Marco Forgione
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Marco Forgione has a total of 38 co-authors. Here are the top co-authors by number of shared publications:
- Dario Piga
Dalle Molle Institute for Artificial Intelligence (IDSIA), SUPSI
15 shared publications - Alberto Bemporad
IMT School for Advanced Studies Lucca
3 shared publications - Simone Formentin
Dipartimento di Elettronica, Informazione e Bioingegneria Politecnico di Milano 20133 Milan Italy
3 shared publications - Alessio Benavoli
Trinity College Dublin
2 shared publications Dario Piga
SUPSI, IDSIA ‐ Dalle Molle Institute for Artificial Intelligence Lugano Switzerland
2 shared publications- Loris Roveda
Dalle Molle Institute for Artificial Intelligence, DTI, SUPSI, IDSIA, Lugano, Switzerland
2 shared publications
Latest Publications
- Open Access
- 17 August 2026
- Published by Wiley in International Journal of Adaptive Control and Signal Processing
dynoGP: Deep Gaussian Processes for Dynamic System Identification
In this work, we present a novel approach to system identification for dynamical systems, based on a specific class of deep Gaussian processes (deepGPs). These models are constructed by interconnecting linear dynamic GPs (equivalent to stochastic linear time‐invariant dynamical systems) and static GPs (to model static nonlinearities). Our approach combines the strengths of data‐driven methods, such as those based on neural network architectures, with the ability to output a probability distribution. This offers a more comprehensive framework for system identification that includes uncertainty quantification. Using both simulated and real‐world data, we demonstrate the effectiveness of the proposed approach.
- 17 August 2026
- Published by Wiley in International Journal of Adaptive Control and Signal Processing
Open AccessdynoGP: Deep Gaussian Processes for Dynamic System Identification
In this work, we present a novel approach to system identification for dynamical systems, based on a specific class of deep Gaussian processes (deepGPs). These models are constructed by interconnecting linear dynamic GPs (equivalent to stochastic linear time‐invariant dynamical systems) and static GPs (to model static nonlinearities). Our approach combines the strengths of data‐driven methods, such as those based on neural network architectures, with the ability to output a probability distribution. This offers a more comprehensive framework for system identification that includes uncertainty quantification. Using both simulated and real‐world data, we demonstrate the effectiveness of the proposed approach.
- Open Access
- 2 March 2026
- Published by Elsevier in Control Engineering Practice
- Vol. 172
Nonlinear system identification for a nano-drone benchmark
- 2 March 2026
- Published by Elsevier in Control Engineering Practice
- Vol. 172
Open AccessNonlinear system identification for a nano-drone benchmark
- 1 January 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Control Systems Letters
- Vol. 10, 2251-2256
ASIA: An Autonomous System Identification Agent
- 1 January 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Control Systems Letters
- Vol. 10, 2251-2256
ASIA: An Autonomous System Identification Agent
- Open Access
- 29 November 2025
- Published by Elsevier in Information Fusion
- Vol. 129
The duality of generative AI and reinforcement learning in robotics: A review
- 29 November 2025
- Published by Elsevier in Information Fusion
- Vol. 129
Open AccessThe duality of generative AI and reinforcement learning in robotics: A review
- Open Access
- 24 October 2025
- Published by Linkoping University Electronic Press in Swedish Aerospace Conference, FT2025
- Vol. 218, 469-477
Zero-Shot Parameter Estimation of Modelica Models using Patch Transformer Networks
This paper introduces a transformer-based generativenetwork for rapid parameter estimation of Modelica buildingmodels using simulation data from a Functional Mock-up Unit(FMU). Utilizing the \texttt{MixedAirCO2} model from theModelica Buildings library, we simulate a single-zonemixed-air volume with detailed thermal and \cotwo dynamics.By varying eight physical parameters and randomizingoccupancy profiles across 100 simulated systems, wegenerate a comprehensive dataset. The transformer encoder,informed by room temperature and \cotwo concentrationoutputs, predicts the underlying physical parameters withhigh accuracy and without re-tuning (hence, ``zero-shot'').This approach eliminates the need for iterativeoptimization or can be used to warm-start suchoptimization-based approaches, enabling real-time control,monitoring, and fault detection in FMU-based workflows.
