Vision-based ZMP detection and stability evaluation in image plane for biped robot

This paper describes a visual ZMP detection and its mapping method to the image plane for evaluating dynamic equilibrium of walking motion in biped robot system. The zero moment point (ZMP) is popular index for evaluating the dynamically balanced gait, and this idea contributes to practical design of walking motion and stability controller for reliable walking. However, the reaction force detection by force or pressure sensors at foot soles requires much implementation costs. In this paper, the appropriate flexibility around ankle joints are given, and the ZMP is detected by only vision sensor mounted on the robot. Furthermore, the proposed idea enables the definition of suitable ZMP region in camera image plane. This means the evaluation of balanced gait is possible only from visual information. The fundamental performance of the proposed method is evaluated by several experimental results.