Posture stabilization of robot by visual feedback

In the future, with progress in robotics, robots will play an important role in the daily lives of human beings. The posture of a robot should be stable so that it does not fall. The posture of robots has been maintained stable mainly by using force and acceleration sensors. However, human beings maintain the stability of their postures on the basis of not only floor reaction force but also visual information. Therefore the use of visual information could be effective in maintaining the stability of the posture of a robot. The rotation invariant phase only correlation (RIPOC) method can be used to measure the rotation angle between two images. The accuracy of the RIPOC method is high in the case of 2D images, for example, in the case of fingerprint recognition and face recognition. However, it is difficult to accurately measure the rotation angle in 3D space by using this method. This paper describes a new extended RIPOC method that can be used to accurately measure the rotation angle in 3D space. By using visual feedback by the proposed method, the posture of robots can be stabilized. Experimental results that confirm the effectiveness of this method are provided in this paper.