI 的机械臂的实际应用](https://developer.download.nvidia.com/images/isaac/accelerated-stereo-visual-odometry.jpg) ### 加速立体视觉测距和 SLAM 加速立体视觉测距和 SLAM 使用 NVIDIA cuVSLAM，在各种传感器和平台上获得低于 1% 的实时 CUDA 加速视觉 SLAM 轨迹错误。 通过融合来自多个视点的输入，无缝浏览具有稀疏视觉特征或重复模式的环境。开始使用[pycuVSLAM](https://github.com/NVlabs/PyCuVSLAM)。 [阅读指南](https://nvidia-isaac-ros.github.io/concepts/visual_slam/cuvslam/index.html) ![通用的端到端移动性](https://developer.download.nvidia.com/images/isaac/generalizable-end-to-end-mobility-ari.jpg) ### 通用的端到端移动性 使用 NVIDIA COMPASS 训练基于视觉的移动性基础模型，实现跨机器人类型和不断变化的环境的导航。 工作流程包括[合成数据生成](https://www.nvidia.cn/glossary/synthetic-data-generation/)通过 NVIDIA Isaac Sim™ 和 Cosmos™ Transfer，在 Isaac Lab 中进行模型训练和后训练，并使用[NVIDIA Jetson Orin™](https://www.nvidia.cn/autonomous-machines/embedded-systems/jetson-orin/)或[Thor](https://www.nvidia.cn/autonomous-machines/embedded-systems/jetson-thor/)。 [阅读指南](https://nvlabs.github.io/COMPASS/) ![打开 Robotics ROS 徽标](https://developer.download.nvidia.com/images/isaac/logo-ros.jpg) #### NVIDIA Isaac ROS NVIDIA Isaac ROS (机器人操作系统) 基于开源 ROS 2 构建。这些 NVIDIA CUDA 加速计算包和 AI 模型的集合简化并加速了先进 AI 机器人应用的开发。 [详细了解 Isaac ROS](https://developer.nvidia.cn/isaac/ros) * * * ## 仿真和机器人学习 在基于物理的虚拟环境中设计、仿真、测试和训练基于 AI 的机器人和自主机器。 ![基于物理性质的虚拟环境中的 NVIDIA Isaac Sim 应用程序](https://developer.download.nvidia.com/images/isaac/nvidia-isaac-sim-1920x1080.jpg) ### NVIDIA Isaac Sim NVIDIA Isaac Sim，基于[NVIDIA Omniverse](https://developer.nvidia.cn/omniverse)因此，您可以在基于物理性质的虚拟环境中更快地开发自主机器。 一起，[NVIDIA Cosmos](https://developer.nvidia.cn/cosmos)和 Isaac Sim 可让您从 3D 场景生成合成数据，用于训练感知机器人。 [详细了解 Isaac Sim](https://developer.nvidia.cn/isaac/sim) ![用于机器人学习和基础模型训练的 NVIDIA Isaac Lab 应用](https://developer.download.nvidia.com/images/isaac/nvidia-isaac-lab-1920x1080.jpg) ### NVIDIA Isaac Lab 此轻量级示例应用基于 Isaac Sim 构建，并针对机器人学习和机器人基础模型训练进行了优化。 [详细了解 Isaac Lab](https://developer.nvidia.cn/isaac/lab) [使用 Isaac Lab-Arena 简化通用机器人策略评估](https://developer.nvidia.cn/blog/simplify-generalist-robot-policy-evaluation-in-simulation-with-nvidia-isaac-lab-arena/) #### Newton，新一代开源物理模拟引擎 Newton 是一个开源、GPU 加速和可扩展的物理引擎，由 Google DeepMind 和 Disney Research 共同开发，以及[由 Linux 基金会管理](https://www.linuxfoundation.org/press/linux-foundation-announces-contribution-of-newton-by-disney-research-google-deepmind-and-nvidia-to-accelerate-open-robot-learning)。Newton 基于 NVIDIA Warp 和 OpenUSD 构建，针对机器人技术进行了优化，并与 MuJoCo Playground 或 Isaac Lab 等学习框架兼容。[牛顿测试版](https://github.com/newton-physics)现在可以使用。 [开始使用 Newton](https://developer.nvidia.cn/newton-physics) ![Newton 开源仿真引擎针对机器人技术进行了优化](https://developer.download.nvidia.com/images/isaac/newton-ari.jpg) ## 用于人形机器人开发的 NVIDIA Isaac GR00T ![用于人形机器人开发的 NVIDIA Project GR00T](https://developer.download.nvidia.com/images/isaac/Isaac-gr00t-for-humanoid-robot-development.jpg) Isaac GR00T 是一项研究计划，也是一个开放式开发平台，适用于通用机器人基础模型 (例如 Isaac GR00T N) 和数据流水线 (例如 Isaac GR00T-Mimic 和 Isaac GR00T-Dreams) ，旨在加速人形机器人开发。 [详细了解 Isaac GR00T](https://developer.nvidia.cn/isaac/gr00t) * * * ## NVIDIA 加速系统 NVIDIA 的三个计算平台可简化并加速开发者的工作流程：用于构建机器人 AI 模型的 NVIDIA DGX™ 系统、用于对其进行仿真、测试和训练的 NVIDIA OVX™，以及用于部署和运行这些模型的 NVIDIA AGX™。 ![NVIDIA Jetson 平台](https://developer.download.nvidia.com/images/isaac/nvidia-dgx-1920x1080.jpg) ### NVIDIA DGX DGX 平台结合了出色的 NVIDIA 软件和基础架构，是训练机器人多模态基础模型的理想选择。 [详细了解 NVIDIA DGX](https://www.nvidia.cn/data-center/dgx-platform/) ![NVIDIA OVX 平台](https://developer.download.nvidia.com/images/isaac/nvidia-ovx-1920x1080.jpg) ### NVIDIA OVX OVX 系统提供行业领先的图形和计算性能，可加速新一代机器人开发。 [详细了解 NVIDIA OVX](https://www.nvidia.cn/data-center/products/ovx/) ![NVIDIA AGX 系统](https://developer.download.nvidia.com/images/isaac/nvidia-jetson-1920x1080.jpg) ### NVIDIA AGX 包括 NVIDIA Jetson™ 在内的 AGX 系统可提供卓越的性能和能效，这使其成为领先的机器人开发平台。将经过训练、测试和优化的机器人 AI 模型部署到这些系统中，以实现实际操作。 [详细了解 NVIDIA AGX](https://www.nvidia.cn/autonomous-machines/embedded-systems/) ![NVIDIA OSMO 云原生工作流编排平台](https://developer.download.nvidia.com/images/isaac/overview/nvidia-osmo.jpg) ### NVIDIA OSMO OSMO 是一个云原生工作流编排平台，可让您轻松地跨分布式环境 (从本地部署到私有云和公有云资源集群) 扩展工作负载。 [详细了解 NVIDIA OSMO](https://developer.nvidia.com/osmo) ## Isaac 学习资源 | title | featured | x_formats | document_url | technologies | document_date | short_summary | document_title | learning_level | x_content_types | | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | | Train Quadruped Locomotion with Isaac Lab & Newton | false | blog | https://developer.nvidia.com/blog/train-a-quadruped-locomotion-policy-and-simulate-cloth-manipulation-with-nvidia-isaac-lab-and-newton/ | Isaac Lab, Newton Physics Engine | 2025-09-29T04:00:00.000Z | Train quadruped locomotion policies and simulate cloth manipulation with Isaac Lab. | Train Quadruped Locomotion with Isaac Lab & Newton | Technical - Intermediate | Tutorial | | Cosmos World Foundation Models for Physical AI | false | blog | https://developer.nvidia.com/blog/advancing-physical-ai-with-nvidia-cosmos-world-foundation-model-platform/ | Cosmos, NVIDIA Isaac GROOT | 2025-01-09T05:00:00.000Z | Explore how Cosmos world foundation models accelerate physical AI development. | Cosmos World Foundation Models for Physical AI | Technical - Beginner | Explainer | | Beginner’s Guide: Robot Sim with ROS 2 & Isaac Sim | false | blog | https://developer.nvidia.com/blog/a-beginners-guide-to-simulating-and-testing-robots-with-ros-2-and-nvidia-isaac-sim/ | Isaac ROS, Isaac Sim | 2024-10-21T00:00:00.000Z | Build and test robot simulations using ROS 2 and Isaac Sim. | Beginner’s Guide: Robot Sim with ROS 2 & Isaac Sim | Technical - Beginner | Tutorial | | Synthetic Motion Generation for Humanoid Robots | false | blog | https://developer.nvidia.com/blog/building-a-synthetic-motion-generation-pipeline-for-humanoid-robot-learning/ | NVIDIA Isaac GROOT | 2025-03-18T04:00:00.000Z | Explore synthetic motion generation pipelines for training humanoid robots. | Synthetic Motion Generation for Humanoid Robots | Technical - Intermediate | Explainer | | Train AMRs with Synthetic Data for Detection | false | blog | https://developer.nvidia.com/blog/how-to-train-autonomous-mobile-robots-to-detect-warehouse-pallet-jacks-using-synthetic-data/ | Isaac Sim | 2023-10-29T04:00:00.000Z | Build an AMR perception model using synthetic data from Isaac Sim. | Train AMRs with Synthetic Data for Detection | Technical - Intermediate | Tutorial | | Amazon Zero-Touch Manufacturing with Isaac & Digital Twins | false | blog | https://blogs.nvidia.com/blog/amazon-zero-touch-manufacturing/ | cuMotion, Isaac ROS, Isaac Sim | 2025-08-11T04:00:00.000Z | See how Amazon uses Isaac Sim and digital twins for autonomous manufacturing. | Amazon Zero-Touch Manufacturing with Isaac & Digital Twins | Technical - Beginner | Overview | | Latest Isaac Platform Release Updates | false | blog | https://developer.nvidia.com/blog/advancing-robot-learning-perception-and-manipulation-with-latest-nvidia-isaac-release/ | Isaac ROS | 2025-01-06T05:00:00.000Z | Learn