Robot localization: comparable performance of EKF and UKF in some interesting indoor settings

Two typical indoor settings are considered, one based on laser range finder readings and the other one on the distances gathered by the robot from a set of artificial landmarks (e.g. RFID tags). It is shown, by comparing the performance of the Extended and of the Unscented Kalman filters, that only the first setting could require the definition of more specific non linear filtering algorithms, not needed so much in the other case. In the second setting, the sensitivity of the estimation error on the number of landmarks has been numerically investigated. Based on these results, a rule of thumb for the filtering approach and for the deployment of the landmarks in the environment is proposed.