A linear-quadratic-Gaussian approach for automatic flight control of fixed-wing unmanned air vehicles

This paper presents the design and implementation of automatic flight controllers for a fixed-wing unmanned air vehicle (UAV) by using a linear-quadratic-Gaussian (LQG) control approach. The LQG design is able to retain the guaranteed closed-loop stability of the linear-quadratic regulator (LQR) while having incomplete state measurement. Instead of feeding back the actual states to form the control law, the estimated states provided by a separately designed optimal observer, i.e. the Kalman filter are used. The automatic flight controllers that include outer-loop controls are constructed based on two independent LQG regulators which govern the longitudinal and lateral dynamics of the UAV respectively. The resulting controllers are structurally simple and thus efficient enough to be easily realized with limited onboard computing resource. In this paper, the design of the LQG controllers is described while the navigation and guidance algorithm based on Global Positioning System (GPS) data is also outlined. In order to validate the performance of the automatic flight control system, a series of flight tests have been conducted. Significant results are presented and discussed in detail. Overall, the flight-test results show that it is highly feasible and effective to apply the computationally efficient LQG controllers on a fixed-wing UAV system with a relatively simple onboard system. On the other hand, a fully automatic 44km cross-sea flight demonstration was successfully conducted using the LQG-based flight controllers. Detailed description regarding the event and some significant flight data are given.