Summary: This project aims to advance human–robot collaboration by enabling robots to perceive, interpret, and anticipate human intentions through visual sensing, learning, and uncertainty-aware decision making. It addresses four interconnected research problems:

(1) perception-enabled intention prediction for safe and fluent human–robot handovers,

(2) visual understanding for interpretable robot imitation learning from human demonstrations,

(3) modeling of coupled human–robot–environment dynamics,

(4) methods and datasets for transferring human–human interaction models to human–robot collaboration.