⚡ electrical engineering

Control of Flight

This paper presents the mathematical fundamentals, state-of-the-art techniques, and technical challenges in controlling atmospheric flight vehicles, with a specific focus on reduced-order modeling, simulation, and the practical implementation of guidance systems that guarantee stability, robustness, and performance.

Original authors: Eugene Lavretsky

Published 2026-06-02

📖 6 min read🧠 Deep dive

CC BY 4.0

Original authors: Eugene Lavretsky

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Based on the presentation by Dr. Eugene Lavretsky from Boeing, here is an explanation of the "Control of Flight" concepts in simple, everyday language.

The Big Picture: Flying in a Storm of Unknowns

Imagine you are driving a car, but the road is constantly changing shape, the engine is sometimes unpredictable, and your speedometer is occasionally lying to you. Now, imagine you have to drive that car at 600 miles per hour. That is what flying an aircraft is like.

The core problem this paper addresses is how to keep a plane stable and safe when you don't have a perfect map of the world.

Traditional flight control systems are like a driver who memorized a specific route perfectly. If the road changes even a little (a new pothole, a sudden wind), the driver might crash. Dr. Lavretsky's work focuses on building a "smart driver" that can handle unknown unknowns—things we didn't even know we needed to worry about, like weird aerodynamic forces at high angles or broken sensors.

The Three Main Tools

The presentation introduces three main "tools" or methods to solve this problem.

1. OBLTR: The "Ghost Pilot" and the "Shadow"

The Concept:
Imagine you are trying to teach a robot to drive. You give it a "Reference Model"—a perfect, ideal version of how the car should behave. This is your "Ghost Pilot."

However, the real car (the "Vehicle") is heavy, has loose parts, and the wind is blowing. The real car will never perfectly match the Ghost Pilot.

The Solution (OBLTR):
Dr. Lavretsky developed a system called Observer-Based Loop Transfer Recovery (OBLTR).

The Observer: This is a "Shadow" that watches the real car and the Ghost Pilot simultaneously. It calculates the difference (the error) between what the car is doing and what the Ghost Pilot says it should be doing.
The Adaptive Augmentation: This is the "Smart Driver" that steps in. When the Shadow sees a difference, it instantly adjusts the steering and gas pedal to force the real car to match the Ghost Pilot, even if the car is broken or the wind is crazy.
The Result: The system recovers the stability of a perfect model, even when the real world is messy. It's like having a co-pilot who knows exactly how much to push the controls to cancel out any surprise bumps.

2. CBF: The "Soft Rubber Band" vs. The "Hard Wall"

The Concept:
Every plane has limits. You can't pull too many Gs, or the wings might snap. You can't turn too sharply, or you'll stall.

Old Way (Hard Saturation): Imagine a car hitting a brick wall. If you turn the wheel too far, the steering wheel just stops moving abruptly. This is "hard saturation." It's jarring and can make the car spin out of control because the sudden stop creates a shockwave of instability.
The New Way (Control Barrier Functions - CBF): Imagine the steering wheel is attached to a soft, strong rubber band. As you turn the wheel toward the limit, the rubber band gets tighter and pulls back gently. It doesn't stop you instantly; it guides you back smoothly.

The Solution:
The paper describes using Control Barrier Functions (CBF) to create these "rubber bands" for flight controls.

Instead of hitting a hard limit and stopping, the system calculates a "soft limit."
It uses math to gently nudge the controls back before they hit the danger zone.
Why it matters: This keeps the plane stable even when the pilot or computer is pushing the limits. It prevents the "shock" of hitting a hard wall, ensuring the plane stays in the "safe zone" without jerking the controls.

3. Unsteady Aerodynamics: The "Wake" Effect

The Concept:
When a plane flies, it pushes air down to create lift. But that air doesn't just disappear; it swirls and creates a "wake" (like the wake behind a boat).

The Problem: If the plane flies through its own wake or a gust of wind, the air hitting the wings changes instantly. This is "Unsteady Aerodynamics." It's like trying to run on a treadmill that suddenly changes speed every second.
The Solution: The paper proposes modeling this "air mass" as a separate, moving part of the system. Instead of just looking at the plane, the control system looks at the plane + the air swirling around it.
The Analogy: Think of it as a dancer (the plane) and their partner (the air). If the partner stumbles, the dancer needs to know immediately to adjust their steps, rather than just reacting after they fall. By modeling the air's movement, the computer can predict the "stumble" before it happens.

