A Novel Fingertip Tactile Sensory System for Humanoid Full-Hand Robotic Grasping

摘要

The humanoid robotic hand is a highly intelligent device capable of providing effective object grasping. A challenging issue is to integrate the tactile perception and adjust the grasping force according to object physical properties or environmental conditions during manipulative tasks. This study collects grasp data from healthy individuals to obtain a pre-grasp dataset and motion trajectories, and introduces a novel fingertip tactile sensory system for humanoid full-hand robotic grasping. A resistive force-sensitive sensor (FSR) signal acquisition system was designed and developed following Anderson loop, and the real-time contact forces between the fingertips and the object was acquired. A feedback control was designed and applied to modulate the fingertip forces for object grasping. The movement trajectories of the robotic hand were planned according to Dynamic movement primitives (DMP) based on human demonstrations captured using a motion capture system (Mocap). A Gaussian mixture regression model (GMM-GMR) was employed to establish a pre-grasping dataset for the manipulator. Results showed that the robotic hand could grasp three types of objects with distinct shapes. The success rate of grasping with the novel tactile system reached up to 85.2%. The study provides the feasibility of the system in humanoid full-hand grasping and may play role in a variety of object grasping and manipulation.