News
icub-models
Official URDF and SDF models of the iCub humanoid robot.
IITHumanoid | DailyMecha
Explore source-linked Humanoid research, news and demonstrations. Short summaries retain the original sources and distinguish reported findings from unverified claims.
News
ROBOTIS-THORMANG-MPC
ROS packages for the motion of THORMANG3, MPC means Motion PC.
ROBOTISNews
ROBOTIS-OP3
ROS packages for the ROBOTIS OP3
ROBOTISNews
roboto_origin
Roboto_origin Fully Open-Source DIY Humanoid Robot/萝博头原型机全开源手搓级人形机器人
RoboPartyNews
microduck
A Tiny biped duck robot 🦆
Pollen RoboticsNews
nimbro-op
Sourcecode & CAD drawings of NimbRo-OP
University of BonnNews
ToddlerBot
toddlerbot
Stanford UniversityNews
berkeley-humanoid-lite
Codebase for Berkeley Humanoid Lite v1
UC BerkeleyNews
Open_Duck_Mini
Making a mini version of the BDX droid. https://discord.gg/UtJZsgfQGe
Open Duck research projectNews
AI Sapiens
ROBOTIS publishes ROS 2 packages for AI Sapiens, including controllers, hardware interfaces, robot descriptions and simulation support. The repository identifies Apache-2.0 as its root software license. This listing confirms the visible software packages; it does not imply that every hardware design, weight file or third-party dependency has the same license.
ROBOTISNews
Generative Bionics plans Italian Physical AI Data Factory for humanoids
Generative Bionics plans a 5,000-square-meter Physical AI Data Factory in Italy, targeted to become operational by Q4 2027. Fleets of its GENE.01 humanoid will perform industrial tasks in reproduced work environments, collecting vision, motion, and full-body tactile data. The company says physical data and digital-twin simulations will support training and validation before customer deployment; AMD is its planned AI infrastructure partner. The facility is an announced plan, not an operational site.
The Robot ReportResearch
Humanoid Mobility via Attention-based Contacts
The framework controls a Unitree G1 humanoid using an attention-based contact planner and a contact-conditioned whole-body tracker. Motion data serves only as a training prior; reinforcement-learning fine-tuning helps the policy handle terrain variations beyond the demonstrations. At deployment, it uses onboard observations without reference inputs, enabling one policy to perform tasks including ladder and stair climbing, crawling, vaulting, and stepping-stone traversal.
arXiv cs.RO