Temporal Tactile Encoding and Compliance for Intent-Aware Robot-to-Human Bimanual Handover

Robotic handovers often falter because visual observations cannot reliably distinguish deliberate taking intent from an accidental brush or a weak grasp. This study pairs a Vision-Language-Action policy with temporal tactile encoding and compliance control for bimanual robot-to-human object transfers. User studies show that temporal tactile history clarifies human readiness before releasing, while active compliance dampens contact forces to deliver a noticeably smoother transfer.