ugo unveils domestically developed semi-humanoid robot ‘ugo Nova,’ with mass production scheduled to begin in 2027

prtimes.jp reported on September 16, 2026, that ugo had unveiled the domestically developed semi-humanoid robot ‘ugo Nova,’ with mass production scheduled to begin in 2027. We have not been able to verify the primary source, so please refer to the original for details.

In its announcement, ugo positioned ugo Nova as a robot that can not only “collect data” but also “run models.” By presenting both a means of acquiring training data and an on-device inference platform, alongside the robot’s specifications, the company makes it easier for prospective users to assess the design of the full operation rather than the robot alone. The announcement also says ugo will handle development, mechanical design and manufacturing; GK Kyoto will handle industrial design; and Asratec Corp. will provide the bilateral controller, offering a glimpse of the division of labor in domestic development.

ugo Nova’s novelty lies less in its form as a dual-arm semi-humanoid than in its combining data collection and inference execution in the robot itself. By bringing together cameras, 100 Hz control, a bilateral controller, a VR controller and Jetson Thor, the design enables a single unit to handle the process from recording work demonstrations to model training and deployment on the robot. It is closer to a development and operations platform for physical AI than a standalone robot for sale. The planned 2027 start of mass production also suggests that the product has moved beyond the concept stage and into pre-production, though many conditions for mass production remain unresolved. Compared with this newspaper’s related article, ugo selected for GENIAC to develop VTLA model for dual-arm semi-humanoid robots (September 9, 2026), this announcement connects information about the selection of software research to the presentation of physical hardware. If VTLA model development lays the groundwork for the “learning” side, ugo Nova makes the “training” and “operation” sides more concrete. However, public materials do not yet show in detail how closely the research framework described in the selection announcement is linked to the roadmap for the mass-produced robot. The announcement also points to a potential shift in the industry, away from separately procuring robots, real-world data, foundation models and field implementation, and toward a single vendor providing them as an integrated offering. The specifications—4 kg per arm and 10 kg for both arms—suggest intended use in tasks such as transport and picking in manufacturing, factories and warehouses. The focus will be on how well six hours of continuous operation and mecanum-wheel mobility address constraints on long-duration use and movement in confined spaces. Equipping the robot with Jetson Thor could reduce reliance on the cloud and provide greater flexibility for deployment in the field, but heat, power consumption, BOM and maintenance will become production considerations. Unresolved issues include the specific production volume, price, manufacturing setup, operating rate in real-world use and range of tasks targeted for training with the collected data ahead of the planned 2027 start of mass production. The announcement presents specifications and a concept; the next questions are how far autonomy will advance in the field and how much responsibility the company will take for operations from data collection through model updates.

The fact that design through manufacturing will take place domestically, with development, industrial design and control systems divided among Japanese companies, has implications for how Japan combines robotics components, design and control technologies. Companies with expertise in dual-arm mechanisms and teleoperation components and control technologies will be watching which processes are handled in-house and which are outsourced once mass production begins.