Plenárias
Structure meets Data – System Representation for Trustworthy and Efficient Control
Sandra Hirche
Professor
Technische Universität München (TUM)
07/10/2026
Resumo: Data-driven models are increasingly employed to complement or replace first-principles models in modern control systems. Yet purely data-driven approaches often lack the guarantees on stability, safety, and generalization that control applications demand, and they typically require large amounts of data to perform well. This talk argues that the key to overcoming these limitations lies not in choosing between physics-based and data-driven modeling, but in systematically combining them: embedding structural knowledge — from physical laws to symmetries — directly into the learning process. We will present approaches to structured dynamic system representation using Gaussian processes. We show how exploiting structure improves data efficiency, yields calibrated uncertainty quantification, and enables formal guarantees on robust closed-loop stability and safety — key ingredients for trustworthy control. We illustrate the benefits of these structured representations on robotic and human-machine systems.
Bio: Sandra Hirche holds the TUM Liesel Beckmann Distinguished Professorship
and heads the Chair of Information-oriented Control in the Department of
Electrical and Computer Engineering at Technical University of Munich (TUM),
Germany (since 2013). She received the diploma engineer degree in
Aeronautical and Aerospace Engineering in 2002 from the Technical
University Berlin, Germany, and the Doctor of Engineering degree in Electrical
and Computer Engineering in 2005 from the Technische Universität München,
Munich, Germany. From 2005-2007 she has been a PostDoc Fellow of the
Japanese Society for the Promotion of Science at the Fujita Laboratory at
Tokyo Institute of Technology, Japan. Prior to her present appointment she
has been an Associate Professor at TUM.
Her main research interests include learning, cooperative, and networked
control with applications in human-robot interaction, multi-robot systems, and
general robotics. She has published more than 200 papers in international
journals, books and refereed conferences. She has received multiple awards
such as the Rohde & Schwarz Award for her PhD thesis, the IFAC World
Congress Best Poster Award in 2005 and – together with students –
Best Paper Awards of IEEE Worldhaptics and IFAC Conference of
Manoeuvring and Control of Marine Craft in 2009 and the Outstanding Student
Paper Award of the IEEE Conference on Decision and Control 2018. In 2013
she has been awarded with an ERC Starting Grant on the “Control based on
Human Models” and in 2019 with the ERC Consolidator Grant on “Safe data-
driven control for human-centric systems”. Sandra Hirche is Fellow of the
IEEE and IFAC Fellow.
Uso de Informação Auxiliar na Identificação de Sistemas Não Lineares
Luis Antonio Aguirre
Professor
Universidade Federal de Minas Gerais (UFMG)
07/10/2026
Plenária Bottura-Castrucci
Resumo: A identificação de sistemas nasceu no contexto de sistemas representados como “caixas pretas”. Nesse cenário o objetivo é construir modelos exclusivamente a partir de dados dinâmicos, normalmente obtidos durante testes específicos. Quando não se tem outras informações sobre o sistema além desse conjunto de dados, a identificação caixa preta é uma ótima opção na construção de modelos. Um leve desvio dessas técnicas, conhecido como identificação semifísica, é composto de técnicas que utilizam conhecimento do sistema para determinar a estrutura matemática do modelo e os dados são empregados para estimar parâmetros. Nesta palestra desejamos dar um passo além da modelagem semifísica. Imaginamos que informação adicional do sistema não determina por completo a estrutura do modelo, sendo que em alguns casos ela está na forma de um conjunto de dados em estado estacionário. Alguns tipos de informação auxiliar úteis em identificação de sistemas são revistos, bem como procedimentos para levar a informação adicional em conta durante a construção de modelos. Esses métodos constituem parte do que se conhece por identificação “caixa cinza”. Dois estudos de caso são discutidos.
