Underactuated dexterous robotic grasping with reconfigurable passive joints
Marek Kopicki, Sainul Islam Ansary, Simone Tolomei, Franco Angelini, Manolo Garabini, Piotr Skrzypczyński
TL;DR
This work tackles dexterous grasping with underactuated hands by introducing reconfigurable passive joints (RP-joints) that lock via tendon tension, enabling flexible pre-grasp configuration without extra actuation. A learning-based framework combines single-example dexterous grasp learning, kernel-density contact models, and kinaesthetic optimization to adapt grasp poses and RP-joint configurations, followed by planning for RP-joint reconfiguration. The approach achieves substantial performance gains, with up to $80\%$ success on IKEA objects and $87\%$ on the YCB dataset, validated over hundreds of grasps across diverse object sets. The results demonstrate that adding RP-joint reconfigurability reduces mechanical complexity while enhancing dexterity and robustness, with potential impact for prosthetics and industrial manipulation.
Abstract
We introduce a novel reconfigurable passive joint (RP-joint), which has been implemented and tested on an underactuated three-finger robotic gripper. RP-joint has no actuation, but instead it is lightweight and compact. It can be easily reconfigured by applying external forces and locked to perform complex dexterous manipulation tasks, but only after tension is applied to the connected tendon. Additionally, we present an approach that allows learning dexterous grasps from single examples with underactuated grippers and automatically configures the RP-joints for dexterous manipulation. This is enhanced by integrating kinaesthetic contact optimization, which improves grasp performance even further. The proposed RP-joint gripper and grasp planner have been tested on over 370 grasps executed on 42 IKEA objects and on the YCB object dataset, achieving grasping success rates of 80% and 87%, on IKEA and YCB, respectively.