[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"project-11683":3},{"id":4,"name":5,"fullName":6,"owner":7,"repo":5,"description":8,"homepage":9,"htmlUrl":10,"language":11,"languages":9,"totalLinesOfCode":9,"stars":12,"forks":13,"watchers":14,"openIssues":15,"contributorsCount":9,"subscribersCount":16,"size":16,"stars1d":17,"stars7d":18,"stars30d":19,"stars90d":16,"forks30d":16,"starsTrendScore":18,"compositeScore":20,"rankGlobal":9,"rankLanguage":9,"license":9,"archived":21,"fork":21,"defaultBranch":22,"hasWiki":21,"hasPages":21,"topics":9,"createdAt":9,"pushedAt":9,"updatedAt":23,"readmeContent":24,"aiSummary":25,"trendingCount":16,"starSnapshotCount":16,"syncStatus":26,"lastSyncTime":27,"discoverSource":28},11683,"GR00T-WholeBodyControl","NVlabs\u002FGR00T-WholeBodyControl","NVlabs","Welcome to GR00T Whole-Body Control (WBC)! This is a unified platform for developing and deploying advanced humanoid controllers. This includes: Decoupled WBC models used in NVIDIA Isaac-Gr00t, Gr00t N1.5 and N1.6 and GEAR-SONIC ",null,"https:\u002F\u002Fgithub.com\u002FNVlabs\u002FGR00T-WholeBodyControl","Python",2261,317,25,36,0,31,93,307,29.51,false,"main","2026-06-12 02:02:33","\u003Cdiv align=\"center\">\n\n  \u003Cimg src=\"media\u002Fgroot_wbc.png\" width=\"800\" alt=\"GEAR SONIC Header\">\n\n  \u003C!-- --- -->\n  \n  \n\u003C\u002Fdiv>\n\n\u003Cdiv align=\"center\">\n\n[![License](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FLicense-Apache%202.0-76B900.svg)](LICENSE)\n[![IsaacLab](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FIsaacLab-2.3.2-orange.svg)](https:\u002F\u002Fgithub.com\u002Fisaac-sim\u002FIsaacLab\u002Freleases\u002Ftag\u002Fv2.3.2)\n[![Documentation](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fdocs-GitHub%20Pages-76B900.svg)](https:\u002F\u002Fnvlabs.github.io\u002FGR00T-WholeBodyControl\u002F)\n[![Demo](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FLive%20Demo-GEAR--SONIC-blue.svg)](https:\u002F\u002Fnvlabs.github.io\u002FGEAR-SONIC\u002Fdemo.html)\n\n\u003C\u002Fdiv>\n\n---\n\n\n\n\n# GR00T-WholeBodyControl\n\nThis is the codebase for the **GR00T Whole-Body Control (WBC)** projects. It hosts model checkpoints and scripts for training, evaluating, and deploying advanced whole-body controllers for humanoid robots. We currently support:\n\n- **Decoupled WBC**: the decoupled controller (RL for lower body, and IK for upper body) used in NVIDIA GR00T [N1.5](https:\u002F\u002Fresearch.nvidia.com\u002Flabs\u002Fgear\u002Fgr00t-n1_5\u002F) and [N1.6](https:\u002F\u002Fresearch.nvidia.com\u002Flabs\u002Fgear\u002Fgr00t-n1_6\u002F) models;\n- **GEAR-SONIC Series**: our latest iteration of generalist humanoid whole-body controllers (see our [whitepaper](https:\u002F\u002Fnvlabs.github.io\u002FGEAR-SONIC\u002F));\n- **MotionBricks**: a real-time latent generative model for interactive motion control in animation and robotics (see the [project page](https:\u002F\u002Fnvlabs.github.io\u002Fmotionbricks\u002F)).