Presentation

Research modern control and optimization technologies to help industry and society control and optimize their processes.

The main mission of the research unit DYSCO (Dynamical Systems, Control, and Optimization) is to develop new methodologies for the design of advanced multivariable controls that make systems react autonomously and optimally. Research revolves around model predictive control of dynamical systems, with specific emphasis on developing new algorithms for real-time convex embedded optimization, system identification and machine learning.

Besides focusing on algorithms and understanding their theoretical properties, research results are constantly applied to real-life problems of industrial, economic, and societal interest, such as in the automotive and aerospace industries, energy, process control, and financial engineering.

DYSCO is part of the research area in Computer Science and Engineering at the IMT School for Advanced Studies Lucca, Italy.

People

Professors

Assistant professors

Filippo Fabiani
Mario Eduardo Villanueva

PhD students

Shokhjakhon Abdufattokhov
Adeyemi Adeoye
Filippo Badalamenti
Chiara Cavalagli
Hasna El Hasnaouy
Matteo Facchino
Yuxuan Guo
Mikalai Korbit
Stefano Menchetti
Francesco Pillitteri
Hamid Salekinia
Saugat Shahi
Kui Xie

Research topics

Model Predictive Control (MPC)
Model Predictive Control (MPC) is widely adopted in industry for real-time control of large multivariable processes to optimize process operations under the best use of limited resources. The main idea of MPC is to choose the control action by solving an optimal control problem on line that minimizes a performance criterion over a future horizon, subject to constraints on process variables. The research unit investigates several issues in MPC, such as: stochastic MPC for constrained linear systems, decentralized and hierarchical MPC for spatially-distributed large-scale systems, MPC of networked systems based on wireless sensor feedback, explicit MPC and multiparametric programming for linear, hybrid, uncertain, and quantized systems.

Numerical optimization Numerical optimization algorithms for solving convex and non-convex mathematical programming problems, related to embedded MPC systems and to a large variety of areas, such as machine learning, various branches of engineering, and economics.

Machine learning and system identification
We develop new algorithms to learn dynamical models from data, with emphasis on linear parameter varying, piecewise affine, and general nonlinear models, and to learn control policies from data (reinforcement learning).

PhD courses

We offer the following PhD courses at IMT Lucca:

Software Tools

MATLAB/Python packages for real-time model predictive control, numerical optimization, system identification:

Model Predictive Control Toolbox (The Mathworks)
Hybrid Toolbox
jax-sysid - Linear and nonlinear system identification of state-space models
IDEAL - Inverse-Distance based Exploration for Active Learning
PARC - Piecewise Affine Regression and Classification
GLIS - Global and preference-based optimization of expensive black-box functions
Fitting Jump Models
ForBES - A MATLAB solver for nonsmooth convex optimization problems
MPCTOOL (European Space Agency)
MPCSofT (European Space Agency)

Applications

Automotive control systems
Autonomous driving, traction control, direct-injection engine control, semi-active suspensions, electromagnetic actuators, adaptive cruise control, robotized gearboxes, air-to-fuel ratio, active steering, idle speed control, power management and thermal management in hybrid electrical vehicles.

Aerospace control
Satellite attitude control, control of sloshing, guidance, navigation and control unmanned aerial vehicles (including formation flying and rendezvous), control of powered descent.

Industrial process control
Control of gas turbines, gas supply systems, flatness in cold tandem rolling, solar plants, cement mill scheduling.

Energy, management, and finance
Optimal bidding on energy markets, optimal power dispatch in smart grids, management of water distribution networks and sewer networks, dynamic hedging of financial options, optimal issuance of public debt securities.

Collaborations

The research unit collaborates with several European and American companies in applied research projects. The unit also has scientific collaborations with several universities worldwide.

Funding

See here.

Events