Tactile Sensing-based Secure and Gentle Grasping of Unknown Objects

This research presents a novel online data-driven predictive control framework for safe dynamic robotic grasping of unknown objects. The system uses compact tactile-sensing modules to characterize gripper-object interactions in real time, enabling robust grasping without prior knowledge. A Gaussian Process model captures interaction dynamics, and model uncertainties are used to form robust constraints within a model predictive control scheme. The method is validated across diverse object types and contact surfaces, showing superior performance compared to baseline strategies.