Forem: MUSTAFA SERDAR SÖKMEN

When State Estimation Fails: How UAVs Actually Crash

MUSTAFA SERDAR SÖKMEN — Fri, 23 Jan 2026 15:04:19 +0000

When a UAV crashes, people often blame:

The pilot

The controller

The AI

But in reality, most crashes begin much earlier — and much quieter.

They start when the system loses confidence in its own state.

🧠 Control Rarely Fails First

Modern flight controllers are robust.
PID loops don’t suddenly “forget” how to stabilize.

What fails first is the input to those controllers:

Wrong attitude

Wrong velocity

Wrong position

Garbage in — crash out.

🌫️ How Estimation Slowly Breaks

State estimation rarely fails catastrophically.
It degrades.

Common patterns:

IMU bias slowly accumulates

GPS latency increases

Magnetometer gets disturbed

Vibrations leak into accelerometers

Each error is small.
Together, they move the system away from reality.

⚠️ The Most Dangerous Moment

The most dangerous phase is not aggressive flight.

It’s hover.

Why?

Low dynamics hide errors

Filters get overconfident

Biases grow silently

Then a sudden maneuver happens —
and the estimated state no longer matches physics.

🔄 Control Obeys the Wrong Reality

The controller is not “wrong”.
It is doing exactly what it was designed to do.

The problem:

It is stabilizing a state that does not exist.

From the controller’s perspective, the crash makes perfect sense.

🤖 Why AI Can Make This Worse

AI-based perception can:

Mask estimation errors

Delay detection

Add false confidence

An AI model may say:

“Everything looks normal.”

While the estimator is already drifting.

🧩 Real Crash Logs Tell the Same Story

If you read enough flight logs, you see patterns:

Sudden attitude jumps

Velocity spikes

Position corrections too late

Crashes are not surprises.
They are unnoticed divergences.

💭 Final Thought

UAVs don’t crash because control fails.

They crash because:

The system stops knowing where it is — and keeps flying anyway.

Understanding this difference is what separates pilots from engineers.

Learning vs. Knowing: Why UAVs Still Need State Estimation

MUSTAFA SERDAR SÖKMEN — Thu, 22 Jan 2026 05:48:34 +0000

Modern UAVs are getting smarter.

They can:

Detect objects

Classify targets

Understand terrain

Learn patterns from data

Yet, despite all this intelligence, a fundamental problem remains:

A UAV does not know where it is.
It estimates.

And that difference matters more than most people realize.

🧠 Learning Is Not Knowing

AI models are excellent at learning patterns.

They answer questions like:

What is this object?

Is this a road or a field?

Where should I go next?

But flight-critical questions are different:

What is my attitude right now?

How fast am I moving?

Am I drifting, or is the wind pushing me?

These are not perception problems.
They are state estimation problems.

⚙️ The Invisible Core of Every UAV

Inside every UAV, there is a continuous process trying to answer one thing:

“What is the most likely state of the system right now?”

This process:

Combines IMU, GPS, barometer, magnetometer

Filters noise and delay

Produces a best guess — not the truth

Kalman filters, EKFs, and complementary filters do not learn.
They infer.

🌫️ Reality Is Noisy and Delayed

Sensors lie:

IMUs drift

GPS lags

Barometers fluctuate

Magnetometers get disturbed

The real world is:

Noisy

Delayed

Incomplete

AI can see the world.
State estimation makes sense of it in real time.

🤖 Why AI Cannot Replace State Estimation

Could an AI model estimate state?

Yes — in theory.

But:

It lacks guarantees

It lacks explainability

It lacks predictable failure modes

A Kalman filter tells you:

“I am uncertain by ±x.”

An AI model usually tells you nothing — until it’s wrong.

🧩 A Healthy Architecture

Robust UAV systems separate responsibilities:

State Estimation:
Physics-based, deterministic, explainable

Control:
Fast, stable, safety-critical

AI:
Perception, prediction, assistance

AI learns.
Estimation knows (approximately).
Control survives.

💭 Final Thought

AI helps UAVs understand the world.

