J-HRTI, short for Japan Humanoid Robot Training & Implementation, opened its Kanto Data Factory in Narashino on August 26 with 35 humanoid robots across about 1,400 square metres. The centre can expand to 40 machines and is designed to generate, label and test the physical data needed to train robots for practical industrial work.
The consortium says the facility is Japan’s first physical-AI and robot data factory operated by a group composed solely of private companies. Its model is intended to reduce the cost and duplication involved when manufacturers try to build large robot-training systems independently, while allowing participating businesses to share common training data and testing infrastructure.
Operators telecontrol the humanoids to create imitation-learning data from tasks such as gripping, moving and handling objects. Cameras mounted on the robots record movements while motor parameters are also collected. Specialist staff then annotate the footage and motion information so AI systems can distinguish actions, errors and acceptable performance before the robots are tested in environments designed to resemble participating companies’ workplaces.
J-HRTI says members divide the cost of equipment and data generation while contributing real-world tasks from their own operations. Common datasets can be reused across companies, reducing the thousands or tens of thousands of training repetitions that an individual business might otherwise need to undertake from scratch. Sensitive company-specific work can be handled separately in a secure area, where information is not shared with other members.
That approach could widen access to humanoid development beyond the largest manufacturers, although J-HRTI has not published membership fees or a standard deployment price for factories. Claims that the system makes humanoids broadly affordable for mid-sized manufacturers therefore remain difficult to quantify. What is established is that participating companies can spread investment and technical risk rather than financing a dedicated data facility alone.
The operating group includes machinery trader Yamazen, pharmaceutical manufacturer Tsumura, direct-mail and logistics company DMS and Fuyo General Lease, alongside another participant that was not publicly identified in the August 26 announcement. INSOL-HIGH serves as the consortium secretariat and operates the data platform and overall system design.
J-HRTI’s programme is aimed initially at straightforward industrial duties. The group plans to place humanoids into workplaces for tasks such as material transport during the 2026 financial year, which ends in March 2027. It then intends to broaden the range of tasks and improve autonomy as more operational data becomes available.
The consortium is also developing a physical-AI foundation model using data collected at the Chiba site. Several thousand datasets had already been accumulated and annotated by the time full operations began, according to INSOL-HIGH officials. Participating companies are expected to use the model and return additional operating data and feedback, creating a cycle in which the shared training resource is continually updated.
Plans call for the data-factory network to expand to 10 locations nationwide during the 2027 financial year. The Chiba facility is intended to serve as the first “mother” site, with future centres positioned closer to industries that can supply specialised tasks and operational conditions. J-HRTI has said those regional facilities could also support maintenance after robots are deployed.
By the 2028 financial year, the group aims to provide foundation models specialised for sectors including manufacturing and logistics. The timetable remains a target rather than a guaranteed commercial rollout, and performance will depend on whether robots trained in controlled environments can operate safely and reliably in varied workplaces.
Follow Arabian Post
Select Arabian Post as your preferred source on Google and MSN News for trusted business news and Arab politics and updates.
