Chair: Nima Fazeli
March 11, 2026, 9am
Ford Robotics Building 2300 and on Zoom

Abstract
Contact is central to dexterous manipulation, yet it remains one of the hardest aspects of robotics to model, control, and perceive. In this dissertation, we investigate how robots can better reason about prehensile contact through a unified perspective of modeling, action, and perception.

We develop analytical models for frictional patch contact in planar manipulation, show how fingertip micro-vibrations can enable in-hand object reconfiguration with simple grippers, and present a multimodal framework that combines vision and active audio sensing to estimate contact under occlusion. We also explore a vibration-based tactile sensing direction for inferring touch location and force from structured acoustic signals.

Together, these contributions help make contact-rich manipulation more predictable, observable, and effective, advancing robotic systems that can interact more robustly with the physical world.