- 24 October 2025
- Published by Linkoping University Electronic Press in Swedish Aerospace Conference, FT2025
- Vol. 218, 469-477
Open AccessZero-Shot Parameter Estimation of Modelica Models using Patch Transformer Networks
This paper introduces a transformer-based generativenetwork for rapid parameter estimation of Modelica buildingmodels using simulation data from a Functional Mock-up Unit(FMU). Utilizing the \texttt{MixedAirCO2} model from theModelica Buildings library, we simulate a single-zonemixed-air volume with detailed thermal and \cotwo dynamics.By varying eight physical parameters and randomizingoccupancy profiles across 100 simulated systems, wegenerate a comprehensive dataset. The transformer encoder,informed by room temperature and \cotwo concentrationoutputs, predicts the underlying physical parameters withhigh accuracy and without re-tuning (hence, ``zero-shot'').This approach eliminates the need for iterativeoptimization or can be used to warm-start suchoptimization-based approaches, enabling real-time control,monitoring, and fault detection in FMU-based workflows.
- 24 June 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Control Systems Letters
- Vol. 9, 1652-1657
Distributionally Robust Minimization in Meta-Learning for System Identification
Meta learning aims at learning how to solve tasks, and thus it allows to estimate models that can be quickly adapted to new scenarios. This work explores distributionally robust minimization in meta learning for system identification. Standard meta learning approaches optimize the expected loss, overlooking task variability. We use an alternative approach, adopting a distributionally robust optimization paradigm that prioritizes high-loss tasks, enhancing performance in worst-case scenarios. Evaluated on a meta model trained on a class of synthetic dynamical systems and tested in both in-distribution and out-of-distribution settings, the proposed approach allows to reduce failures in safety-critical applications.
- 24 June 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Control Systems Letters
- Vol. 9, 1652-1657
Distributionally Robust Minimization in Meta-Learning for System Identification
Meta learning aims at learning how to solve tasks, and thus it allows to estimate models that can be quickly adapted to new scenarios. This work explores distributionally robust minimization in meta learning for system identification. Standard meta learning approaches optimize the expected loss, overlooking task variability. We use an alternative approach, adopting a distributionally robust optimization paradigm that prioritizes high-loss tasks, enhancing performance in worst-case scenarios. Evaluated on a meta model trained on a class of synthetic dynamical systems and tested in both in-distribution and out-of-distribution settings, the proposed approach allows to reduce failures in safety-critical applications.
- 22 June 2025
- Published by Springer Nature
- p. 283-287
Physics-Informed Neural Networks for Hidden Insulin Dynamics Estimation from Glucose Data
- 22 June 2025
- Published by Springer Nature
- p. 283-287
Physics-Informed Neural Networks for Hidden Insulin Dynamics Estimation from Glucose Data
- 5 May 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Journal of Biomedical and Health Informatics
- Vol. 29 (10) , 7387-7395
Contactless Estimation of Respiratory Frequency Using 3D-CNN on Thermal Images
Monitoring physiological parameters such as respiratory rate (f$_{R}$) is essential for diagnosing and managing various pathological conditions. Thermal imaging offers a promising contactless alternative to traditional methods, which often rely on partially invasive sensors or obtrusive wearable systems. However, existing approaches for f$_{R}$ estimation from thermal signals typically require extensive pre-processing and manual or semi-automatic region-of-interest (ROI) tracking, limiting their practical applicability. This study proposes a deep learning-based method for estimating f$_{R}$ directly from thermal videos, eliminating the need for complex pre-processing and ROI tracking. A 3D Convolutional Neural Network (3D-CNN) is developed to operate on raw thermal video data. To address challenges related to small datasets, the model is trained using data augmentation and transfer learning from synthetic datasets. Experimental results demonstrate that the proposed approach achieves a validation $R^{2}$ score of approximately 0.61 on both pre-processed and raw thermal videos. By simplifying the workflow, this method holds promise for enhancing the feasibility of thermal imaging in real-world applications, such as remote healthcare and driver monitoring in automotive applications.
