Practical Vision-Based Walking Navigation for the Humanoid Robot NAO in the Maze-like Environment

Abstract

The motion features that accompanied by the alternate walking of the humanoid robot, which include body swing, foot slip, and fear of collision, are important factor limiting the indoor applications. In this manuscript, the humanoid robot NAO is taken as the platform, and the reliable pass through the maze-like indoor region is set as specific goal. Firstly, the specific intersection scene types are summarized, and the walking strategies for different intersection scenarios are discussed. Then, the general algorithm of path correction is developed. And the long straight walking, yaw correction, and L-shape intersection steering experiments are carried out. Furthermore, the corresponding optimization methods are discussed. The proposed strategy realizes a comprehensive technical framework for NAO to perform visual perception and stable walking in an unknown environment, which provide practical reference for efficient indoor navigation.