How It All Fits Together

The presentation outlines a workflow that looks like this:

Simplify the Model: Start with a simple, clean math model of the plane (like a toy car).
Design the Controller: Build the "Ghost Pilot" (OBLTR) to make the toy car fly perfectly.
Add the Safety Net: Attach the "Rubber Bands" (CBF) so that if the toy car gets too close to a wall, it gently bounces back instead of crashing.
Add the Real World: Throw in the "Wind and Swirls" (Unsteady Aerodynamics) and broken sensors.
The Magic: The "Shadow" (Observer) sees the difference between the toy car and the real plane, and the "Smart Driver" (Adaptive Control) instantly fixes it.

The Bottom Line

Dr. Lavretsky's work is about making flight control robust. It's about admitting that we can't know everything about the air, the plane, or the sensors. Instead of trying to build a perfect map, we build a system that can adapt in real-time to whatever surprises come its way, keeping the plane stable and the passengers safe, even when the math gets messy.

The paper claims these methods have been tested on various Boeing projects (like the Phantom Eye and Scan Eagle) and are ready to be used on real aircraft to handle the "unknown unknowns" of flight.

Technical Summary: Control of Flight

Problem Statement

Flight control systems for modern aerial platforms must operate in real-time under "unknown unknowns," including incorrect assumptions about system dynamics, corrupted sensors, degraded controls, and environmental disturbances. Traditional model-based design relies on simplified linear models (e.g., gain-scheduled LQR) which, while guaranteeing stability for the model, often struggle with significant uncertainties such as:

Uncertain Aerodynamics: Particularly at high angles of attack, ground effects, and flexible body dynamics.
Actuator and Propulsion Effects: Including induced propulsion system effects, thrust vectoring, and actuator dynamics.
Operational Constraints: The need to enforce strict limits on control inputs and regulated outputs (e.g., angle of attack, structural loads) without resorting to "hard" saturation (sat-functions) that can degrade stability margins or cause windup.
Unsteady Aerodynamics: The tightly coupled dynamical interaction between aircraft dynamics and airmass vorticity, which manifests as gust-like effects.

The core challenge is to design control architectures that maintain stability, performance, and robustness while reducing dependency on highly accurate models and ensuring safety through formal constraint enforcement.

Methodology

The presentation outlines a comprehensive framework for flight control design, analysis, and implementation, focusing on three primary technical pillars:

1. Observer-Based Loop Transfer Recovery with Adaptive Augmentation (OBLTR)

OBLTR is a robust and adaptive flight control architecture designed to handle uncertain systems.

Baseline Design: Utilizes a state observer with Loop Transfer Recovery (LTR) to recover optimal stability margins (LQR margins) at the plant input. This involves "squaring-up" tall systems (where outputs > inputs) by adding fictitious inputs to ensure stable transmission zeros, overcoming minimum-phase restrictions.
Adaptive Augmentation: A direct adaptive control layer augments the baseline. It utilizes a closed-loop reference model concept to minimize transients during adaptation. The adaptive law is Lyapunov-based, ensuring stability while compensating for uncertainties in aerodynamics, flexible modes, and control failures.
Key Mechanism: The observer serves as a reference model for the adaptive controller. As the parameter $v$ (a weighting factor in the Algebraic Riccati Equation) approaches zero, the system recovers the robustness properties of the state-feedback design, effectively bridging robust and adaptive control.

2. Control Barrier Functions (CBF) for Operational Constraints

To address the need for enforcing operational limits (e.g., max angle of attack, control surface deflection) without the instability risks of hard saturation, the paper proposes a CBF-based design.

