Long-term aerial robot localization based on visual odometry and radio-based ranging

Visual odometry approaches have demonstrated to work very well for localization in GPS-denied areas, however they could quickly diverge depending on the environment. In order to achieve reliable long-term localization, it is a good practice to complement vision-based estimations with an additional source of information able to correct odometry drifts. This paper presents an approach that complements stereo visual odometry with distance measurements from several radio beacons located in an indoor scenario. The computational efficiency of the whole processing pipeline allows for real-time and bounded drift localization. Field experiments have been conducted using an indoor testbed with a very accurate tracking system in order to validate the algorithms.