Vibrations At First Contact Encode Object Stiffness Before Grasp Completion

Current methods for stiffness estimation in artificial dexterous systems depend on force and displacement measurements after an object is fully grasped. This delays grasp adjustments until the grasping process is complete. The proposed approach in this study introduces a novel method for stiffness estimation at first contact, exploiting the time window created by “asymmetry” in grasp dynamics when one finger contacts an object before the others. We integrated a biomimetic tactile sensor inspired by human skin anatomy into an artificial dexterous hand that simultaneously captures both force and vibration information. Vibrational signals from the first finger's initial contact were found to encode stiffness information, achieving high-accuracy discrimination (98.6%) and regression (error: 2.39 Shore A) of object stiffness using machine learning models. Despite the brief temporal window (mean: 16.65 ms), models could infer stiffness in under 1.5 ms, allowing time for grasp modulation with the remaining fingers before the object is fully secured. Vibration sensing in object grasping establishes a new paradigm for rapid stiffness estimation, with significant potential for advancing dexterous manipulation in delicate tasks for robotics.