G. Dudek, K. Lonji We are developing a teleoperated mobile robot system that functions in a known indoor environment using vision and a geometric model of the environment. This geometric model(internal map) is used to construct a CAD model of the environment. An uncertainty estimated in the position/orientation of the robot during navigation is maintained using the camera images in conjunction with the internal map. This will be useful for vision based self-location, future motion planning, and for updating the internal map with new objects extracted from the stereo images. In particular we are interested on matching landmarks derived from the internal map with monocular images from the camera so that if during navigation simulation our sensor data(stereo images) experiences transmission delays or becomes too noisy, we can superpose artificially generated graphics onto a regular video picture. Once transmission contact is resumed the stereo data can once more be used for further navigation.