Tracking for humanoid robot based on Kinect

Abstract

According to the high requirement in environment of tracking for humanoid robot based on vision, a recognition algorithm and control method is present using depth information extraction. The depth information is gathered by a Kinect camera from Microsoft, and an accelerated algorithm is proposed to improve the real-time requirements of image recognition. The obstacles information and robot status are gathered and an artificial potential field method based on safe region is planned to control the tracking. Finally, experiments are achieved on a real humanoid robot Nao.