Mobile robot navigation using vision and odometry

Abstract

A cooperation strategy between odometry and a novel visual self-location method for indoor autonomous robot navigation is described. The strategy exploits the a-priori knowledge about the path that the vehicle is going to run, whether to model the odometer uncertainity in order to a-priori define the odometer re-calibration position or to verify the experimental uncertainity following the one modelled in order to update the re-calibration position. The visual self-location method capitalizes on the excellent angular resolution of CCD TV cameras in order to estimate the camera position and orientation and, therefore, the vehicle position.