Japan advances physical AI through humanoid robot data hubs and development programs

Japan is seeing a surge in developments within the physical AI and humanoid robotics sectors, driven by new infrastructure and specialized support programs.

In Narashino, the J-HRTI consortium, operated by INSOL-HIGH, has launched the Kanto Data Factory. This facility utilizes up to 40 humanoid robots to collect essential training data for physical AI, allowing operators to synchronize movements via VR to teach robots complex tasks like sorting and grasping.

AWS Japan is accelerating this ecosystem through its Physical AI Development Support Program. Recent results from the program show significant progress in robotics foundation models. For instance, FastLabel demonstrated that mixing simulation data with real-world data can improve robot task success rates by an average of 14.2%, potentially reducing the high costs of physical data collection.

On the implementation side, Brains Technology has introduced a service specifically for manufacturing sites. Their solution integrates robot foundation models with AI anomaly detection to assist in the deployment, verification, and maintenance of various robotic systems, including humanoid and quadruped robots.