{"version":1,"type":"story","url":"https://digestai.news/story/noitom-robotics-releases-hiphi-dataset-for-humanoid-robot-training","json":"https://digestai.news/story/noitom-robotics-releases-hiphi-dataset-for-humanoid-robot-training.json","markdown":"https://digestai.news/story/noitom-robotics-releases-hiphi-dataset-for-humanoid-robot-training.md","slug":"noitom-robotics-releases-hiphi-dataset-for-humanoid-robot-training","headline":"Noitom Robotics releases HiPHI dataset for humanoid robot training","summary":"Noitom Robotics has published a white paper introducing HiPHI, a large-scale motion capture dataset designed to address data limitations in humanoid robot learning. The dataset contains 617.5 hours of whole-body human motion recorded with optical motion capture at sub-millimeter accuracy. It specifically includes 245.7 hours of human-object interaction data, featuring synchronized object trajectories and meshes to help robots learn tasks like carrying, pushing, and pulling.\nThe paper explains that internet video lacks precise physical state data, while existing lab motion capture often covers too narrow a range of actions. To solve this, the researchers used FrameNet, a linguistic framework for human action, to guide the collection process and ensure broad coverage of whole-body movements. The white paper also presents a benchmark suite for measuring motion diversity and interaction grounding.\nResults show that reinforcement learning policies trained on this data improve with scale. The authors report that these policies successfully transfer from simulation to a physical Unitree G1 humanoid robot, demonstrating the dataset's utility for real-world deployment. This work aims to close the data gap that currently limits the development of embodied AI and Physical AI systems.","keyPoints":["HiPHI is a 617.5-hour dataset with sub-millimeter accuracy for humanoid robot training.","It includes 245.7 hours of human-object interaction with synchronized object trajectories.","Policies trained on HiPHI transfer successfully to a physical Unitree G1 robot."],"whyItMatters":"This dataset provides the high-precision, diverse motion data needed to train humanoid robots for real-world tasks, addressing a key bottleneck in embodied AI development.","category":{"slug":"robotics","name":"Robotics & Physical AI","url":"https://digestai.news/category/robotics"},"entities":{"companies":["Noitom Robotics","Unitree","IEEE Spectrum","Wiley"],"models":[],"people":[]},"firstPublishedAt":"2026-10-07T13:49:48Z","updatedAt":"2026-10-07T13:49:48Z","sourceCount":1,"hasPrimarySource":false,"sources":[{"outlet":"IEEE Spectrum Robotics","title":"HiPHI: A Large-Scale Benchmark for High-Precision Human Motion and Object Interaction","url":"https://content.knowledgehub.wiley.com/hiphi-a-large-scale-benchmark-for-high-precision-human-motion-and-object-interaction","publishedAt":"2026-10-07T13:49:48Z","type":"press","primary":false,"lead":true}],"sourceNotes":null,"discussions":[],"thread":{"title":"Humanoid Robots Learn Adaptive Motion Planning","url":"https://digestai.news/thread/ieee-spectrum-outlines-adaptive-motion-planning-for-humanoid-robots","storyCount":2},"cite":{"text":"Digest AI, \"Noitom Robotics releases HiPHI dataset for humanoid robot training\", 7 October 2026, https://digestai.news/story/noitom-robotics-releases-hiphi-dataset-for-humanoid-robot-training","publisher":"Digest AI","title":"Noitom Robotics releases HiPHI dataset for humanoid robot training","datePublished":"2026-10-07T13:49:48Z","url":"https://digestai.news/story/noitom-robotics-releases-hiphi-dataset-for-humanoid-robot-training"},"generatedBy":"Written by Digest AI's editorial model from the linked sources; the sources are the record.","license":"Headlines, digests and key points are written by Digest AI and may be quoted with a link to the story page. Linked articles belong to their publishers. Terms: https://digestai.news/terms#reuse"}