휴머노이드 로봇을 위한 시각 기반 계단 보행 패턴 생성 방법

Abstract

This paper proposes a method of a staircase walking pattern generation for humanoid robots based on stereo vision processing. A footprint generating method is suggested using the clamped cubic spline for generating a desired trajectory of the Center of Mass (CoM) and a desired trajectory of the Zero Moment Point(ZMP). Here, the conventional PZCSA in [6] is adopted for the walking pattern generation. However, since the PZCSA method makes use of a series approximation in order to remove unstable zero. it is inevitable to include the errors, if the order of series approximation is not enough to remove unstable zeros. As an alternative, we utilize the low-pass filter for removing the error. Also, the features of staircase are extracted using the RANSAC algorithm to obtain kinematic data based on stereo vision processing. Finally, the effectiveness of the suggested method is shown through the simulations.