Kufusha and 5 Other Organizations Selected for METI Decommissioning Project to Develop Physical AI Robots for Hazardous Environments

Kufusha Inc. (Headquarters: Sagamihara City; Minamisoma Office: Minamisoma City, Fukushima; CEO: Iichiro Onishi), in a joint development with the Japan Atomic Energy Agency (JAEA), the Zhao Laboratory at the University of Tokyo's Graduate School of Engineering, EQUES Inc. (a startup from the Matsuo-Iwasawa Lab at the University of Tokyo), DigiLabo Holdings Inc., and i-PLANTS systems Inc., has been selected for a subsidy project by the secretariat for the Decommissioning and Contaminated Water Management Program.

This project will commence the development of physical AI robot technology aimed at supporting investigation and work in high-radiation environments.

## Adopted Project Name

FY2026 Subsidy for Decommissioning and Contaminated Water Management Project (Third Round)
'Development of Robot Utilization Technology with Physical AI for Remote Work in Hazardous Environments'

## Background

In some areas of the Fukushima Daiichi Nuclear Power Plant, there are locations where the extent of equipment damage is unknown due to the nuclear accident, and environments with persistently high radiation levels remain. In such conditions, while investigation and preparatory work are needed to reduce radiation levels and secure work areas, human entry itself poses a significant risk.

Furthermore, remote operations require advanced piloting skills and extensive preparation, creating a dependency on skilled operators. Establishing next-generation robot technology capable of handling these harsh environments is an urgent task to safely and steadily advance decommissioning work.

## Objective

With recent advancements in AI technology, 'Physical AI,' where robots autonomously act by perceiving and making judgments about their surroundings, is gaining attention. This project will develop robot technology utilizing Physical AI to safely and reliably conduct investigations and tasks in high-radiation environments.

The project will work to verify the feasibility of robots supporting or replacing tasks by perceiving and judging the on-site environment, in place of skilled operators and preparation workers.

## What is Physical AI?

Physical AI is a technology where an AI perceives the real-world environment and acts physically as a robot. As a next-generation robot technology that can move, investigate, and work on behalf of humans while understanding its surroundings, it is expected to be applied in a wide range of fields, from manufacturing, construction, and infrastructure maintenance to decommissioning work.

## Participating Organizations

- Kufusha Inc.
- Japan Atomic Energy Agency (JAEA)
- Zhao Laboratory, Graduate School of Engineering, The University of Tokyo
- EQUES Inc. (A startup from the Matsuo-Iwasawa Laboratory at The University of Tokyo)
- DigiLabo Holdings Inc.
- i-PLANTS systems Inc.

Each organization will leverage its expertise to advance the project: Kufusha (integrated development of the physical AI robot system), JAEA (specialized knowledge in decommissioning), the University of Tokyo's Zhao Lab (swarm intelligence and flying robot technology), EQUES (point cloud processing and data infrastructure), DigiLabo HD (digital simulation), and i-PLANTS (automatic 3D model generation).

Kufusha, JAEA, and DigiLabo Holdings are based in the Hamadori region of Fukushima and will advance the project in close collaboration with the region's industrial recovery and decommissioning research efforts.

## About Kufusha Inc.

Kufusha Inc. is a robotics venture dedicated to 'implementing Physical AI in society.' The company focuses on R&D of physical AI using deep reinforcement learning and imitation learning, primarily with humanoid, quadruped, and dual-arm robots.

With development bases in Sagamihara City, Kanagawa, and Minamisoma City, Fukushima, the company is engaged in development and demonstration in realistic environments at the Fukushima Robot Test Field in Minamisoma.

Kufusha will continue to tackle various on-site challenges, including hazardous environments inaccessible to humans, through physical AI and robotics technology.