VTAP Gripper combines fingertip sensing and active palm for dexterous robot manipulation

The VTAP Gripper, developed by Yuhao Zhou and collaborators at Purdue University and Columbia University, introduces a novel approach to robotic dexterity by synergizing fingertip sensing with an active palm that combines vision and touch. The gripper features a Visuo-Tactile Active Palm (VTAP) that seamlessly switches between long-range visual localization and contact-rich tactile feedback, extending manipulation capabilities beyond traditional grippers. Its three compliant, reconfigurable fingers are equipped with tactile array sensors, allowing the system to finely adjust grip during tasks.

The system was validated through a series of challenging tasks: reactive grasping of YCB benchmark objects and fragile items, in-hand syringe reorientation and plunger actuation, singulation of clustered objects down to 3 mm in diameter, and vision-tactile peg-in-hole insertion. A key innovation is a staged, gesture-conditioned retargeting framework that bridges the embodiment gap between human hand motion and the heterogeneous three-finger structure, enabling dexterous teleoperation. The work demonstrates that high manipulation performance can be achieved through coordinated finger-palm interaction and multi-modal sensing without resorting to high-DOF anthropomorphic designs, offering a practical architecture for learning-based robotics.