1) Knowledge and understanding Simulation intelligence: You will gain comprehension of state-of-the-art techniques to learn physics-informed (PI) surrogate models and an understanding of their strengths and weaknesses. You will learn how to choose the architecture that better suits your data and your physical model. Learning for autonomous systems: Know different kinds of problems that can be tackled with learning-based (LB) control. Know the general working schemes of the most common LB approaches. Know the pros and cons of LB control. Know how to design and assess LB controller's performance. 2) Applying knowledge and understanding Simulation intelligence: You will be able to state and develop the inference problem of a physicinformed surrogate model. Learning for autonomous systems: Apply LB control approaches in practical control problems. 3) Making judgments Judge the theoretical and technical soundness of a PI surrogate model and of an LB-controlled system. 4) Communication skills Describe the reasoning behind the design, development, and assessment choices of PI surrogate models and LB-controlled systems. 5) Learning skills Navigate the scientific literature regarding simulation intelligence techniques and retrieve information from scientific publications about techniques handling LB-based control applications that have not been explicitly presented in this course

Knowledge from the "Introduction to Machine Learning" course. Knowledge of Python and scientific Python. Basic knowledge of matrix operations, differential equations, and Calculus. Fundamentals of Control Theory.

This course offers a deep exploration of cutting-edge scientific fields, providing students with a solid foundation and advanced insights. The Simulation Intelligence (SI) module will delve into the intriguing realm of Physics-Infused Surrogate Modeling where probabilistic and differentiable programming are key. Students will gain a thorough understanding of how surrogate models can be effectively applied in simulation-based inference techniques, enabling to address complex problems. Additionally, students will explore the critical aspects of causality and uncertainty, equipping them with the necessary tools to navigate these challenges. Transitioning to the Learning for Autonomous Systems (LAS) module, students will embark on a comprehensive study of autonomous systems. They will delve into optimal control problems, examining the underlying principles and methodologies. Furthermore, they will explore the evolution from model-based to model-free control strategies, providing them with a deep understanding of their advantages and limitations. Through practical applications and real-world examples, students will develop a holistic perspective on the capabilities and challenges of autonomous systems. Throughout the course, students will engage in a variety of activities designed to enhance your learning experience. These include hands-on sessions where they will apply theoretical concepts using practical tools and techniques.

The Simulation Intelligence part builds on Lavin, Alexander, et al. "Simulation intelligence: Towards a new generation of scientific methods." arXiv preprint arXiv:2112.03235 (2021). All the relevant references can be found therein. The Learning for autonomous systems part builds on: Bertsekas, D. (2019). Reinforcement learning and optimal control. Athena Scientific. Other relevant references as well as slides will be provided throughout the course.

A) Simulation Intelligence 1) Introduction to Probabilistic and Differentiable Programming 2) Physics-Infused Surrogate Modeling: - Neural Ordinary Differential Equations (Neural ODE), - Physic-Informed Neural Network (PINN), - Neural Operators (NO) 3) Overview of Simulation-Based Inference Techniques 4) Causal Reasoning 5) Uncertainty Estimation B) Learning for autonomous systems 1) Introduction to Autonomous Systems 2) Optimal control problems 3) From model-based to model-free control 4) Reinforcement Learning - Reinforcement Learning terminology - Value-Based Methods - Policy Optimization - Deep Reinforcement Learning

Frontal lectures and hands-on sessions, both individually and in groups. Hands-on activities typically involve experimenting with Python libraries and developing or using/testing tools implementing the methodologies seen during lectures.

Bring your own laptop

The examination will consist of an oral presentation of two relevant papers or libraries investigating and testing respectively a simulation intelligence approach and a learning methodology for autonomous systems not seen in detail during the lectures. The main assessment points are the clarity and precision of the answers, Testi in inglese Language English This course offers a deep exploration of cutting-edge scientific fields, providing students with a solid foundation and advanced insights. The Simulation Intelligence (SI) module will delve into the intriguing realm of Physics-Infused Surrogate Modeling where probabilistic and differentiable programming are key. Students will gain a thorough understanding of how surrogate models can be effectively applied in simulation-based inference techniques, enabling to address complex problems. Additionally, students will explore the critical aspects of causality and uncertainty, equipping them with the necessary tools to navigate these challenges. Transitioning to the Learning for Autonomous Systems (LAS) module, students will embark on a comprehensive study of autonomous systems. They will delve into optimal control problems, examining the underlying principles and methodologies. Furthermore, they will explore the evolution from model-based to model-free control strategies, providing them with a deep understanding of their advantages and limitations. Through practical applications and real-world examples, students will develop a holistic perspective on the capabilities and challenges of autonomous systems. the technical understanding of the methods, and the understanding of their conditions of applicability.

This course explores topics closely related to one or more goals of the United Nations 2030 Agenda for Sustainable Development (SDGs)