Next: Map Representation Up: Localization Previous: Model matching localization
In these approaches landmarks are used as references to compute the location of the robot. Landmarks can range from a set of sensor readings to artificial landmarks such as beacons or bar-codes or natural landmarks detected by vision systems. Because of its computational simplicity and also its close relationship with human navigational abilities, this approach is the most widely used, and it has been used with both grid-based and topological approaches.
This approach also suffers from the problem of ambiguity among landmarks that look similar. Again, the probabilistic approach can help solving this problem. Thrun [65] and Dissanayake et al [21], among others, use this approach together with grid-based maps, and Simmons and Koenig [60] and Kaelbling et al [33] combine it with the topological approach.