State estimation helps UAVs understand themselves.

And in flight,
self-awareness comes before intelligence.

AI Is Not the Pilot: Where Artificial Intelligence Actually Fits in UAVs

MUSTAFA SERDAR SÖKMEN — Wed, 21 Jan 2026 17:45:51 +0000

Artificial Intelligence is everywhere in modern engineering discussions.
Especially in UAVs.

Vision-based navigation.
Object detection.
Autonomous decision-making.
Swarm intelligence.

But there is a dangerous misconception hiding underneath all this excitement:

AI does not fly a drone.
And it never should.

🧠 What Actually Keeps a UAV in the Air

A drone stays airborne because of:

State estimation

Control loops

Real-time deterministic systems

Flight controllers operate at:

Hundreds or thousands of Hertz

With strict timing guarantees

Under hard real-time constraints

AI models:

Are probabilistic

Have variable latency

Can fail silently

That alone disqualifies them from low-level flight control.

⚠️ Why AI Is a Terrible Pilot

Imagine a neural network responsible for:

Attitude stabilization

Motor mixing

Failsafe recovery

Now add:

Sensor noise

EMI

Voltage drops

Edge-case scenarios

A PID controller fails predictably.
An AI model fails creatively.

In aviation, creativity is a liability.

🤖 Where AI Actually Belongs

AI shines at high-level cognition, not reflexes.

Good use cases:

Target detection and classification

Terrain understanding

Path planning

Mission-level decision making

Anomaly detection

In other words:

AI decides what to do — not how to stay alive.

🧩 The Control–AI Boundary

A healthy UAV architecture looks like this:

Flight Controller:
Stability, control, safety (deterministic)

Autonomy Stack:
State machines, logic, rule-based systems

AI Modules:
Perception, prediction, assistance

AI suggests.
Autonomy decides.
Control executes.

Reverse this hierarchy, and you lose reliability.

🚀 The Future Is Hybrid, Not AI-Only

The most robust UAVs will not be:

Fully rule-based

Fully AI-driven

They will be hybrid systems:

Classical control for survival

AI for understanding the world

The smartest drones will still trust physics more than data.

💭 Final Thought

AI is not the pilot.
It’s the advisor.

And in aviation,
the advisor is never allowed to touch the controls.

The Myth of Stability: Why a Drone Is Always About to Fall

MUSTAFA SERDAR SÖKMEN — Tue, 20 Jan 2026 08:40:27 +0000

At first glance, a hovering drone looks calm.
Steady. Balanced. Almost effortless.

But this is one of the biggest illusions in UAV engineering.

A drone is never truly stable.
It is constantly falling — just very well corrected.

🌀 Stability Is an Illusion

Unlike an airplane cruising forward, a multirotor has no natural stability.

No wings generating passive lift

No restoring aerodynamic moments

No equilibrium without active control

If you turn off the flight controller:

The drone doesn’t “slowly drift”

It falls immediately

Hovering is not a state of rest.
Hovering is an ongoing emergency handled in real time.

⚙️ The Control Loop That Keeps It Alive

What we call “stable flight” is actually the result of:

IMU measurements (accelerometers + gyros)

State estimation

Control algorithms (PID, LQR, etc.)

Motor commands updated hundreds of times per second

At 400–1000 Hz, the flight controller:

Detects a tiny deviation

Predicts what happens next

Applies corrective thrust

Repeats — forever

Miss a few cycles, and gravity wins.

🧠 Why Balance Is the Wrong Mental Model

Many beginners think:

“If the center of mass is right, the drone will be stable.”

That logic works for static objects.
A drone is not a static system.

It is:

Underactuated

Nonlinear

Highly sensitive to delay and noise

Stability does not come from balance.
It comes from continuous decision-making.

🌬️ The World Is Actively Trying to Kill Your Drone

Wind gusts.
Motor mismatches.
Vibrations.
Battery voltage drops.
Sensor noise.

The environment is hostile.

A “stable” drone is simply one whose flight controller is:

Fast enough

Smart enough

Tuned well enough

to fight reality every millisecond.

