Optimization is increasingly pervasive in modern technology, driven by major advances in computation, storage, and communication. These advances have made it possible to run optimization algorithms online, integrate them into control like in dynamic programming [1] and model predictive control [2], and process large-scale data in both centralized and distributed settings [3,4]. At the same time, these new paradigms pose challenges that remain far from fully resolved. In online optimization, performance guarantees often depend on feasibility and stability properties of the underlying algorithms [5, 6], while modeling errors and disturbances can degrade long-term behavior. In distributed optimization, key difficulties include privacy preservation, adaptation to time-varying environments, and resilience against malicious agents injecting faulty information [7–9]. Moreover, modern datasets are often large, heterogeneous, and governed by complex nonlinear relationships that simple parametric models may fail to capture [10]. This motivates the use of multilayer neural networks with nonlinear activation functions, which in turn leads to highly nonlinear optimization problems [11]. These challenges arise in strategic domains such as resource management, robotics and automation, power and energy systems, and distributed machine learning. They call for solid theoretical foundations, both in the formulation of meaningful frameworks and in the establishment of rigorous guarantees.
Several research directions tackle these challenges by combining optimization with system theoretic tools. For instance, algorithms can be modeled as dynamical systems, enabling the study of stability, robustness, and performance through control-theoretic methods [12–16]. Online and feedback optimization [17, 18] incorporate measurements into closed-loop schemes, leading to questions of sensitivity, model mismatch, and disturbance rejection. Distributed optimization adopts a multi-agent perspective, with applications in cooperative robotics [19], cyber-physical networks [20], and power and energy systems [21]. In parallel, constrained and safety-critical control focuses on feasibility, constraint satisfaction, robustness, and performance guarantees under uncertainty [22].
[1] Dimitri P. Bertsekas. Dynamic Programming and Optimal Control. Athena Scientific, 1995.
[2] James B. Rawlings and David Q. Mayne. Model Predictive Control: Theory and Design. Nob Hill Publishing, 2009.
[3] Stephen Boyd and Lieven Vandenberghe. Convex Optimization. Cambridge University Press, 2004.
[4] Yurii Nesterov. Introductory Lectures on Convex Optimization. Springer, 2004.
[5] Martin Zinkevich. Online convex programming and generalized infinitesimal gradient ascent. Proceedings of ICML, 2003.
[6] Elad Hazan. Introduction to Online Convex Optimization. Foundations and Trends in Optimization, 2016.
[7] John N. Tsitsiklis, Dimitri P. Bertsekas, and Michael Athans. Distributed asynchronous deterministic and stochastic gradient optimization algorithms. IEEE Transactions on Automatic Control, 1986.
[8] Reza Olfati-Saber, J. Alex Fax, and Richard M. Murray. Consensus and cooperation in networked multi-agent systems. Proceedings of the IEEE, 2007.
[9] Angelia Nedic and Asuman Ozdaglar. Distributed subgradient methods for multi-agent optimization. IEEE Transactions on Automatic Control, 2009.
[10] Kurt Hornik, Maxwell Stinchcombe, and Halbert White. Multilayer feedforward networks are universal approximators. Neural Networks, 1989.
[11] Yann LeCun, Yoshua Bengio, and Geoffrey Hinton. Deep learning. Nature, 2015.
[12] Gianluca Bianchin and Bryan Van Scoy. The internal model principle of time-varying optimization. IEEE Transactions on Automatic Control, PP(99):1–16, 2026.
[13] Michelangelo Bin, Ivano Notarnicola, Lorenzo Marconi, and Giuseppe Notarstefano. A system theoretical perspective to gradient-tracking algorithms for distributed quadratic optimization. In Proceedings of the 58th Conference on Decision and Control (CDC), pages 2994–2999, Nice, France, 2019.
[14] Umberto Casti, Nicola Bastianello, Ruggero Carli, and Sandro Zampieri. A control theoretical approach to online constrained optimization. Automatica, 176:112107, 2025.
[15] Vito Cerone, Sophie M. Fosson, Simone Pirrera, and Diego Regruto. A new framework for constrained optimization via feedback control of Lagrange multipliers. IEEE Transactions on Automatic Control, 70(11):7141–7156, November 2025.
[16] Laurent Lessard, Benjamin Recht, and Andrew Packard. Analysis and design of optimization algorithms via integral quadratic constraints. SIAM Journal on Optimization, 2016.
[17] Marcello Colombino, Emiliano Dall’Anese, and Andrey Bernstein. Online optimization as a feedback controller: Stability and tracking. IEEE Transactions on Control of Network Systems, 7(1):422–432, 2020.
[18] Adrian Hauswirth, Zhiyu He, Saverio Bolognani, Gabriela Hug, and Florian Dorfler. Optimization algorithms as robust feedback controllers. Annual Reviews in Control, 57:100941, 2024.
[19] Guido Carnevale, Nicola Mimmo, and Giuseppe Notarstefano. Nonconvex distributed feedback optimization for aggregative cooperative robotics. Automatica, 167:111767, 2024.
[20] Giuseppe Notarstefano, Ivano Notarnicola, and Andrea Camisa. Distributed optimization for smart cyber-physical networks. Foundations and Trends in Systems and Control, 7(3):253–383, 2019.
[21] Francesco Bullo, Jorge Cortes, and Sonia Martınez. Distributed Control of Robotic Networks. Princeton University Press, 2009.
[22] Aaron D. Ames, Xiangru Xu, Jessy W. Grizzle, and Paulo Tabuada. Control barrier functions: Theory and applications. IEEE Transactions on Automatic Control, 2017.
The aim of this action is to bring together researchers working on, or interested in contributing to, modern optimization challenges through control-theoretic tools. This will initially be pursued through a workshop planned for spring–summer 2027, with the goal of strengthening national momentum on the topic and fostering cross-collaboration. In the longer term, we will consider the possibility to establish a Technical Committee (TC) within SAGIP dedicated to the use of system-theoretic tools in modern optimization, meeting once or twice a year to discuss the latest developments in the field and identify methodological challenges. Possible outcomes of this activity include an overview paper, as well as the development of an ANR collaborative grant application.
Félicitations car votre dépôt d’action MACS a été accepté.
C’est un paisir de vous voir porter ce sujet très intéressant pour notre communauté.
Quelques éléments issus des discussions:
* Il vous faudra rapidement établir des contacts de personnes à impliquer dans cette action, identifier les personnes participant au workshop prévu. Ce sera bien de mettre leurs noms dans la fiche. Je peux vous suggérer Sophie Tarbouriech er Samuele Zobolli.
* Sur ces questions je vous encourage à regarder les derniers travaux de Carsten Scherer (U Stuttgart)
* Ca peut-êre intéressant de créer des liens avec des collègues qui s’intéressent aux convergences des algorithmes d’apprentissage tels que ceux présent dans la communauté du traitement du signal (GdR IASIS), je pense par exemple à Gersende Fort.
* La dernière rubrique de la fiche de votre action nommée “Evolution” vouspermet de raconter en ligne l’actualité de votre action. Ca peut être utile et plus simple que de cérer une page web dédiée.
* Concernant le budget, l’argent du GdR est annualisé. POur des dépenses en 2026 vous me faites une demande et je vous transfère l’argent vers un de vos labos. Idem pour 2027, mais ce sera en 2027.
* Je ferai une brève dans la prochaine lettre du GdR en annonçant votre action et inviterai les collègues à vous contacter s’ils sont intéressés à vous rejoindre.
Dimitri