- 5 May 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Journal of Biomedical and Health Informatics
- Vol. 29 (10) , 7387-7395
Contactless Estimation of Respiratory Frequency Using 3D-CNN on Thermal Images
Monitoring physiological parameters such as respiratory rate (f$_{R}$) is essential for diagnosing and managing various pathological conditions. Thermal imaging offers a promising contactless alternative to traditional methods, which often rely on partially invasive sensors or obtrusive wearable systems. However, existing approaches for f$_{R}$ estimation from thermal signals typically require extensive pre-processing and manual or semi-automatic region-of-interest (ROI) tracking, limiting their practical applicability. This study proposes a deep learning-based method for estimating f$_{R}$ directly from thermal videos, eliminating the need for complex pre-processing and ROI tracking. A 3D Convolutional Neural Network (3D-CNN) is developed to operate on raw thermal video data. To address challenges related to small datasets, the model is trained using data augmentation and transfer learning from synthetic datasets. Experimental results demonstrate that the proposed approach achieves a validation $R^{2}$ score of approximately 0.61 on both pre-processed and raw thermal videos. By simplifying the workflow, this method holds promise for enhancing the feasibility of thermal imaging in real-world applications, such as remote healthcare and driver monitoring in automotive applications.
- Open Access
dynoGP: Deep Gaussian Processes for dynamic system identification
In this work, we present a novel approach to system identification for dynamical systems, based on a specific class of Deep Gaussian Processes (Deep GPs). These models are constructed by interconnecting linear dynamic GPs (equivalent to stochastic linear time-invariant dynamical systems) and static GPs (to model static nonlinearities). Our approach combines the strengths of data-driven methods, such as those based on neural network architectures, with the ability to output a probability distribution. This offers a more comprehensive framework for system identification that includes uncertainty quantification. Using both simulated and real-world data, we demonstrate the effectiveness of the proposed approach.
Open AccessdynoGP: Deep Gaussian Processes for dynamic system identification
In this work, we present a novel approach to system identification for dynamical systems, based on a specific class of Deep Gaussian Processes (Deep GPs). These models are constructed by interconnecting linear dynamic GPs (equivalent to stochastic linear time-invariant dynamical systems) and static GPs (to model static nonlinearities). Our approach combines the strengths of data-driven methods, such as those based on neural network architectures, with the ability to output a probability distribution. This offers a more comprehensive framework for system identification that includes uncertainty quantification. Using both simulated and real-world data, we demonstrate the effectiveness of the proposed approach.
- Open Access
One controller to rule them all
Imagine having a system to control and only know that it belongs to a certain class of dynamical systems. Would it not be amazing to simply plug in a controller and have it work as intended? With the rise of in-context learning and powerful architectures like Transformers, this might be possible, and we want to show it. In this work, within the model reference framework, we hence propose the first in-context learning-based approach to design a unique contextual controller for an entire class of dynamical systems rather than focusing on just a single instance. Our promising numerical results show the possible advantages of the proposed paradigm, paving the way for a shift from the "one-system-one-controller" control design paradigm to a new "one-class-one-controller" logic.
Open AccessOne controller to rule them all
Imagine having a system to control and only know that it belongs to a certain class of dynamical systems. Would it not be amazing to simply plug in a controller and have it work as intended? With the rise of in-context learning and powerful architectures like Transformers, this might be possible, and we want to show it. In this work, within the model reference framework, we hence propose the first in-context learning-based approach to design a unique contextual controller for an entire class of dynamical systems rather than focusing on just a single instance. Our promising numerical results show the possible advantages of the proposed paradigm, paving the way for a shift from the "one-system-one-controller" control design paradigm to a new "one-class-one-controller" logic.