Online Detection of Map Discrepancies in Landmark-Based Robotic Localization

Many robotic navigation tasks assume that a static map of the environment is sufficient for all future navigation tasks. This paper describes a method to detect the case where some features in the environment have changed with respect to the original map. The proposed method runs a localization filter and a mapping filter in parallel and finds discrepancies based on differences between the two map representations. Then, the map is repaired online to reflect the current environment.