Obstacle crossing with stereo vision for a quadruped robot

Abstract

Autonomous locomotion in unstructured environments is a great challenge for quadruped robots, especially when there is little knowledge about the environments. Based on stereo vision, this paper proposes a locomotion control system for a quadruped robot that is able to perform adaptive locomotion in unknown environments cluttered with many obstacles. To reduce the negative influence of poor light, surface reflection, and textureless conditions, combined with the sum of absolute difference (SAD) method, an efficient post-processing algorithm is investigated to fulfill disparity estimation. According to the calculated terrain maps, a novel gait planning algorithm, which contains different gaits generation and gait transition, is designed to cross over terrains with enough stability and considerable locomotion speed. Experiments on the real robot of traversing complicated surface with stereo vision demonstrate the effectiveness of our system. Video footage can be seen at: http://irdc.ia.ac.cn/robot/.