Topological localization using appearance-based recognition

Localization is a fundamental problem of autonomous mobile robots. Localization is the determination of the position and orientation of a robot. Most localization systems are made up of several sensors and a map of the environment. Sophisticated localization systems can solve both the global location problem and the kid-napped robot problem. Global localization is the process of placing the robot into an unknown location within the map, and the robot should be able to locate itself within a relatively short period of time. The kidnapped robot problem is similar to global localization, as it is a test of how well the robot is able to recover after becoming lost. The robot is “teleported” to a new location, and the robot should again be able to determine its new location within a relatively short amount of time. CReSIS (Center for Remote Sensing of Ice Sheets) is developing autonomous robots in an effort to measure ice sheets characteristics in Greenland and Antarctica. These robots currently rely on differential GPS for localization and navigation. In order to survive for long periods of time in these environments, however, the robots need to be able to return to campsites in order to refuel and unload the data that has been acquired. In order to perform this task effectively and safely, a more elaborate system is required. A localization system that can recognize the dfferent locations of the campsites is the beginning of this process. The approach is to use a single camera for use in multiple types of large-scale environments, indoors and outdoors using a topological map. The system described uses an appearance-based approach for recognizing the different locations. The appearance-based methods attempt to recognize the appearance of a scene rather than specific objects. Several different types of features are tested including histograms, eigenimages, and Hu Moments. Using these simple features, the system is able to determine its location within the map at 95% accuracy.