🧩 Same Frame, Different Reality

Take two identical drones:

Same frame

Same motors

Same propellers

Change only:

Control gains

Filtering

Sensor fusion logic

One will feel:

Smooth and confident

The other:

Nervous, twitchy, unpredictable

The difference isn’t mechanical.
It’s how close to falling they are allowed to get.

🚁 Pilots Don’t Fly — They Intervene

This is why experienced pilots say:

“You don’t fly a drone. You prevent it from crashing.”

The flight controller does 99% of the work.
The pilot (or autonomy logic) only nudges the system away from disaster.

💭 Final Thought

A drone in hover is not stable.
It is dynamically surviving.

And the flight controller?
That’s not a stabilizer.

It’s a system that continuously answers one question:

“How do I avoid falling — again — right now?”

Mechanics vs. Electronics: Why the “Soul” of the Drone Lies in the Flight Controller

MUSTAFA SERDAR SÖKMEN — Mon, 19 Jan 2026 09:01:24 +0000

When people look at a drone, the first things they notice are usually the frame, motors, and propellers.
Carbon fiber arms, thrust-to-weight ratios, sleek aerodynamics… all impressive.

But here’s a controversial take:

The real soul of a drone is not mechanical — it’s electronic.

More specifically, it lives inside the flight controller.

🔩 Mechanics: The Body

Mechanics give the drone its physical existence.

The frame defines strength and geometry

Motors and propellers generate thrust

Aerodynamics affect efficiency and endurance

Without solid mechanics, a drone can’t fly — that’s true.
But mechanics alone don’t decide how it flies.

A perfectly designed airframe without intelligence is just a falling object with spinning motors.

🧠 Electronics: The Mind (and Soul)

The flight controller is where a drone becomes alive.

It:

Interprets sensor data (IMU, GPS, barometer, magnetometer)

Makes thousands of decisions every second

Maintains stability in chaos (wind, vibration, payload changes)

Translates human intent into precise motion

Two drones with identical frames and motors can fly completely differently
just because of:

Control algorithms

Sensor fusion quality

Tuning philosophy

Software architecture

That difference is not mechanical — it’s behavioral.

⚙️ Why the Flight Controller Is the “Soul”

Think about it this way:

Mechanics define limits

Electronics define character

A drone’s:

Smoothness

Aggressiveness

Precision

Fault tolerance

Autonomy level

are all shaped by the flight controller.

That’s why:

A racing drone feels “angry”

A cinematic drone feels “calm”

An autonomous UAV feels “confident”

Same physics. Different souls.

🚀 From Hobby to Aerospace

As drones evolve toward:

Swarm operations

Autonomous missions

Electronic warfare–resilient systems

AI-assisted flight

The importance of mechanics decreases relatively,
while avionics and software dominate performance.

In aerospace-grade UAVs, the flight controller isn’t just a board —
it’s a decision-making system.

💭 Final Thought

Mechanical engineers build the body.
Electrical and avionics engineers give it a soul.

And the flight controller?
That’s where the drone decides who it wants to be.

Simulating Physics: Using MATLAB/Simulink for UAV Dynamics

MUSTAFA SERDAR SÖKMEN — Sun, 18 Jan 2026 07:18:21 +0000

Before a UAV ever takes flight, it must "fly" in a digital environment. Using MATLAB/Simulink, we can model complex flight dynamics and test control algorithms like LQR (Linear Quadratic Regulator) or PID without risking expensive hardware.

The real power of Simulink lies in its ability to handle multi-domain physics. By defining mass, inertia, and aerodynamic coefficients, we create a mathematical twin of the aircraft. In my projects, such as the VTOL mothership developed with BTU-ALFA, simulation allowed us to refine the transition logic between hover and forward flight.

The ultimate goal is "Model-in-the-Loop" (MIL) and "Hardware-in-the-Loop" (HIL) testing. When the simulated response matches our real-world flight logs, we gain the confidence to push the boundaries of autonomous flight.

