Indoor localization using visual information and passive landmarks

This thesis proposes a localization algorithm for Automatically Guided Vehicles (AGVs) based on a vision system and simple passive markers. The pose is estimated using trilateration and triangulation techniques. Then the results are combined with heterogeneous data provided by odometry using an Extended Kalman Filter. The tests have shown that even with a non fully optimized algorithm, a precision of 0.2m can be reached, confirming the validity of this technology