Human position tracking for side by side walking mobile robot using foot positions

Abstract

This work proposes a method for tracking human position from side way. The method takes streaming of depth images from a Microsoft Kinect sensor as its input. The sensor is set up at the height of a human and tilted downward to cover the human's feet. The system tracks the position of the feet and uses the position along with the feet's direction to compute the human position and heading. The underlying idea relies on the fact that depth sensor can easily identify floor plane and the assumption that feet are located just above the floor plane. The method finds its use on side-by-side walking mobile robot. Real world examples are presented showing that the method gives fairly accurate tracking performance.