Learning of the 5-finger robot hand using deep learning for stable grasping

The human hand is one of the most complex organs of the human body, capable of performing skilled tasks. Manipulation, especially grasping is a critical ability for robots. However, grasping objects by a robot hand is a challenging issue. Many researchers have used deep learning and computer vision methods to solve this problem. This paper presents a humanoid 5-degree-of-freedom robot hand for grasping objects. The robotic hand is made using a 3D printer and 5 servo motors are used to move the fingers. In order to simplify the robotic hand, a tendon-based transmission system was chosen that allows the robot's fingers to flexion and extension. The purpose of this article is to use deep learning algorithm to grasping different objects semi-automatically. In this regard, a convolutional neural network structure is trained with more than 600 images. These images were collected by a camera mounted on the robot's hand. Then, the performance of this algorithm is tested on different objects in similar conditions. Finally, the robot hand is able of successfully grasping with 85% accuracy.