Humanoid robot control system utilizing cost-oriented automation (COA) and edge detection

Abstract

Humanoid robots are machines designed to resemble the human body. They emulate specific aspects of human physiology, cognition, and social behavior to facilitate perception, processing, and action. One of the world’s most notable humanoid robot competitions is the HuroCup organized by The Federation of International Sports Association (FIRA). Although the race hosts various categories, this study only considers the obstacle run. This study has multiple objectives, including enhancing the ability of the humanoid robot’s vision system to detect objects using the Edge Detection algorithm. Additionally, the study aims to integrate the Cost-Oriented Automation (COA) and Edge Detection algorithms to enable the robot to detect and avoid obstacle objects. The COA algorithm serves a role in the robot’s body structure, while the Edge Detection algorithm detects objects through the Canny operator’s edge detection capabilities within the robot’s visual range. The Canny operator functions to reduce edge ambiguity for improved object detection. The results of the test indicate that in images with dark light intensity conditions in the HSV (Value Channel) color space, the average detection accuracy of the system reaches 71.43%. The detection accuracy increases to 82.86% in images with bright light intensity conditions. However, in images with dark light intensity conditions in the RGB (Blue Channel) color space, the detection accuracy is 50%, and it increases to 61.42% in images with bright light intensity conditions. The data confirms that using HSV color space images (Value Channel) provides better detection accuracy results than using RGB color space images (Blue Channel). However, the accuracy of robot movement remains a challenge that requires consideration. The implementation of the Cost-Oriented Automation (COA) and Edge Detection algorithms was able to successfully detect all box-shaped objects; however, the robot’s movements were inaccurate in avoiding obstacles. This was due to the robot’s unstable balance, and there were a few servos in its legs that stopped themselves, resulting in undesired movement. Therefore, the implementation of the Cost-Oriented Automation (COA) algorithm to the robot frame is suboptimal. Further refinement is necessary to improve the accuracy of robot movements. This requires replacing several faulty components, including servos and robot frames, to enhance the system’s overall performance, especially robot movements.