Neurofuzzy grasp control of a robotic hand

Abstract

A best first grasp for a robotic hand with pressure sensing is determined by assigning fuzzy membership values to aspects of candidate grasps attempted with a modified genetic backpropagation neural network controller. Fuzzy logic control is employed to guide finger adjustments as the grasped object begins to trip or slip. Extensions to three dimensions, as well as controller optimization with neural networks, are explored.< >