We present an accurate metric localization
method using a simple artificial landmark for the navigation of
indoor mobile robots. The proposed landmark model is designed
to have a three-dimensional, multi-colored structure and the
projective distortion of the structure encodes the distance and
heading of the robot with respect to the landmark. Catadioptric
vision is adopted for the robust and easier acquisition of the
bearing measurements for the landmark. We propose a practical
EKF based self-localization method that uses a single artificial
landmark and runs in real time.