Featureless robot navigation using optical flow

Abstract

In this paper, we aim to develop an algorithm for featureless navigation of an autonomous robot using optical flow observed through a vision system mounted on a mobile robot. Furthermore, we show that this optical flow observed from a mobile robot for navigation of it is also used in the construction of environmental maps. Our algorithm first detects an area corresponding to the dominant plane from a sequence of images employing optical flow and, second, backprojects the dominant plane to the environment for the construction of the environmental map. Moreover, our algorithm estimates ego-motion of the mobile robot using projected optical flow. The planar area detection by our method does not require any physical assumptions concerning the robot motion of any camera calibration. Therefore, our algorithm allows obstacles to be detected without using any feature in the environment and any heuristics rule. We show the results of the construction of an environmental map using the mobile robot.