Stability on orientation motion of biped walking robot aiming at a target object

A definition of the visual walking proposed by authors is that the robot autonomously walks by making decision based on the image feature motion. One of the achievements is "visual tracking walk". In the past conference, authors proposed a hybrid control for "visual walking", in which a biped walking robot with a CCD camera mounted on the head follows a red target. Authors used median and the area of the target image in the image plane as the feature value. Therefore, the biped walking robot followed the translation of the target. However, the robot could not follow orientation of the target. In this paper, the four vertexes of a square surrounding the red target in the image plane are selected as image features, and the visual walking makes the rotation motion possible. Considering the practical application, a new estimated feature value is proposed including the effect of the video capturing processing time and the fluctuation of COM of the walking robot. The camera motion is calculated form the estimated image features by multiplying the inverse image Jacobian matrix. Consequently the walking trajectory reference of the robot foot is generated. The proposed time delay compensation or the disturbance rejection approach makes the visual walking stable in the practical situation. Simulation studies clarified several walking patterns, including translation motion and rotation motion of the target