Felix Renard

Wednesday, March 25, 2026, 9:30 - 10:30

Room 01-012, Georges-Köhler-Allee 102

Abstract: This thesis presents the design, implementation, and numerical evaluation of a real-time nonlinear model predictive control (NMPC) scheme for the Maverix flight system,a small fixed-wing unmanned aerial vehicle developed at RWTH Aachen University. The controller is coupled with an Extended Kalman Filter (EKF) for joint state and wind estimation, forming a closed-loop system that operates without full state measurement or prior knowledge of the wind conditions. The dynamic model of the aircraft is formulated as a 22-dimensional state space system comprising six-degree- of-freedom rigid body dynamics and a polynomial aerodynamic model approximated from computational fluid dynamics data. 

The NMPC controller is formulated as a nonlinear least-squares tracking problem and solved using the real-time iteration scheme provided by the open-source acados framework. Closed-loop simulations over a reference trajectory demonstrate accurate trajectory tracking in nominal conditions as well as under a constant wind disturbance. All state and control constraints are satisfied in both scenarios, with the EKF converging rapidly to the true states and wind. The simulations further reveal that the aerodynamic model validity bounds, particularly on the forward body velocity, limit the operational wind speed envelope.