Dynamic map building for autonomous mobile robot

Presents an algorithm for autonomous map building and maintenance for a mobile robot. With each geometric target in the map the authors associate a validation measure to represent the belief in the validity of a target, in addition to the usual covariance matrix to represent spatial uncertainty. At each position update cycle, predicted features are generated for each target in the map and compared to features actually observed. Successful matches to targets with high validation measure are used for localization. Unpredicted observations are used to initialize target tracks for new environment features, while unobserved predictions result in a target's validation measure being decreased. They describe experimental results obtained with the algorithm that demonstrate successful map-building using real sonar data.>