Physical AI News (Japan)

Physical AI, robotics and compute — reported from Japan.

日本語
· Policy & regulation· 日本語で読む

METI and NEDO report selection of 13 companies and groups for GENIAC robot foundation model development support

Thirteen companies and groups were selected for a robot foundation model research and development project under GENIAC, led by the Ministry of Economy, Trade and Industry and NEDO. At a project plan presentation on Sept. 9, plans were outlined for preparing field data and collecting task data, including tactile data. The focus is on how to implement data collection and structuring to support foundation models that control robots.

What happened

The Ministry of Economy, Trade and Industry and NEDO held a project plan presentation and other events in Tokyo on Sept. 9, 2026, under GENIAC, a generative AI development support project. According to a report by RoboStar, 13 companies and groups were selected for the “Robot Foundation Model Research and Development Project (Multipurpose Robots, etc.).” At the event, plans were presented to collect field data from manufacturing, lifestyle services, medical care and robot operations. A consortium led by Kawasaki Heavy Industries explained its plan to collect data for a “VTLA” foundation model integrating vision, touch, language and action, together with Osaka University, FANUC, FingerVision and Yaskawa Electric.

Implications for Japan

In Japan, companies and institutions with field data such as manufacturing equipment operating data, work logs and surgical video are likely to play an increasingly important role in determining the inputs for robot foundation models. Plans involving Komatsu, the National Institute of Advanced Industrial Science and Technology, Kawasaki Heavy Industries, Osaka University, FANUC and Yaskawa Electric will help show whether existing machinery, sensor and control data can be turned into training data.

Sources

Translated from the Japanese original: /news/keisan-shou-to-nedo-no-geniac-saitaku-jigyousha-ga-deeta-ren-e92f5f. The Japanese version is authoritative if they differ.