In this paper we compare the use of a Kalman filter and a Robust State Observer for the localization and mapping of an underwater vehicle using range-only measurements between the vehicle and a set of beacons lying on the seafloor. As expected, we show that the Kalman filter performs great when we have a reasonably good prior information on the location of the vehicle and the beacons. Based on set-membership methods, the Robust State Observer demonstrates an outstanding capacity to provide consistent estimates (where the true solution is in the estimated confidence domain) in the presence of outliers at the cost of a very coarse precision. The source of this lack of precision will be discussed.