Application of intelligent robust adaptive flight control laws for the Cessna Citation X business jet

Abstract

The flight control systems design for commercial aircraft remains constrained by the reliance of classical and modern controllers on accurate modeling, extensive gain scheduling, and expert driven tuning. While adaptive and intelligent control techniques offer promising alternatives, their application to safety-critical aviation systems is limited by concerns related to robustness, numerical stability, and certification compatibility. This thesis addresses these challenges through the development and validation of a hybrid adaptive flight control architecture applied to the Cessna Citation X business jet. The proposed control method combines Proportional Integral–Derivative (PID) controllers for baseline stability with Recursive Least Squares–based Dynamic Inversion and Neural Network adaptation to compensate for modeling uncertainties, nonlinearities, and time-varying dynamics. Dynamic Inversion is achieved using online estimation of local aircraft dynamics from measured state and control increments, thus eliminating the need for explicit aerodynamic models or gain scheduling. Neural network adaptation further compensates for residual inversion errors and actuator nonlinearities, enabling dual online adaptation within a structured and interpretable control architecture. The methodology is validated using a high-fidelity nonlinear simulation model of the Cessna Citation X across 64 cruise flight conditions covering a wide range of altitudes and airspeeds. Inner-loop controllers for pitch rate, roll rate, and yaw stabilization, as well as outer-loop autopilots for vertical speed, altitude, and heading control, are designed and evaluated using a single controller configuration. Extensive robustness analyses are conducted under wind gusts, Dryden turbulence, actuator noise, and actuator loss-of-effectiveness. Numerical stabilization strategies, including covariance reinitialization and threshold-based adaptation, are introduced to ensure reliable online estimation and dynamic inversion. Performance evaluation demonstrates accurate signal tracking with steady-state errors below 2%, fast settling times for inner-loop dynamics, and robust disturbance rejection across the full flight envelope. Flying qualities are explicitly assessed using MIL-STD 1797A criteria, and Level 1 requirements are consistently satisfied for short-period, roll, and Dutch-roll dynamics modes. The Lyapunov-based stability analysis confirms boundedness and convergence of all adaptive control elements. Comparative results validation using a Level D research flight simulator data shows improved performance of the controller relative to the baseline onboard controller. This thesis demonstrates that hybrid adaptive control based on PID stabilization, RLS dynamic inversion, and neural network adaptation provides a scalable, robust, and certification compatible solution for envelope-wide flight control of commercial aircraft. The results establish a practical pathway toward the deployment of adaptive flight control systems in safety-critical aviation applications.

Date	19 Jun 2026
Original language	American English
Awarding Institution	École de technologie supérieure
Supervisor	Ruxandra Botez (Supervisor) & Georges Ghazi (Co-supervisor)