In order to address the flexible requirements on load, flight speed, and maneuverability in actual flight conditions, a hand-launched fixed-wing unmanned aerial vehicle (UAV) is designed with high performance in portability, lightweight, and maneuverability. On this basis, a novel system-decomposition-based robust control method is proposed to achieve a satisfactory control performance for this designed UAV under internal/external uncertainties. First, considering the feature that the coupling terms are usually small in most flight conditions and may be regarded as disturbance, the complex flight system is decomposed into a group of simple subsystems with disturbance, namely, roll-yaw subsystem, pitch-altitude subsystem, and airspeed subsystem. Since the complexity of each subsystem is much smaller than the original system, this will reduce the difficulty of the controller design and complexity of the controller. Besides, to improve the tracking performance and robustness to disturbances, an extended state observer based robust control method is proposed for each subsystem. The stability and robustness of the proposed control method are effectively demonstrated by the theoretical analysis and the actual flight experiments on the designed UAV.