Assuming that landmarks, such as mountain peaks, in the environment have been identified and matched to a map, we have shown that it is possible to limit localization errors before they occur. A simple algorithm can be used to exploit the geometric properties of the landmarks in order to decrease errors in localization. The goal is to choose landmarks which will provide the best localization regardless of measurement error.
The assumption on knowledge about the landmarks is then reconsidered and it is shown that geometric properties can again be used to order or partially order landmarks, avoiding the time consuming task of using more subtle characteristics to identify individual mountain peaks.