The Tactile SoftHand: An Anthropomorphic Robot Hand with a Biomimetic Sense of Touch

点击翻译按钮获取中文摘要

This project is a continuation of my first year project in which rudimentary closed loop grasp control of the IIT/Pisa SoftHand was implemented using only feedback from a fingertip sized version of the BRL TacTip on the smallest digit.The main objectives of this project will be to create a robust tactile based controller for the SoftHand (using multiple TacTip sensors on different digits) that can accurately perform relatively complex in-space and in-hand dexterous manipulation tasks despite the hand's under-actuation and the non-rigid nature of the digit joints. This will be achieved primarily from feedback of multiple TacTip sensors on several digits of the hand as well as secondary measurements from other sensors, with the tactile hand mounted on a UR5 industrial robotic arm.Activities will include: developing the software of the controller, implementing further sensors into the hand and applying appropriate methods for processing their outputs together, path planning with the UR5 and planning grasp experiments, investigating higher degrees of actuation.Milestones will include: successful integration with the UR5, implementation of framework for using multiple TacTips and execution of dexterous manipulation tasks, potentially with higher degrees of actuation.If successful, this project will bridge the gap between the hardware complexity of fully-actuated anthropomorphic hands and limited manipulation capabilities of under-actuated hands.