Flight Testing of Real-Time Model Predictive Flight Control for Unmanned Flexible Aircraft

Citation
EuroGNC conference, CEAS 2024
Contribution to Conference
EuroGNC conference, CEAS 2024
This paper describes the design, real-time implementation, and flight testing of a model predictive flight controller for the longitudinal and lateral-directional control of a slightly flexible 25kg unmanned aircraft. The controller is responsible for stabilization of the aircraft and tracking of reference commands. Reduced-order linear prediction models are derived from a high-fidelity nonlinear model of the flexible aircraft at different points in the flight envelope. To allow for the consideration of aeroelastic effects throughout the control design process, relevant structural modes are included in these models. From the set of prediction models, multiple linear model predictive controllers are designed off-line, applying the MPC formulation using a velocity-form model for offset-free tracking. During on-line operation, they are scheduled similar to classical gain-scheduling. The number of decision variables is effectively reduced by approximating the incremental control trajectory with Laguerre functions. An observer is designed to estimate non-measureable states. The predictive flight controller is finally validated in multiple flight test experiments with predefined tracking scenarios. The controller is shown to provide very good tracking performance even in the presence of wind and gust disturbances.