\n\n## News\n\n- **[2026-05-07]** 🤖 **End-to-end VLA workflow on G1** — collect teleop data, fine-tune Isaac-GR00T N1.7, and deploy with SONIC whole-body control. See [Data Collection](https:\u002F\u002Fnvlabs.github.io\u002FGR00T-WholeBodyControl\u002Ftutorials\u002Fdata_collection.html), [VLA Workflow](https:\u002F\u002Fnvlabs.github.io\u002FGR00T-WholeBodyControl\u002Ftutorials\u002Fvla_workflow.html), and [VLA Inference](https:\u002F\u002Fnvlabs.github.io\u002FGR00T-WholeBodyControl\u002Ftutorials\u002Fvla_inference.html).\n- **[2026-04-27]** 🧩 **MotionBricks preview** — interactive G1 demo, pretrained checkpoints (VQVAE · pose · root), synthetic training code, and motion-representation docs. See [`motionbricks\u002F`](motionbricks\u002F) and the [project page](https:\u002F\u002Fnvlabs.github.io\u002Fmotionbricks\u002F).\n- **[2026-04-14]** 🌐 **[Live web demo](https:\u002F\u002Fnvlabs.github.io\u002FGEAR-SONIC\u002Fdemo.html)** — try SONIC interactively in your browser. Features [Kimodo](https:\u002F\u002Fgithub.com\u002Fnv-tlabs\u002Fkimodo) text-to-motion generation.\n- **[2026-04-10]** 🚀 Released **SONIC training code and checkpoint** on [HuggingFace](https:\u002F\u002Fhuggingface.co\u002Fnvidia\u002FGEAR-SONIC). Train from scratch or finetune. **Additional embodiment support** and **VLA data collection pipeline**. See [Training Guide](https:\u002F\u002Fnvlabs.github.io\u002FGR00T-WholeBodyControl\u002Fuser_guide\u002Ftraining.html).\n- **[2026-03-24]** 🔧 C++ inference stack update: motor error monitoring, TTS alerts, ZMQ protocol v4, idle-mode readaptation. **ZMQ header size changed to 1280 bytes.**\n- **[2026-03-16]** 📦 [BONES-SEED](https:\u002F\u002Fhuggingface.co\u002Fdatasets\u002Fbones-studio\u002Fseed) open-sourced — 142K+ human motions (~288 hours) with G1 MuJoCo trajectories.\n- **[2026-02-19]** 🎉 Released GEAR-SONIC: pretrained checkpoints, C++ inference, VR teleoperation, and documentation.\n- **[2025-11-12]** 🏁 Initial release with Decoupled WBC for GR00T N1.5 and N1.6.\n\n## Table of Contents\n\n- [News](#news)\n- [GEAR-SONIC](#gear-sonic)\n- [VR Whole-Body Teleoperation](#vr-whole-body-teleoperation)\n- [Kinematic Planner](#kinematic-planner)\n- [SONIC Training](#sonic-training)\n- [TODOs](#todos)\n- [What's Included](#whats-included)\n  - [Setup](#setup)\n- [Documentation](#documentation)\n- [Citation](#citation)\n- [License](#license)\n- [Support](#support)\n- [MotionBricks](#motionbricks)\n- [Decoupled WBC](#decoupled-wbc)\n\n\n## GEAR-SONIC \n\n\u003Cp style=\"font-size: 1.2em;\">\n    \u003Ca href=\"https:\u002F\u002Fnvlabs.github.io\u002FGEAR-SONIC\u002F\">\u003Cstrong>Website\u003C\u002Fstrong>\u003C\u002Fa> | \n    \u003Ca href=\"https:\u002F\u002Fhuggingface.co\u002Fnvidia\u002FGEAR-SONIC\">\u003Cstrong>Model\u003C\u002Fstrong>\u003C\u002Fa> | \n    \u003Ca href=\"https:\u002F\u002Farxiv.org\u002Fabs\u002F2511.07820\">\u003Cstrong>Paper\u003C\u002Fstrong>\u003C\u002Fa> | \n    \u003Ca href=\"https:\u002F\u002Fnvlabs.github.io\u002FGR00T-WholeBodyControl\u002F\">\u003Cstrong>Docs\u003C\u002Fstrong>\u003C\u002Fa>\n  \u003C\u002Fp>\n\n\u003Cdiv align=\"center\">\n  \u003Cimg src=\"docs\u002Fsource\u002F_static\u002Fsonic-preview-gif-480P.gif\" width=\"800\" >\n  \n\u003C\u002Fdiv>\n\nSONIC is a humanoid behavior foundation model that gives robots a core set of motor skills learned from large-scale human motion data. Rather than building separate controllers for predefined motions, SONIC uses motion tracking as a scalable training task, enabling a single unified policy to produce natural, whole-body movement and support a wide range of behaviors — from walking and crawling to teleoperation and multi-modal control. It is designed to generalize beyond the motions it has seen during training and to serve as a foundation for higher-level planning and interaction.\n\nIn this repo, we release SONIC's training code, deployment framework, model checkpoints, and teleoperation stack for data collection.\n\n\n## VR Whole-Body Teleoperation\n\nSONIC supports real-time whole-body teleoperation via PICO VR headset, enabling natural human-to-robot motion transfer for data collection and interactive control.\n\n\u003Cdiv align=\"center\">\n\u003Ctable>\n\u003Ctr>\n\u003Ctd align=\"center\">\u003Cb>Walking\u003C\u002Fb>\u003C\u002Ftd>\n\u003Ctd align=\"center\">\u003Cb>Running\u003C\u002Fb>\u003C\u002Ftd>\n\u003C\u002Ftr>\n\u003Ctr>\n\u003Ctd align=\"center\">\u003Cimg src=\"media\u002Fteleop_walking.gif\" width=\"400\">\u003C\u002Ftd>\n\u003Ctd align=\"center\">\u003Cimg src=\"media\u002Fteleop_running.gif\" width=\"400\">\u003C\u002Ftd>\n\u003C\u002Ftr>\n\u003Ctr>\n\u003Ctd align=\"center\">\u003Cb>Sideways Movement\u003C\u002Fb>\u003C\u002Ftd>\n\u003Ctd align=\"center\">\u003Cb>Kneeling\u003C\u002Fb>\u003C\u002Ftd>\n\u003C\u002Ftr>\n\u003Ctr>\n\u003Ctd align=\"center\">\u003Cimg src=\"media\u002Fteleop_sideways.gif\" width=\"400\">\u003C\u002Ftd>\n\u003Ctd align=\"center\">\u003Cimg src=\"media\u002Fteleop_kneeling.gif\" width=\"400\">\u003C\u002Ftd>\n\u003C\u002Ftr>\n\u003Ctr>\n\u003Ctd align=\"center\">\u003Cb>Getting Up\u003C\u002Fb>\u003C\u002Ftd>\n\u003Ctd align=\"center\">\u003Cb>Jumping\u003C\u002Fb>\u003C\u002Ftd>\n\u003C\u002Ftr>\n\u003Ctr>\n\u003Ctd align=\"center\">\u003Cimg src=\"media\u002Fteleop_getup.gif\" width=\"400\">\u003C\u002Ftd>\n\u003Ctd align=\"center\">\u003Cimg src=\"media\u002Fteleop_jumping.gif\" width=\"400\">\u003C\u002Ftd>\n\u003C\u002Ftr>\n\u003Ctr>\n\u003Ctd align=\"center\">\u003Cb>Bimanual Manipulation\u003C\u002Fb>\u003C\u002Ftd>\n\u003Ctd align=\"center\">\u003Cb>Object Hand-off\u003C\u002Fb>\u003C\u002Ftd>\n\u003C\u002Ftr>\n\u003Ctr>\n\u003Ctd align=\"center\">\u003Cimg src=\"media\u002Fteleop_bimanual.gif\" width=\"400\">\u003C\u002Ftd>\n\u003Ctd align=\"center\">\u003Cimg src=\"media\u002Fteleop_switch_hands.gif\" width=\"400\">\u003C\u002Ftd>\n\u003C\u002Ftr>\n\u003C\u002Ftable>\n\u003C\u002Fdiv>\n\n## Kinematic Planner\n\nSONIC includes a kinematic planner for real-time locomotion generation — choose a movement style, steer with keyboard\u002Fgamepad, and adjust speed and height on the fly.