UWB localization system for partially GPS-denied robotic applications

Most common robotic outdoor navigation systems typically rely upon global positioning system, never the less there are several scenarios in which the GPS alone is not enough. The presence of high buildings, trees or other obstacles can cause a substantial deterioration of the estimation of the position of the robot due to blockage of line-of-sight and multipath effects. So, dealing with partially GPS-denied environments makes it necessary to develop ad hoc solutions to aid navigation. This work moves the first steps towards the development of a localization system based on drones and ultra-wide band technologies. The idea is to use three or four drones hovering above the working area of the rover where they can be localized using GPS. Moreover, both the UAVs and the ground vehicle are equipped with ultra-wide band sensors in order to precisely measure the relative distance between all the agents. With this information, it is possible to compute the position of the UGV using trilateration algorithms. The first part of this thesis is a theoretical analysis of the trilateration problem. Two algorithms are compared, taking into account some simple geometries for the positioning of the anchor nodes. The first goal is to evaluate which are the points in space that minimize the positioning error. The second issue is to understand how the number of anchors affect the precision of the position estimation. In the second part of the work, several measurements are performed using two different sets of commercial sensors: Decawave TREK1000 and Pozyx. The precision of both systems is evaluated through ranging and positioning measurements. The computed parameters are then used in simulation in combination with information on the positioning of a drone hovering in fixed position. In this way it is finally possible to compute the expected error in the localization of the ground vehicle.