⚡ electrical engineering

Control Algorithms for Quadcopter Motion in Dynamic Positioning Mode

This paper derives a complete motion model for quadcopter dynamic positioning and proposes two control algorithms: one that generalizes previous results to varying yaw angles, and another that utilizes a simplified regulator tuning methodology.

Original authors: Stanislav Kim, Anton Pyrkin, Oleg Borisov

Published 2026-05-18

📖 4 min read☕ Coffee break read

CC BY 4.0

Original authors: Stanislav Kim, Anton Pyrkin, Oleg Borisov

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

Imagine a quadcopter (a drone with four propellers) as a very sensitive, high-speed dancer trying to hold a perfect pose in mid-air. The goal of this paper is to teach that dancer how to stay perfectly still at a specific spot, even if it needs to twist its body (change its "yaw" or spin direction) while doing so.

Here is a breakdown of the paper's ideas using simple analogies:

The Problem: The "Crude" Map vs. The Real World

The authors start by saying that most people try to control drones by using a "crude map." They simplify the complex physics of the drone into a straight line to make the math easier. But in the real world, the drone is a complex, non-linear machine. If you use a simple map for a complex journey, you might get there, but the ride will be bumpy and shaky.

Furthermore, previous attempts to fix this used "incomplete maps." They forgot to account for the drone spinning around (the yaw angle). It's like trying to drive a car while ignoring the steering wheel; you might move forward, but you can't control where you're pointing.

The Solution: Two New "Navigation Systems"

The authors built a complete, high-definition map of the drone's movement that includes every twist, turn, and spin. Based on this perfect map, they designed two different ways to control the drone.

Algorithm 1: The "Expert Navigator" (Generalizing the Old Way)

Think of this as upgrading an old GPS system.

How it works: It takes the best parts of previous research and adds the missing piece: the spinning (yaw) motion.
The Catch: To make this work, the "navigator" (the controller) has to be tuned very carefully. The authors had to prove mathematically that if you pick the right numbers, the drone will stabilize. It's like tuning a radio to a specific frequency; if you are slightly off, the signal is static. This method works, but setting it up is tricky and requires satisfying complex mathematical conditions.

Algorithm 2: The "Auto-Pilot with a Simple Manual" (The Simplified Approach)

This is the paper's main innovation. The authors wanted a control system that is just as stable but much easier to set up.

The Metaphor: Imagine the first algorithm is like a Formula 1 car where you have to manually adjust the suspension, tire pressure, and fuel mixture for every single turn. The second algorithm is like a modern car with a "Sport Mode" button. You press it, and the car automatically adjusts everything for you.
How it works: They created a "dynamic" control system. Instead of trying to force the drone to behave like a simple straight line, they treat the drone's motion as a chain of reactions.
The Magic Trick: They proved that by using a specific mathematical structure (called a "normal form"), they can tune the controller using standard, easy-to-find patterns (like a "Butterworth" filter, which is just a standard recipe for smooth movement).
The Benefit: You don't need to solve a complex puzzle to make it work. You just follow a simple recipe to pick your numbers, and the drone stays stable.

The "Safety Net" (Stability)

A major worry in drone control is: "What if the drone tilts too far and crashes?"
The authors showed that their new "simple manual" approach includes a built-in safety net. They proved that as long as the drone doesn't tilt too wildly (which is a natural physical limit), the control system will always bring it back to the center. It's like a self-correcting gyroscope that knows exactly how hard to push to keep the drone upright, even when the wind (or in this case, the math of the drone's own movement) tries to knock it over.

Summary

In short, the authors took a very difficult math problem (controlling a spinning drone in 3D space) and solved it in two ways:

The Upgrade: They fixed the old, incomplete methods to include spinning.
The Simplification: They invented a new method that is just as powerful but much easier for engineers to tune, removing the need for complex, guesswork-heavy adjustments.

They didn't invent a new drone; they invented a smarter, easier-to-use "brain" for the drone to keep it perfectly still in the air.