\n\n\u003Cdiv align=\"center\">\n\u003Ctable>\n\u003Ctr>\n\u003Ctd align=\"center\" colspan=\"2\">\u003Cb>In-the-Wild Navigation\u003C\u002Fb>\u003C\u002Ftd>\n\u003C\u002Ftr>\n\u003Ctr>\n\u003Ctd align=\"center\" colspan=\"2\">\u003Cimg src=\"media\u002Fplanner\u002Fplanner_in_the_wild_navigation.gif\" width=\"800\">\u003C\u002Ftd>\n\u003C\u002Ftr>\n\u003Ctr>\n\u003Ctd align=\"center\">\u003Cb>Run\u003C\u002Fb>\u003C\u002Ftd>\n\u003Ctd align=\"center\">\u003Cb>Happy\u003C\u002Fb>\u003C\u002Ftd>\n\u003C\u002Ftr>\n\u003Ctr>\n\u003Ctd align=\"center\">\u003Cimg src=\"media\u002Fplanner\u002Fplanner_run.gif\" width=\"400\">\u003C\u002Ftd>\n\u003Ctd align=\"center\">\u003Cimg src=\"media\u002Fplanner\u002Fplanner_happy.gif\" width=\"400\">\u003C\u002Ftd>\n\u003C\u002Ftr>\n\u003Ctr>\n\u003Ctd align=\"center\">\u003Cb>Stealth\u003C\u002Fb>\u003C\u002Ftd>\n\u003Ctd align=\"center\">\u003Cb>Injured\u003C\u002Fb>\u003C\u002Ftd>\n\u003C\u002Ftr>\n\u003Ctr>\n\u003Ctd align=\"center\">\u003Cimg src=\"media\u002Fplanner\u002Fplanner_stealth.gif\" width=\"400\">\u003C\u002Ftd>\n\u003Ctd align=\"center\">\u003Cimg src=\"media\u002Fplanner\u002Fplanner_injured.gif\" width=\"400\">\u003C\u002Ftd>\n\u003C\u002Ftr>\n\u003Ctr>\n\u003Ctd align=\"center\">\u003Cb>Kneeling\u003C\u002Fb>\u003C\u002Ftd>\n\u003Ctd align=\"center\">\u003Cb>Hand Crawling\u003C\u002Fb>\u003C\u002Ftd>\n\u003C\u002Ftr>\n\u003Ctr>\n\u003Ctd align=\"center\">\u003Cimg src=\"media\u002Fplanner\u002Fplanner_kneeling.gif\" width=\"400\">\u003C\u002Ftd>\n\u003Ctd align=\"center\">\u003Cimg src=\"media\u002Fplanner\u002Fplanner_hand_crawling.gif\" width=\"400\">\u003C\u002Ftd>\n\u003C\u002Ftr>\n\u003Ctr>\n\u003Ctd align=\"center\">\u003Cb>Elbow Crawling\u003C\u002Fb>\u003C\u002Ftd>\n\u003Ctd align=\"center\">\u003Cb>Boxing\u003C\u002Fb>\u003C\u002Ftd>\n\u003C\u002Ftr>\n\u003Ctr>\n\u003Ctd align=\"center\">\u003Cimg src=\"media\u002Fplanner\u002Fplanner_elbow_crawling.gif\" width=\"400\">\u003C\u002Ftd>\n\u003Ctd align=\"center\">\u003Cimg src=\"media\u002Fplanner\u002Fplanner_boxing.gif\" width=\"400\">\u003C\u002Ftd>\n\u003C\u002Ftr>\n\u003C\u002Ftable>\n\u003C\u002Fdiv>\n\n## SONIC Training\n\nSONIC can be trained from scratch on the [Bones-SEED](https:\u002F\u002Fhuggingface.co\u002Fdatasets\u002Fbones-studio\u002Fseed)\nmotion capture dataset (142K+ motions, ~288 hours, Unitree G1 retargeted), or finetuned\nfrom the released checkpoint on [Hugging Face](https:\u002F\u002Fhuggingface.co\u002Fnvidia\u002FGEAR-SONIC).\n\n### Quick start\n\n```bash\n# Install training dependencies (Isaac Lab must be installed separately — see docs)\npip install -e \"gear_sonic\u002F[training]\"\n\n# Download checkpoint + SMPL data from Hugging Face\npip install huggingface_hub\npython download_from_hf.py --training\n\n# Download Bones-SEED G1 CSVs from huggingface.co\u002Fdatasets\u002Fbones-studio\u002Fseed, then convert and filter\npython gear_sonic\u002Fdata_process\u002Fconvert_soma_csv_to_motion_lib.py \\\n    --input \u002Fpath\u002Fto\u002Fbones_seed\u002Fg1\u002Fcsv\u002F \\\n    --output data\u002Fmotion_lib_bones_seed\u002Frobot --fps 30 --fps_source 120 --individual --num_workers 16\npython gear_sonic\u002Fdata_process\u002Ffilter_and_copy_bones_data.py \\\n    --source data\u002Fmotion_lib_bones_seed\u002Frobot --dest