Publications

IFAC World Congress 2026 · Paper TuA16.4

Internal-Model-Based Design of Disturbance Observers for a Class of Linear Systems with Modeled Disturbances

Tersoo Samuel Awai, Youngjun Joo, Hongkeun Kim

Korea University of Technology and Education; Sookmyung Women’s University. Proceedings of the IFAC World Congress, 2026.

Abstract

This paper addresses the design of disturbance observers for a class of uncertain linear plants affected by external disturbances, where the disturbance is generated by a linear model whose eigenvalues all lie in the closed right-half complex plane. The proposed method asymptotically rejects the modeled disturbance’s effect on the closed-loop system, where conventional disturbance observers only attenuate it approximately by implicitly embedding an internal model of the disturbance into one of the observer’s low-pass filters. Unlike existing internal-model-based designs, which require solving certain linear equations and so restrict disturbances to time polynomials and sinusoids with distinct frequencies, this method imposes no such restriction. The closed-loop system is shown to be robustly stable, with supporting simulations.

Robust controller synthesis Robust LMIs Uncertain systems

MSc Thesis · KoreaTech · 2026

Design and Stability Analysis of Internal-Model-Based Disturbance Observers for a Class of Linear Systems with Modeled Disturbances

Tersoo Samuel Awai

MSc thesis, Dept. of Mechatronics Engineering (Future Convergence Engineering), Graduate School, Korea University of Technology and Education (KoreaTech), Feb 2026. Advisor: Hongkeun Kim.

Abstract

A disturbance observer (DOB) estimates and compensates for unknown disturbances acting on a system to recover nominal performance. Conventional DOBs typically only attenuate, rather than completely reject, disturbances, so recent research has focused on achieving exact rejection of modeled disturbances. Existing internal-model-based DOB methods often embed the disturbance model through the solution of Vandermonde-type matching equations. These approaches impose restrictive algebraic conditions, particularly the non-singularity of Vandermonde matrices that limit applicability to disturbances with distinct polynomial or sinusoidal frequencies, and so cannot accommodate repeated eigenvalues, coupled modes, or general exosystems whose minimal polynomials contain multiplicities.

This thesis develops an internal-model-based disturbance observer (IM-DOB) that achieves exact rejection for a broader class of disturbances generated by arbitrary linear exosystems whose eigenvalues lie in the closed right-half complex plane. Unlike Vandermonde-based constructions, the proposed method embeds the internal model directly through a polynomial-shaping approach derived from the known minimal polynomial of the disturbance generator. This structure bypasses all solvability constraints, admits repeated eigenvalues naturally, and guarantees realizability for any admissible disturbance model, significantly generalizing existing designs.

Using singular perturbation theory and Lyapunov analysis, the stability of the resulting fast–slow closed-loop dynamics is rigorously established, proving that nominal performance is recovered under bounded plant uncertainties. The thesis further details the construction of the predictor and observer filters, the internal-model embedding mechanism, and the boundary-layer dynamics. MATLAB–Simulink simulations validate the design, demonstrating asymptotic disturbance rejection and robust tracking despite parameter variation and divergent disturbance growth.

초록 (국문)

외란 관측기(Disturbance Observer, DOB)는 시스템에 작용하는 미지의 외란(disturbance)을 추정하고 이를 보상하여 기준 성능을 회복하도록 설계된다. 그러나 기존 DOB는 일반적으로 외란을 완전히 제거하기보다는 감쇠하는 수준에 머무르며, Vandermonde 방정식의 해 존재성과 같은 제약적인 대수 조건에 의존하기 때문에 서로 다른 주파수를 갖는 다항식 또는 사인파 형태의 외란에만 적용 가능한 한계가 있다.

본 논문에서는 폐우반평면(closed right-half plane)에 고윳값을 갖는 임의의 선형 외인성시스템(exosystem)에 의해 생성되는 보다 일반적인 외란에 대해 이를 정확히 제거할 수 있는 내부모델 기반 외란 관측기(IM-DOB)를 제안한다. 제안된 방법은 외인성시스템의 최소다항식(minimal polynomial)으로부터 유도된 행렬 방정식 접근법을 사용하여 내부모델을 저대역 통과 필터에 직접 포함시키므로, 기존의 Vandermonde 방정식 풀이가 필요 없고 모든 허용 가능한 외란 모델에 대해 설계 가능하다. 이 방식은 반복 고유값, 결합 모드, 일반적인 외부시스템 등 기존 기법이 처리할 수 없었던 경우까지 자연스럽게 포함한다.

또한 특이섭동이론(singular perturbation theory)과 Lyapunov 해석을 통해 제안된 폐루프 시스템의 빠른-느린 동역학(fast-slow dynamics)이 안정적임을 엄밀히 증명하였으며, 이를 통해 모델 불확실성이 존재하더라도 기준 성능이 회복됨을 보였다. 본 논문에서는 저대역 통과 필터의 구성, 내부모델 포함 기법, 경계층(boundary-layer) 동역학 등 설계 절차를 상세히 제시하고, MATLAB-Simulink 시뮬레이션을 통해 외란의 점근적 제거와 강인한 추종 성능을 검증하였다.

Disturbance observers Linear systems Robust stability Singular perturbation
In progress

Further work from my doctoral research on decentralized multi-agent cooperation and safe control under unreliable communication is in preparation. This page is updated as new work is published.