Formulation: The problem is formulated as finding a "min-norm" optimal control augmentation policy that enforces soft limits on outputs via feedback.
Analytic Solution: For Linear Time-Invariant (LTI) systems with linear constraints, the CBF solution is derived analytically using Karush-Kuhn-Tucker (KKT) conditions. This results in a continuous, piece-wise linear state feedback augmentation.
Advantages:
- No Online Optimization: Unlike Model Predictive Control (MPC), this approach does not require solving a Quadratic Program (QP) in real-time; the solution is pre-computed and implemented as a static feedback law.
- Anti-Windup: The method naturally manages anti-windup for MIMO-coupled integrators in PI servo-controllers.
- Stability Guarantees: A "CBF-Ability Criterion" is defined to guarantee closed-loop stability and uniform boundedness, independent of the baseline controller, provided the open-loop data and CBF parameters satisfy specific conditions.
Implementation: The CBF augmentation acts as a continuous static feedback that cancels (subtracts) the baseline controller near prescribed limits, similar to dynamic inversion, ensuring the system remains within the safe set defined by Nagumo's Theorem (forward invariance).

3. Unsteady Aerodynamics Modeling

The paper addresses the coupling between aircraft dynamics and airmass vorticity.

Physics: Based on Newton's Third Law, lift generation creates vorticity in the airmass, which in turn exerts a force back on the aircraft.
Modeling Approach: This interaction is modeled as a tightly coupled dynamical system where airmass vorticity dynamics act as "gust-like" effects on the aircraft. The presentation advocates for reduced-order modeling of these unsteady aerodynamic effects to be integrated into control applications, moving beyond static aerodynamic assumptions.

Key Contributions

OBLTR Architecture: The development and flight-proven transition of OBLTR, which combines robust observer-based control with adaptive augmentation. This technology has been transitioned to various platforms (e.g., Phantom Eye, Vulture, Scan Eagle, 777X Autoflight) to handle flexible body dynamics and uncertain aerodynamics.
Analytic CBF Augmentation: The derivation of an analytic, continuous piece-wise linear state feedback solution for enforcing operational constraints. This method avoids the computational burden of online optimization while providing formal stability guarantees and superior performance compared to hard saturation or projection operators.
Squaring-Up Technique: The application of "squaring-up" modifications to handle tall systems in observer design, relaxing restrictive minimum-phase assumptions and allowing the recovery of LQR margins for systems with non-minimum phase individual transfer functions.
Integration of Robustness and Adaptation: Demonstrating how to use control theory to extend robustness properties to achieve mission assurance, reducing the reliance on extensive wind tunnel testing or perfect models.

Results and Validation

The paper presents the technology as "Flight Proven" and "Verified, Validated."

Flight Testing: The OBLTR technology has been successfully flight-tested on multiple platforms, including the Dominator (2013), Pivot UAS (2014), JDAM 2kER (2015), Scan Eagle (2016-2021), and the 777X Autoflight system (2020).
Simulation Comparisons:
- CBF vs. Projection: Simulation data on an unstable scalar system demonstrates that CBF augmentation enforces soft limits via feedback more effectively than a rectangular projection operator, maintaining stability and tracking performance without the chattering or instability associated with hard limits.
- Transient Minimization: The closed-loop reference model concept in OBLTR is shown to minimize transients during adaptive augmentation.
Robustness: The methodology claims to recover optimal stability margins at the input and provides increased robustness at high frequencies compared to standard state feedback.

Significance and Claims

The paper positions these methodologies as essential for advancing the state-of-the-art in flight control for aerial platforms facing "unknown unknowns."

Mission Assurance: The primary significance lies in using control theory to guarantee stability and performance in the presence of significant uncertainties (aerodynamic, structural, and environmental) without requiring perfect models.
Safety and Certification: The CBF approach offers a formal, mathematically rigorous method for enforcing safety constraints (limits) that is compatible with standard flight control design practices (e.g., gain/phase margin analysis) and certification processes.
Efficiency: By providing analytic solutions for constraint handling (CBF) and robust adaptive augmentation (OBLTR), the approach reduces the dependency on "more testing" or "more wind tunnel" data, shifting the burden to rigorous mathematical design and verification.
Scalability: The technologies are presented as applicable to a wide range of systems, from small UAS to large commercial airliners (777X) and hypersonic vehicles, addressing challenges from flexible dynamics to high-angle-of-attack aerodynamics.

In summary, the work claims to provide a unified, mathematically grounded framework for designing flight control systems that are simultaneously robust, adaptive, and safe, capable of operating effectively in complex, uncertain environments where traditional linear control methods may fall short.