Yukiyasu Domae

The embodiment gap reveals that even advanced robot foundation models require substantial adaptation to function on specific robotic platforms, challenging the assumption that scaling alone suffices for generalization.

Immediate feedback-driven corrections can enhance robotic manipulation performance without the overhead of retraining entire policies.

TWINS revolutionizes manipulation learning by enabling robots to learn from body-surface contact, a previously overlooked aspect of human-robot interaction.

Correctable visual attention can drastically enhance robot manipulation performance, especially in unpredictable environments.

YUBI's finger-driven design enables seamless data collection for bimanual manipulation, yielding a dataset that allows for effective policy transfer across diverse robotic systems.