02
BEFORE / AFTER

What changed from existing technology?

Existing technology

This study teaches robots to move by learning how their movements relate to motor inputs, without needing to know their physical structure beforehand.

This research

The method learns the relationship between movement and motor inputs without prior knowledge of the robot's structure.

03 · REALITY CHECK

Current stage: Basic research

Classified from the research published in Nature and its publicly reported validation. Manufacturing and revenue evidence require separate confirmation.

Current stageBasic research
Basic researchExperimental validationPrototypeField validationNear commercialization
04
REMAINING CHALLENGES

What still needs to be solved?

Follow-up validation points are being checked.

Solving these issues would increase the chance of reaching products and industry use.

05
INDUSTRY OPPORTUNITY

What could change if this succeeds?

Follow the path from technology to materials, components, equipment, and services.

06 · ONE MORE POSSIBILITYOne possibility beyond the paper

This approach could be used to control robots in environments where the robot's structure is unknown or varies significantly.

08 · EVIDENCE & SOURCES

Evidence and sources

View paper and company evidenceKey paperControlling diverse robots by inferring Jacobian fields with deep networksNature · 2025.06.25 ↗