This report discusses the use of vision feedback for autonomous navigation by a mobile robot in indoor environments. In particular, we have discussed in detail the issues of camera calibration and how binocular and monocular vision may be utilized for self-location by the robot. A noteworthy feature of monocular vision is that the camera image is compared with a CAD model of the interior of the hallways using the PSEIKI reasoning system; this reasoning system allows the comparison to take place at different levels of geometric detail.

Date of this Version