A Data-Driven Based Blind Path Recognizing and Tracking Control Strategy for a Guided Quadruped Robot

This study addresses the motion control issues of a guiding quadruped robot and proposes a novel data-driven approach for blind path recognition and tracking. The proposed approach consists of two parts, including blind path recognition and robot tracking control. In terms of visual principles, the guiding quadruped robot identifies blind paths based on color using the Otsu threshold segmentation algorithm, combined with Canny edge detection and Hough transform for precise positioning of the blind path. Next, this study presents a model-free adaptive control method utilizing a dynamic linearization technique to follow the identified blind path. The controller is configured with a pseudo partial derivative (PPD) parameter. The proposed method enables accurate recognition and tracking of blind paths by the robot, thereby achieving precise following of blind paths by the guiding quadruped robot. The control aspect of the proposed method relies on a model-free strategy, eliminating the requirement to establish an exact mathematical model of the guiding robot during the controller design process. Instead, the controller can be achieved solely using input and output data of the robot. The proposed method was applied to a real guiding quadruped robot experimental platform, and the experimental results demonstrated a 96.24% accuracy in blind path recognition and an average deviation angle of 0.17 degree in tracking control. The effectiveness of this approach has been validated.