Implementation of Biped Robot with Smart Vision and Gait Controller

Abstract

In this study, the deep learning technology is applied to recognize various objects and to reduce the computation burden of the biped robot. The YOLOV4 neural network was adopted to recognize the obstacles in the first part of this research. Because of YOLOV4’s computational power, no need to connect to a server computer, the recognizing accuracy and fast computational characteristics, it satisfies the basic demand of smart biped robot. In order to achieve higher performance of obstacle avoidance, the second part of this research is to add the depth camera D435i which combines with the YOLOV4 neural network, the proposed architecture can more accurately measure the width and depth of the obstacles ahead. The third part of this research is to design a set of smart lightweight controller to reach above mentioned obstacle recognition ability. The fourth part is the mechanical design and gait control of biped robot. Some simulation and experimental results are provided in this paper. Furthermore, a biped robot prototype is implemented. We hope the implemented biped robot can not only replace our human beings to carry out the high risk jobs, but also can assist the transporting system.