Technical Summary: Control Algorithms for Quadcopter Motion in Dynamic Positioning Mode

Problem Statement
The paper addresses the challenge of controlling a quadcopter for dynamic positioning at a specified point in space. The authors identify that existing approaches often rely on model linearization, which can be crude and fail to provide satisfactory transient performance due to model inaccuracies. Furthermore, previous works that utilize nonlinear motion models (e.g., [1, 2]) are criticized for employing incomplete models that neglect yaw angle dynamics and for proposing control algorithms with non-obvious tuning methodologies. In these prior works, regulator parameters must be selected to satisfy complex conditions derived from stability proofs, making practical implementation difficult.

Methodology
The authors derive a complete mathematical model of quadcopter motion that accounts for all linear and angular coordinates, including the dynamics of the yaw angle. This model is expressed in a nonstationary normal form, representing the system as a cascade of integrators with time-varying matrix coefficients. Based on this complete model, two distinct control algorithms are proposed:

State-Feedback Control with Generalized Yaw Dynamics:
- This algorithm generalizes the results of previous work [1] to the case of a varying yaw angle.
- It utilizes feedback linearization to decouple the system into altitude/yaw and horizontal coordinate loops.
- For the altitude and yaw loops, a PD controller with saturation is employed to ensure the system remains within a bounded control region, satisfying specific stability conditions.
- For the horizontal plane coordinates, the control law involves a complex state-feedback structure. The authors prove that asymptotic stability can be achieved despite the nonstationary nature of the system matrix (specifically the time-varying coefficient $\beta(t)$ ) by utilizing an extended version of the Dayawansa lemma. This involves a backstepping-based proof to determine stabilizing feedback coefficients.
Dynamic State-Feedback Control with Simplified Tuning:
- This algorithm introduces a dynamic extension where the primary control inputs ( $u_1, u_2$ ) are treated as outputs of integrators driven by new virtual inputs ( $v_1, v_2$ ).
- This transformation allows the aggregated quadcopter model to be represented in a strict normal form with a nonsingular input matrix.
- The resulting closed-loop system is a linear time-invariant system with a block-diagonal structure.
- Crucially, this approach simplifies the regulator tuning process. The feedback parameters ( $\gamma_1, \gamma_2, \gamma_3, \gamma_4$ ) can be chosen directly based on standard performance criteria (e.g., Butterworth or Newton characteristic polynomials) without needing to satisfy complex inequalities associated with nonstationary gains.

Key Contributions and Results

Complete Modeling: The paper provides a rigorous derivation of a complete quadcopter motion model that includes yaw dynamics, addressing a gap in previous literature.
Generalization: The first algorithm successfully extends existing robust control results to scenarios with varying yaw angles, ensuring asymptotic stability of the zero equilibrium for position and orientation errors.
Simplified Tuning: The second algorithm's primary contribution is the simplification of the regulator tuning methodology. By restructuring the control problem into a dynamic state-feedback form, the authors eliminate the need for complex, nonstationary gain calculations in the state observer synthesis. This makes the selection of controller parameters more straightforward and less dependent on proving specific stability inequalities for time-varying coefficients.
Stability Proofs: The paper includes formal proofs (Statement 1, Statement 2, and Statement 3) demonstrating that the proposed control laws ensure the asymptotic stability of the quadcopter at a desired point in space, even in the presence of the system's nonstationary characteristics.

Significance
The authors claim that their work resolves identified shortcomings in current quadcopter control literature. By providing a complete model and two control strategies, the study offers a more robust framework for dynamic positioning. Specifically, the second algorithm is highlighted for its practical advantage: it facilitates the design of robust control algorithms by removing the complexity associated with nonstationary gains in state observers, thereby significantly easing the parameter selection process for engineers and researchers. The work is supported by grants from the President of the Russian Federation and the Ministry of Science and Higher Education of the Russian Federation.