Defending the Skies: Anti-Jamming and GPS Spoofing in UAVs

MUSTAFA SERDAR SÖKMEN — Sat, 17 Jan 2026 15:50:05 +0000

In the modern era of electronic warfare, Unmanned Aerial Vehicles (UAVs) face two major invisible threats: Jamming and Spoofing.

Jamming is the act of drowning out GPS signals with high-powered noise, causing the drone to lose its position and often trigger a "failsafe" mode.

Spoofing is more dangerous; it involves sending fake GPS coordinates to the drone, tricking it into thinking it is somewhere else, which can lead to the aircraft being hijacked or crashed.

To counter these, we develop systems like NavGuard. By integrating multi-constellation GNSS receivers, inertial navigation systems (INS), and advanced filtering algorithms like Extended Kalman Filters (EKF), we can ensure mission continuity even in contested environments. Protecting the data link is no longer an option—it is a necessity for the future of avionics.

Pilot vs. Engineer: How Flying a UAV Changes the Way I Write Code

MUSTAFA SERDAR SÖKMEN — Fri, 16 Jan 2026 09:05:00 +0000

**As an Electrical-Electronics Engineering student and a licensed UAV pilot, I’ve spent countless hours behind two different screens: the IDE where I write flight control logic, and the Ground Control Station (GCS) where I monitor that logic in the sky.

Early in my journey, I thought that if the simulation in Gazebo looked perfect, the flight would be perfect. I was wrong. Transitioning from the desk to the field taught me that "perfect code" can fail if it doesn't account for the "chaos" of reality.

Inertia is More Than a Variable In a simulator, inertia is just a number in a physics engine. In the field, it’s the reason your VTOL drifts an extra meter during a transition because of a sudden gust of wind. Being a pilot taught me to write "anticipatory" code. Now, when I implement an LQR or PID controller like: u(t)=Kpe(t)+Ki∫e(t)dt+Kddtde(t)

I don't just see gain values (Kp,Ki,Kd); I visualize how the aircraft will physically struggle or snap into position.

The "What If" Mindset (Fail-safes) A programmer thinks about "if-else" statements. A pilot thinks about "what if the link breaks right now?" This perspective shifted my focus toward robust Fail-safe mechanisms. If my NavGuard layer detects a GPS spoofing attempt, the code must decide in milliseconds: Trust the IMU or initiate an emergency landing? Pilots don't want "clever" code; they want "predictable" code.
Respecting the Environment Electronic Warfare (EW), signal interference, and sensor noise are invisible enemies. My interest in avionics grew from seeing these "invisible" factors disrupt my flight missions. Now, I write code that doesn't just process data—it questions it.

Conclusion Being a pilot made me a better engineer because it gave me empathy for the operator. If you want to write better firmware, step away from the IDE and go to the flight field. Seeing your code fight the wind will teach you more than any textbook ever could.**

Why Avionics? The Invisible Brain of Modern Aircraft

MUSTAFA SERDAR SÖKMEN — Thu, 15 Jan 2026 12:33:02 +0000

When we look at a drone or a high-tech aircraft, we are usually captivated by its physical presence: the sleek wings, the powerful motors, or the aerodynamic frame. But as an Electrical-Electronics Engineering student who has spent the last two years developing UAVs, I’ve realized that the true magic isn't in the parts you see, but in the "invisible brain" that orchestrates it all—the Avionics.

Early in my journey at BTÜ-ALFA, I learned that while mechanics define the physical limits of an aircraft, avionics define its intelligence. A wing provides lift, but it is the complex control loops (like PID or LQR) that decide how to handle a sudden gust of wind or how to execute a precision landing.

My passion for this field stems from a simple realization: Designing the flight controller is like writing the soul of the machine. Whether it’s optimizing power management in a Hybrid VTOL or ensuring navigation in GPS-denied environments, avionics is where mathematics meets the sky.

In this blog, I won’t just talk about code or hardware; I’ll share the journey of making "machines that think." Because in modern aviation, the brain is just as important as the wings.

I’d love to hear your thoughts! Do you think the soul of a drone lies in its aerodynamics or its control algorithms? If you have questions about my transition from being a pilot to an avionics enthusiast, feel free to ask in the comments!