data\u002Fmotion_lib_bones_seed\u002Frobot_filtered\n\n# Finetune from released checkpoint (64+ GPUs recommended)\naccelerate launch --num_processes=8 gear_sonic\u002Ftrain_agent_trl.py \\\n    +exp=manager\u002Funiversal_token\u002Fall_modes\u002Fsonic_release \\\n    +checkpoint=sonic_release\u002Flast.pt \\\n    num_envs=4096 headless=True \\\n    ++manager_env.commands.motion.motion_lib_cfg.motion_file=data\u002Fmotion_lib_bones_seed\u002Frobot_filtered \\\n    ++manager_env.commands.motion.motion_lib_cfg.smpl_motion_file=data\u002Fsmpl_filtered\n```\n\nFor the full guide including multi-node training, evaluation, ONNX export, and SOMA encoder setup:\n📖 [Installation (Training)](https:\u002F\u002Fnvlabs.github.io\u002FGR00T-WholeBodyControl\u002Fgetting_started\u002Finstallation_training.html) |\n[Training Guide](https:\u002F\u002Fnvlabs.github.io\u002FGR00T-WholeBodyControl\u002Fuser_guide\u002Ftraining.html)\n\n\n## TODOs\n\n- [x] Release pretrained SONIC policy checkpoints\n- [x] Open source C++ inference stack\n- [x] Setup documentation\n- [x] Open source teleoperation stack and demonstration scripts\n- [x] Release training scripts and recipes for motion imitation and fine-tuning\n- [ ] Open source large-scale data collection workflows and fine-tuning VLA scripts. \n- [ ] Publish additional preprocessed large-scale human motion datasets\n\n\n\n## What's Included\n\nThis release includes:\n\n- **`gear_sonic_deploy`**: C++ inference stack for deploying SONIC policies on real hardware\n- **`gear_sonic`**: Full SONIC training stack — PPO training, data processing pipeline, and configuration system for training on Bones-SEED and custom motion datasets\n- **`motionbricks`**: Preview release of the MotionBricks real-time latent generative stack — interactive G1 demo, pretrained checkpoints, synthetic training code, and motion-representation docs\n\n### Setup\n\n> **Git LFS required.** This repo contains large binary assets (meshes, ONNX\n> models). Without Git LFS, you will get small pointer files instead of actual\n> data, causing silent failures. Install Git LFS first if you don't have it:\n> `sudo apt install git-lfs && git lfs install`\n>\n> MotionBricks pretrained checkpoints are skipped by default to avoid an extra\n> ~2.2 GiB download during normal monorepo setup. MotionBricks GIFs and meshes\n> still download normally. Fetch the checkpoints explicitly if you plan to run\n> the MotionBricks demo.\n\n```bash\ngit clone https:\u002F\u002Fgithub.com\u002FNVlabs\u002FGR00T-WholeBodyControl.git\ncd GR00T-WholeBodyControl\ngit lfs pull\n\n# Optional: fetch MotionBricks pretrained checkpoints.\ngit lfs pull --include=\"motionbricks\u002Fout\u002F**\" --exclude=\"\"\n\n# Verify your environment\npython check_environment.py\n```\n\n### Which environment do I need?