Bio: Luis A. Aguirre recebeu os títulos de Bacharel e Mestre em Engenharia Elétrica pela Universidade Federal de Minas Gerais em 1987 e 1990, respectivamente. Obteve o grau de doutor em 1994 pela Universidade de Sheffield, Inglaterra por trabalho em Identificação de Sistemas Não Lineares. Desde 2010 é professor Titular da Universidade Federal de Minas Gerais, no Departamento de Engenharia Eletrônica, onde coordenou o Programa de Pós-Graduação em Engenharia Elétrica por cinco anos. Serviu como membro e coordenador da Câmara de Tecnologia em Engenharia (FAPEMIG) e no Comitê Assessor do CNPq. Foi cordenador de área adjunto junto à CAPES. Atualmente é Editor Chefe do periódico Nonlinear Science (Elsevier) desde sua fundação em 2024 e Editor Regional do periódico Modelling, Identification and Control (Inderscience) desde sua fundação em 2006. No passado foi Editor Chefe do periódico Controle & Automação (SBA) e Editor Associado do periódico International Journal of Systems Science (Taylor & Francis). É bolsista PQ-1A desde 2012. Foi vice-presidente da SBA em dois mandatos.
Stability and Signal Generator Agnostic Moment Matching
Alessandro Astolfi
Professor, Computer, Electrical and Mathematical Science and Engineering
Division
King Abdullah University of Science and Technology (KAUST)
08/10/2026
Abstract: Moment matching is a model reduction technique that allows the
construction of reduced-order models preserving specific moments. These moments are associated with the steady-state output response of the system to be reduced, interconnected in an open-loop fashion with a signal generator. Model matching thus relies on strong stability properties and on the availability of a model of the signal generator.
To relax these requirements, in the first part of the talk we introduce a data-driven
procedure for computing reduced-order models from input-output data generated by an agnostic signal generator. The moments are directly identified from the output of the system to be reduced, and the resulting reduced-order models achieve asymptotic matching.
To relax the stability requirements, in the second part of the talk we revisit the notion of moment matching by introducing the concept of closed-loop interpolation. This notion relies upon the construction of a novel class of signal generators, which are
feedback-interconnected with the plant to be reduced, and allows the definition of
moments and models for unstable systems. The existence of a family of models that
parameterise all, possibly unstable, systems achieving moment matching in a
closed-loop fashion is also presented.
Bio: Alessandro Astolfi (IFAC Fellow, FIEEE, Academia Europaea, FIET,
ITATEC, ACA Fellow, EUCA Fellow) received the Laurea in Electronic Engineering
from the University of Rome La Sapienza, Italy, in 1991; the M.Sc. degree in
Information Theory and the Ph.D. degree with Medal of Honor with a thesis on
discontinuous stabilization of holonomic systems from ETH Zurich, Switzerland, in
1995; and the Ph.D. degree for his work on nonlinear robust control from the
University of Rome “La Sapienza” in 1996.
From 1992 to 1996 he was a Research Associate at ETH Zurich. From 1996 to
2026, he was with the Electrical and Electronic Engineering Department, Imperial
College London. In particular, he was Professor of Nonlinear Control Theory; College
Consul for the Faculty of Engineering and Business School (2022–2025); and Head
of the Control and Power Group (2010–2022). From 1998 to 2003 he was an
Associate Professor with the Department of Electronics and Information, Politecnico
di Milano. From 2005 to 2025 he was a Professor with the Dipartimento di Ingegneria
Civile e Ingegneria Informatica, University of Rome Tor Vergata, Rome. Since
February 2026, he is a Professor of Applied Mathematics and Computational
Science at KAUST, Saudi Arabia.
His research interests include mathematical control theory and control applications,
with special emphasis on the problems of discontinuous stabilization, robust and
adaptive control, observer design, optimal control, game theory, and model
reduction.
Dr. Astolfi was the recipient of the IEEE CSS A. Ruberti Young Researcher Prize
(2007); the IEEE RAS Googol Best New Application Paper Award (2009); the IEEE
CSS George S. Axelby Outstanding Paper Award (2012); the Automatica Best Paper
Award (2017); and the IEEE Transactions on Control Systems Technology
Outstanding Paper Award (2023). He is a “Distinguished Member” of the IEEE CSS
and a recipient of the IFAC Outstanding Service Award (2017). He is the recipient of
the Institute of Measurement and Control Sir Harold Hartley Medal for “outstanding
contributions to the technology of measurement and control.”
