Optimal trajectory tracking control design

Abstract

Aircraft lifespan can be extended by upgrading and modernizing the electrical subsystems and instruments. However, this often introduces increased power demands and heat generation that the aircraft was not originally designed for and can result in a reduced flight performance and increased wear. Netherlands Aerospace Centre (NLR) and the Netherlands Ministry of Defence have set up a project to investigate power and thermal management of different aircraft platforms and to optimise the operational capabilities of the F-16. As part of this project, flight tests with the F-16 fighter jet performing various challenging aerobatic maneuvers were conducted and also replicated in a simulation environment by the pilot to validate the simulator accuracy. In order to evaluate the effect of different flight configurations on the power and thermal loads, it is necessary to track the same trajectories numerous times while changing the aircraft settings. Having a pilot in the loop introduces undesired variability to the results and is infeasible. Instead, a controller is required that is able to use the recorded reference trajectories and closely reproduce these flights in a simulated environment.

In this thesis, a publicly available F-16 model was used and extended with a feedback linearization controller to generate an analogous set of aerobatic reference flight trajectories. To track the reference trajectories two controllers were developed. A nonlinear model predictive controller and a nonlinear model predictive controller combined with a feedback linearization controller in the form of incremental nonlinear dynamic inversion. Adding feedback linearization reduced computation time and improved tracking of engine dynamics but made the controller dependent on reference actuator angles while the nonlinear model predictive controller alone is slower, but is able to track accurately with and without thrust and actuator angle reference signals. Tracking accuracy was tested on a set of well known aerobatic maneuvers such as Barrel rolls, loops and Half Cuban Eights. The NMPC controller was able to track the aircraft position in these maneuvers with a mean error of 0.44 ft (σ = 0.39 ft) while the NMPC-INDI controller achieved a mean position tracking error of 0.20 ft (σ = 0.10 ft). Further tests included mismatches in initial fuel weight and position and finally both controllers were validated by tracking a reference generated by NLR F-16 simulator. Here both controllers were able to adjust the thrust level to counter the effects of the speed brakes that were present for the trajectory generation but not for tracking. The NMPC was able to track the aircraft position with a mean position error of 20.01 ft (σ = 13.74 ft) while the NMPC-INDI achieved a mean position error of 8.70 ft (σ = 2.38 ft). These tests showed that both controllers are able to handle significant differences in the aircraft models and still keep the mean position tracking errors within one wingspan length (30 ft) which was the size of the reference tunnel for the pilot to track.