A Neural Network for Planning Hand Shapes in Human Prehension

Abstract

Quantifying human hand movement is a problem that interests both motor psychologists, in studying human behavior, and robot designers, in reproducing it. We attempt to capture the functionality of human prehensile movement using abstracted concepts such as virtual fingers and opposition space. We describe a neural network that maps object/task properties into a prehensile posture, relating the mapping to empirical evidence.