Intelligent Control for Fixed-Wing eVTOL Aircraft | Dr. Xichen Shi, Caltech

Added:

Deterministic Robust Control and the Aerospace Certification Challenge

While intelligent and learning-based control architectures offer high adaptability, a major counterpoint in aerospace engineering favors deterministic, classical robust control methods due to stringent safety and certification standards. Critics argue that learning-based systems, particularly those utilizing neural networks or online adaptation, act as "black boxes" whose behavior is difficult to predict under edge-case scenarios. Regulatory bodies like the FAA and EASA require strict Verification and Validation (V&V) pathways. Classical control methods (such as gain scheduling and robust H-infinity control) rely on rigorous, static mathematical proofs of stability. Proponents of this traditional approach argue that the complexity, lack of interpretability, and potential unpredictability of adaptive, learning-based controllers pose unacceptable risks for passenger-carrying eVTOLs. Consequently, they advocate for simpler, highly verifiable deterministic models as the only viable path for commercial flight certification.

Fundamental Flight Mechanics and Aerodynamics: Understanding the physics of both fixed-wing flight and rotary-wing hover, including the aerodynamically complex transition phase of eVTOL aircraft.
State-Space Control Theory: Familiarity with linear and non-linear control systems, state-feedback, and multi-input multi-output (MIMO) system representations.
Adaptive Control Theory: Core concepts of adaptive control architectures (such as Model Reference Adaptive Control) designed to handle system uncertainties and dynamic changes.
Basic Machine Learning for Control: An introductory understanding of how neural networks or reinforcement learning can be integrated into dynamic system controllers.
Robust and Nonlinear Control Design: Exploring advanced techniques like Sliding Mode Control or Lyapunov-based safety guarantees for systems with high non-linearities.
Hardware-in-the-Loop (HIL) Testing: Translating theoretical control architectures into embedded software for real-time simulation and physical flight-testing on eVTOL prototypes.
Aviation Certification for AI/ML: Investigating the regulatory challenges and verification methods (e.g., FAA/EASA standards) for deploying learning-based control algorithms in commercial aviation.
Fault-Tolerant Control (FTC): Designing systems that can dynamically adapt to actuator failures, structural damage, or extreme meteorological hazards during flight.
256 views7likes58:05@galcitmedia800Original Release: 2021-01-28

This video presents a comprehensive intelligent control framework for fixed-wing eVTOL (electric vertical takeoff and landing) aircraft, addressing the challenges of urban air mobility applications. The framework integrates physics-based modeling with data-driven approaches, including a unified control architecture that uses force and control allocation to manage transitions between VTOL and fixed-wing flight modes. Key innovations include adaptive control methods that compensate for environmental disturbances using novel airflow sensors, fault-tolerant rotor configurations optimized for failure scenarios, and neural network-based dynamics approximation for enhanced control intelligence. The research also addresses practical implementation challenges such as actuation delay compensation through predictive control methods, enabling robust performance in real-world digital control systems.