Neural Network based Control Method Implemented on Ambidextrous Robot Hand

Abstract

Human hands can precisely perform a wide range of tasks. This paper investigates key performance differences when conventional robotic hand controllers are combined with Neural Networks (NN). Tests are performed on a novel 3D printed multi-finger ambidextrous robot hand. The ambidextrous hand is actuated using pneumatic artificial muscles (PAMs) and can bend its fingers both left and right, offering full ambidextrous functionality. Force sensors are placed on the fingertips. In our control method, the grasping trajectory of each finger combines its data with that of the neighboring fingers to obtain accurate results.