UAV are becoming increasing interesting for many applications, also in the civil field. While UAV are usually expected to be in constant contact with basis stations, a critic aspect of their usability is related to the autonomy in case of an interruption of communications. This autonomy includes the ability of both a sense-andavoid algorithm as well as local pilot capabilities. In this paper we present the design of a suitable autopilot design based on a 12th order nonlinear model. Two waypoint tracking strategies are discussed and compared in simulation in lateral and longitudinal direction using the Aerosonde simulation model. One solution is based on model inversion while the second one relies on a very simple controller. Under windy conditions the tracking results of both controllers look similar which makes the simpler controller more attractive due to its simplicity. This controller has finally been extended to control a UAV realized with a Multiplex Twinstar II. Measurements show that the presented and implemented control law is appropriate for smooth flight maneuvers.