\n\n| I want to... | Environment | How to install |\n|---|---|---|\n| **Train \u002F finetune SONIC** | Isaac Lab's Python env | [Install Isaac Lab](https:\u002F\u002Fisaac-sim.github.io\u002FIsaacLab\u002Fmain\u002Fsource\u002Fsetup\u002Finstallation\u002Findex.html), then `pip install -e \"gear_sonic\u002F[training]\"` |\n| **Run MuJoCo simulation** | `.venv_sim` (auto-created) | `bash install_scripts\u002Finstall_mujoco_sim.sh` |\n| **VR teleoperation** | `.venv_teleop` (auto-created) | `bash install_scripts\u002Finstall_pico.sh` |\n| **Collect data** | `.venv_data_collection` (auto-created) | `bash install_scripts\u002Finstall_data_collection.sh` |\n| **Deploy on real robot** | C++ build | See [deployment docs](https:\u002F\u002Fnvlabs.github.io\u002FGR00T-WholeBodyControl\u002Fgetting_started\u002Finstallation_deploy.html) |\n\nEach use case has its own lightweight environment. The install scripts use `uv`\nand create isolated venvs automatically — you don't need to manage them manually.\nTraining is the only one that requires Isaac Lab (installed separately).\n\n## Documentation\n\n📚 **[Full Documentation](https:\u002F\u002Fnvlabs.github.io\u002FGR00T-WholeBodyControl\u002F)**\n\n### Getting Started\n- [Installation Guide](https:\u002F\u002Fnvlabs.github.io\u002FGR00T-WholeBodyControl\u002Fgetting_started\u002Finstallation_deploy.html)\n- [Quick Start](https:\u002F\u002Fnvlabs.github.io\u002FGR00T-WholeBodyControl\u002Fgetting_started\u002Fquickstart.html)\n- [VR Teleoperation Setup](https:\u002F\u002Fnvlabs.github.io\u002FGR00T-WholeBodyControl\u002Fgetting_started\u002Fvr_teleop_setup.html)\n\n### Tutorials\n- [Keyboard Control](https:\u002F\u002Fnvlabs.github.io\u002FGR00T-WholeBodyControl\u002Ftutorials\u002Fkeyboard.html)\n- [Gamepad Control](https:\u002F\u002Fnvlabs.github.io\u002FGR00T-WholeBodyControl\u002Ftutorials\u002Fgamepad.html)\n- [ZMQ Communication](https:\u002F\u002Fnvlabs.github.io\u002FGR00T-WholeBodyControl\u002Ftutorials\u002Fzmq.html)\n- [ZMQ Manager \u002F PICO VR](https:\u002F\u002Fnvlabs.github.io\u002FGR00T-WholeBodyControl\u002Ftutorials\u002Fvr_wholebody_teleop.html)\n\n### Training\n- [Installation (Training)](https:\u002F\u002Fnvlabs.github.io\u002FGR00T-WholeBodyControl\u002Fgetting_started\u002Finstallation_training.html)\n- [Training Guide](https:\u002F\u002Fnvlabs.github.io\u002FGR00T-WholeBodyControl\u002Fuser_guide\u002Ftraining.html)\n- [Training Data](https:\u002F\u002Fnvlabs.github.io\u002FGR00T-WholeBodyControl\u002Fuser_guide\u002Ftraining_data.html)\n\n### Best Practices\n- [Teleoperation](https:\u002F\u002Fnvlabs.github.io\u002FGR00T-WholeBodyControl\u002Fuser_guide\u002Fteleoperation.html)\n\n\n\n\n\n\n---\n\n## Citation\n\nIf you use GEAR-SONIC in your research, please cite:\n\n```bibtex\n@article{luo2025sonic,\n    title={SONIC: Supersizing Motion Tracking for Natural Humanoid Whole-Body Control},\n    author={Luo, Zhengyi and Yuan, Ye and Wang, Tingwu and Li, Chenran and Chen, Sirui and Casta\\~neda, Fernando and Cao, Zi-Ang and Li, Jiefeng and Minor, David and Ben, Qingwei and Da, Xingye and Ding, Runyu and Hogg, Cyrus and Song, Lina and Lim, Edy and Jeong, Eugene and He, Tairan and Xue, Haoru and Xiao, Wenli and Wang, Zi and Yuen, Simon and Kautz, Jan and Chang, Yan and Iqbal, Umar and Fan, Linxi and Zhu, Yuke},\n    journal={arXiv preprint arXiv:2511.07820},\n    year={2025}\n}\n```\n\n---\n\n## License\n\nThis project uses dual licensing:\n\n- **Source Code**: Licensed under Apache License 2.0 - applies to all code, scripts, and software components in this repository\n- **Model Weights**: Licensed under NVIDIA Open Model License - applies to all trained model checkpoints and weights\n\nSee [LICENSE](LICENSE) for the complete dual-license text.