Learning and Control (LC)


A massive amount of data is generated by sensors, smartphones, computers, and web platforms, inside home automation systems and intelligent devices, artifacts such as vehicles and robots, production processes, smart energy grids, and many others. This availability of data, combined with the possibility of having increasingly powerful and pervasively distributed computing units both within the devices (embedded) or in multiprocessor boards connected to them and in the cloud, now makes new skills necessary to know how to use the data to predict the behavior of the system that generates them and to make decisions autonomously based on the information contained therein, preferably in an efficient and robust way from a computational point of view. These skills are paramount for engineering industrial automation systems and robots and in various other contexts, such as critical infrastructures (energy networks, urban mobility, water networks), self-driving vehicles, financial systems, biomedical systems, home automation, etc.

The Learning and Control (LC) doctoral track provides interdisciplinary doctoral training to graduate students who want to develop algorithms for machine learning and for controlling dynamical systems based on numerical optimization techniques. By utilizing these methodologies, it becomes possible to comprehend the dynamics of the system that produces the data. Using mathematical models learned from the data, one can analyze the behavior of the system; predict potential future scenarios; and diagnose malfunctions. Moreover, such models allow one to improve the overall system behavior using real-time control algorithms, making the system autonomous in acting optimally and safely to pursue pre-established objectives. Since these methodologies are not dependent on the physical nature of the system being studied, they can be applied to a wide range of real-world problems. For instance, they can enable a vehicle to drive autonomously while avoiding obstacles, allow a satellite to adjust its attitude, and help a smart electricity grid optimize the use of energy from renewable sources.

Ph.D. courses for the Learning and Control Track

Advanced Topics in Machine Learning
Applications of Stochastic Processes
Dynamics on Complex Networks
Foundations of Probability and Statistical Inference
Fundamentals of Academic Entrepreneurship
Fundamentals of Numerical Analysis
Funding Opportunities and Management of Intellectual Property
Game Theory
Introduction to Machine Learning
Introduction to Network Science
Markov Processes
Model Predictive Control
Network Reconstruction
Numerical Methods for Optimal Control
Numerical Methods for the Solution of Partial Differential Equations
Numerical Optimization
Optimal Control and Differential Games
Publication Strategies and Scientific Dissemination
Principles of Programming with Python
Python for Data Science
Reduced Order Models and Applications
Reinforcement Learning
Stochastic Processes

Track structure

The program's curriculum comprises fundamental courses that provide comprehensive training on various topics such as machine learning techniques, numerical optimization, analysis and control of dynamical systems, and computer programming. In particular:

Apart from the fundamental courses, students will have access to advanced research seminars covering cutting-edge topics, and they will have the opportunity to attend thematic doctoral schools. The program also allows students to spend a research period abroad.

Input and Output Profiles 

Prospective students should preferably have training in engineering, mathematics, computer science, physics, statistics, or a related field. Potential students are offered frontier research topics or are free to propose a research topic of interest to them.

The LC track prepares researchers and professionals capable of analyzing and proposing solutions to various real problems of industrial, economic, and social interest, making them qualified to work in high-profile professional roles within universities, research centers, and in the private sector, such as in automotive, aerospace, chemical, manufacturing, infrastructure, energy, urban mobility, biomedical, and various other sectors. Professional figures able to manipulate data using mathematical algorithms are also particularly sought after in emerging sectors such as electronic commerce, social networks, finance, and many others. These Ph.D. figures are particularly appreciated for their extreme versatility, mastering methodologies for approaching the formulation and resolution of problems, and very general algorithmic and computer skills.

Ph.D. students have the opportunity to collaborate with other institutions and companies with which the Research Units of the IMT School have established partnerships.

For more information regarding the research activities and the researchers related to the LC track, please refer to the home page of the Research Unit DYSCO (DYnamical Systems, Control, and Optimization).