about the latest Isaac platform updates for robot learning and manipulation. | Latest Isaac Platform Release Updates | Technical - Beginner | News | | Isaac ROS Webinar Series | false | video | https://gateway.on24.com/wcc/experience/elitenvidiabrill/1407606/3998202/isaac-ros-webinar-series | Isaac ROS | 2026-01-01T05:00:00.000Z | Watch the Isaac ROS webinar series covering robotics development workflows. | Isaac ROS Webinar Series | Technical - Beginner | Tutorial | | GR00T N1: Foundation Model for Humanoid Robots | false | webpage | https://arxiv.org/abs/2503.14734 | NVIDIA Isaac GROOT | 2025-03-18T04:00:00.000Z | Explore the GR00T N1 vision-language-action model for humanoid robot control. | GR00T N1: Foundation Model for Humanoid Robots | Technical - Advanced | Explainer | | ReMEmbR: Spatio-Temporal Memory for Robot Navigation | false | webpage | https://arxiv.org/abs/2409.13682 | Isaac ROS | 2024-09-20T04:00:00.000Z | Explore spatio-temporal memory systems for long-horizon robot navigation. | ReMEmbR: Spatio-Temporal Memory for Robot Navigation | Technical - Advanced | Explainer | | Isaac ROS Release Notes | false | webpage | https://nvidia-isaac-ros.github.io/releases/index.html | Isaac ROS | 2026-02-19T05:00:00.000Z | Reference Isaac ROS release history, changelogs, and version details. | Isaac ROS Release Notes | Technical - Beginner | Documentation | | Isaac Lab: GPU Simulation for Robot Learning | false | webpage | https://research.nvidia.com/publication/2025-09_isaac-lab-gpu-accelerated-simulation-framework-multi-modal-robot-learning | Isaac Lab | 2025-09-29T04:00:00.000Z | Explore Isaac Lab’s GPU-accelerated simulation architecture for robot learning. | Isaac Lab: GPU Simulation for Robot Learning | Technical - Advanced | Explainer | | Universal Robots: AI Cobots with Isaac & Jetson | false | webpage | https://www.nvidia.com/en-us/customer-stories/universal-robots-accelerates-cobot-development-with-nvidia/ | Isaac ROS | 2026-03-04T05:00:00.000Z | See how Universal Robots uses Isaac and Jetson for AI-powered cobots. | Universal Robots: AI Cobots with Isaac & Jetson | Technical - Beginner | Overview | | End-to-End Robot AI with Isaac Sim & TAO | false | blog | https://developer.nvidia.com/blog/developing-and-deploying-ai-powered-robots-with-nvidia-isaac-sim-and-nvidia-tao/ | Isaac ROS, Isaac Sim, TAO Toolkit | 2022-05-05T04:00:00.000Z | Develop an end-to-end robot AI workflow with Isaac Sim and TAO. | End-to-End Robot AI with Isaac Sim & TAO | Technical - Intermediate | Tutorial | | Whole-Body Control & Teleoperation in Isaac Lab 2.3 | false | blog | https://developer.nvidia.com/blog/streamline-robot-learning-with-whole-body-control-and-enhanced-teleoperation-in-nvidia-isaac-lab-2-3/ | Isaac Lab | 2025-09-29T04:00:00.000Z | Explore whole-body control and teleoperation features in Isaac Lab 2.3. | Whole-Body Control & Teleoperation in Isaac Lab 2.3 | Technical - Intermediate | Explainer | [下载原始结果数据 (JSON)](https://developer.nvidia.cn/search-data/robotics.json) ## 开发者资源和支持 ![代表论坛的装饰性图像](https://developer.download.nvidia.com/images/omniverse/m48-people-group.svg) ### 探索社区 ![](https://developer.download.nvidia.com/images/isaac/m48-certification-ribbon-2-256px-blk.png) ### 获取培训和认证 ![](https://developer.download.nvidia.com/images/isaac/m48-ai-startup-256px-blk.png) ### 加入面向初创公司的计划 * * * ## 最新机器人开发新闻 [查看所有新闻 ](https://developer.nvidia.cn/blog/search-posts/?categories=Robotics)