Artificial intelligence system for controlling the capture of a humanoid robot

Abstract

In this dissertation, a new approach to intelligent hand control was proposed, which can simulate the natural movement of a human hand. The implementation of such an intelligent hand can be used both in a humanoid and in personal robots. In this work, vision-based interaction methods (cameras) are used to track finger movements and obtain accurate and fast hand movements. Accuracy and efficiency play a key role in real-time motion-based applications. The Raspberry pi processor is used to control the entire installation to which the motion capture camera module is attached. To describe the hands in the video, the removal of the face, the definition of skin color is used. The mean shift algorithm is used to track finger movements. The relationship between the mechanical structure and the number of drives and degrees of freedom (DOF) is optimized to withstand the rigid dimensions and weight restrictions that are common applications of artificial hands for the consequences of humanoid robotics. The proposed arm structure has a structure similar to a natural arm, with articular fingers controlled by permanent arm motors (each finger has 4 degrees of freedom).