Intelligence begins in the hand.
CLAVIR explores how robotic hands can be part of the learning system.
What kind of hand is suitable for learning?
CLAVIR starts from a simple view: the hand should not be only an actuator. It should be part of the conditions that make learning possible.
A hand suited for learning should expose meaningful interaction signals rather than hide them behind rigid abstraction.
Learning works better when contact is not treated as a failure mode, but as a structured part of manipulation.
The right hand should allow rich, continuous control over contact and force for learning to explore.
Robotic hands are not built for learning.
Most robotic hands are designed as precise actuators, not as systems that enable learning. This makes contact-rich manipulation extremely difficult to learn.
The hand itself must be intelligent.
A dexterous hand should not only execute commands. Its physical structure should expose information, stabilize interaction, and enable learning to explore.
Hardware designed for learning.
CLAVIR explores robotic hands whose physical structure improves the quality and stability of learning in contact-rich manipulation.