\n\nPlease review both licenses before using this project. The NVIDIA Open Model License permits commercial use with attribution and requires compliance with NVIDIA's Trustworthy AI terms.\n\nAll required legal documents, including the Apache 2.0 license, 3rd-party attributions, and DCO language, are consolidated in the \u002Flegal folder of this repository.\n\n---\n\n## Support\n\nFor questions and issues, please contact the GEAR WBC team at [gear-wbc@nvidia.com](mailto:gear-wbc@nvidia.com) to provide feedback! \n\n## MotionBricks\n\n\u003Cp style=\"font-size: 1.2em;\">\n  \u003Ca href=\"https:\u002F\u002Fnvlabs.github.io\u002Fmotionbricks\u002F\">\u003Cstrong>Project page\u003C\u002Fstrong>\u003C\u002Fa> |\n  \u003Ca href=\"motionbricks\u002FREADME.md\">\u003Cstrong>Subproject README\u003C\u002Fstrong>\u003C\u002Fa>\n\u003C\u002Fp>\n\n\u003Cdiv align=\"center\">\n  \u003Cimg src=\"motionbricks\u002Fassets\u002Fgifs\u002Fteaser_animation.gif\" width=\"400\">\n  \u003Cimg src=\"motionbricks\u002Fassets\u002Fgifs\u002Fteaser_robotics.gif\" width=\"400\">\n\u003C\u002Fdiv>\n\nMotionBricks is a real-time generative framework that transforms interactive motion control for animation and robotics. It combines a large-scale latent backbone with intuitive \"smart primitives\" to deliver high-quality, zero-shot motion synthesis at 15,000 FPS — complementing the tracking-based GEAR-SONIC controllers in this repo.\n\nThis preview release ships an interactive G1 demo (keyboard-driven, MuJoCo viewer), pretrained checkpoints (VQVAE · pose · root), a synthetic training pipeline, and motion-representation docs. Its pretrained checkpoints are opt-in for monorepo clones; run `git lfs pull --include=\"motionbricks\u002Fout\u002F**\" --exclude=\"\"` from the repo root before using the demo. A full release — fully embedded in the GEAR-SONIC pipeline — is targeted for approximately one month out. See [`motionbricks\u002FREADME.md`](motionbricks\u002FREADME.md) for setup, demo, and training instructions.\n\n## Decoupled WBC\n\nFor the Decoupled WBC used in GR00T N1.5 and N1.6 models, please refer to the [Decoupled WBC documentation](docs\u002Fsource\u002Freferences\u002Fdecoupled_wbc.md).\n\n\n## Acknowledgments\nWe would like to acknowledge the following projects from which parts of the code in this repo are derived from:\n- [Beyond Mimic](https:\u002F\u002Fgithub.com\u002FHybridRobotics\u002Fwhole_body_tracking)\n- [Isaac Lab](https:\u002F\u002Fgithub.com\u002Fisaac-sim\u002FIsaacLab)\n","GR00T Whole-Body Control (WBC) 是一个统一的平台，用于开发和部署先进的人形机器人控制器。该项目的核心功能包括解耦的WBC模型（如NVIDIA Isaac-Gr00t、Gr00t N1.5和N1.6以及GEAR-SONIC系列），这些模型结合了强化学习和逆运动学来控制机器人的下身和上身动作。此外，它还支持MotionBricks，这是一种实时潜变量生成模型，适用于动画和机器人中的交互式运动控制。此项目适合于需要高度灵活性和精确度的人形机器人控制场景，例如在研究实验室、教育机构或工业自动化中进行复杂任务的执行。",2,"2026-06-11 03:32:16","trending"]