A University of Electro-Communications prototype reproduced human-like wrist changes in controlled laboratory tests, not a complete humanoid demonstration.
Researchers at the University of Electro-Communications in Tokyo built a soft robotic forearm to test whether wrist anatomy can perform part of the control work. The arXiv paper describes eight independently movable carpal bones, 22 actuated muscles, articulated fingers, and compliant silicone fingertips.
Stiffness means how strongly a joint resists movement. It can vary by direction: a wrist may stay flexible along one path while resisting force along another. The researchers measured that behavior with a force gauge under four muscle-activation patterns.
With the anatomically structured wrist, activating the finger muscles left the wrist relatively compliant along the “dart-throwing” path, a diagonal motion from radial extension to ulnar flexion. Activating the wrist and finger muscles together made that same direction much stiffer. The paper reports a 48.2-degree shift in the direction of maximum stiffness.
The comparison designs showed why separate bones mattered. Fusing the proximal carpal row removed the low-stiffness behavior under finger-muscle activation. A simpler ellipsoid-shaped skeleton produced high stiffness but barely changed its orientation as activation changed.
Motion measurements found significant condition-dependent changes in the proximal carpal row. Changes at the midcarpal joint did not reach statistical significance, pointing to the proximal row as the stronger measured contributor.
This remains a laboratory design study, not a deployable humanoid product. The team tested one physical prototype under slow, quasi-static loading—controlled motion rather than rapid manipulation—and the findings have not been independently replicated. The useful next test is whether the same wrist behavior holds during fast tasks and across multiple builds.