He was Editor-in-Chief of the IEEE Transactions on Automatic Control (2018–2025)
and Chair of the IEEE CSS Conference Editorial Board (2010–2017). He is Vice
Chair of the IFAC Technical Board (2020–2026); and a Member of the IEEE
Thesaurus Editorial Board (2025–) and of the IEEE PRAC (2026–).
Differentiable Control and GPU Parallelism: A New Optimization Framework for Robotics
Fabio Ramos
Professor
The University of Sydney
08/10/2026
Abstract: Massively parallel hardware has unlocked unprecedented computational power in robotics, shifting our analytical focus from finding a single motion to generating a diverse and resilient set of solutions. But how do we mathematically define or measure diversity in motion planning for robotics manipulators? In this talk, I will provide a probabilistic interpretation for diversity and show how tools designed for machine learning—such as differentiable programming languages and parallel computation on GPUs—can be utilized for large-scale probabilistic inference in planning and control. I will describe a powerful nonparametric inference method that leverages both differentiability and parallelism to provide posterior approximations for real-time Model Predictive Control (MPC), and task and motion planning. Finally, I will formulate diversity in trajectory planning through a new mathematical tool—signature transforms—that leads to highly resilient, next-generation motion planning methods capable of handling the uncertainties of real-world robotics manipulation.
Bio: Fabio Ramos is a Professor of Robotics and Machine Learning at The University of Sydney in Australia, and a Principal Research Scientist at NVIDIA, USA. He is an alumnus of the Polytechnic School, University of São Paulo (USP), where he received his BSc and MSc in Mechatronics Engineering, before completing his PhD at the University of Sydney. His research centers on statistical machine learning for large-scale Bayesian inference and decision-making in robotics and heavy industry—notably leading the design of the world’s first autonomous open-pit iron mine. He has authored over 200 peer-reviewed publications and holds multiple Best Paper Awards from premier venues like RSS, IROS, and ECML.
Controle de Sistemas Comutados: Uma Perspectiva Baseada em Grafos
Thiago Alves Lima
Professor
Instituto Tecnológico de Aeronáutica (ITA)
09/10/2026
Plenária Jovem Pesquisador
Resumo: Esta palestra apresenta resultados recentes sobre o controle de sistemas comutados por meio de ferramentas baseadas em grafos. Para a classe de sistemas comutados com modos lineares e em tempo discreto, são discutidas a construção de funções de Lyapunov e a síntese de controladores a partir de famílias de grafos com estruturas particulares. Será mostrado que, para uma classe específica de grafos, obtêm-se, pela primeira vez na literatura, condições convexas, expressas por LMIs, que são assintoticamente necessárias e suficientes para a estabilização sob comutação arbitrária. Em seguida, utilizando teoria de linguagens formais e autômatos, será apresentada uma extensão dessas ferramentas para cenários em que a estabilização sob comutação arbitrária não é possível. A palestra será concluída com uma breve discussão sobre perspectivas e desafios atuais em sistemas comutados.
Bio: Thiago Alves Lima é professor do Instituto Tecnológico de Aeronáutica (ITA) no novo campus de Fortaleza, Ceará. Realizou graduação, mestrado e doutorado em Engenharia Elétrica pela Universidade Federal do Ceará (UFC). Concluiu o doutorado em 2021, tendo realizado período sanduíche de 15 meses no laboratório LAAS (CNRS), em Toulouse, França. Em seguida, realizou pós-doutorados na Université Catholique de Louvain (ICTEAM), Bélgica, e na Université de Lorraine (CRAN-CNRS), França. Entre 2023 e 2025, foi Maître de Conférences (Professor Associado) na CentraleSupélec, junto ao Laboratoire des Signaux et Systèmes (L2S), França. Posteriormente, foi bolsista do programa Conhecimento Brasil, voltado à repatriação de pesquisadores brasileiros, antes de assumir o atual cargo de professor no ITA. Seus interesses de pesquisa atuais concentram-se no controle de sistemas híbridos e não lineares, com ênfase no desenvolvimento de métodos baseados na teoria de Lyapunov, métodos formais e aplicações de ferramentas grafo-teóricas à análise de estabilidade e síntese de controladores via LMIs.
