Research profile dr. P. (Pietro) Pesi

Description of research:

Keywords: Systems and control theory; cyber-physical systems; dynamical networks; data-driven control design.

Resilient control of cyber-physical networks

Energy, telecommunications, and transportation infrastructures are becoming increasingly complex. There is in fact a trend to build modern infrastructures as large-scale networks, where multiple computational and physical elements interact with one another. Due to the tight conjoining of and coordination between computational and physical resources, these systems are referred to as Cyber-Physical Systems (CPSs). CPSs pose several challenges for their efficient, secure and reliable operation. Resilient control systems consider all of these elements. In my group, we analyze and design resilient control systems with applications spanning from distributed sensor/actuator networks and electrical power systems to district heating systems and data networks. My main interest lies in designing control solutions that secure the resilience of CPSs against failures in the communication layer, possibly of a malicious nature.

(Cyber-physical system security).

Detecting topology variations in dynamical networks

To achieve full resilience, it is fundamental to have algorithms that can reveal changes in the network infrastructure (physical and communication topology) with speed and accuracy. In fact, this enables better predictions, timely warnings and control counteractions. I am interested in the design of methodologies and algorithms for real-time monitoring of dynamical networks, with main focus on the following issues: i) what are the conditions under which we can detect topology changes and identify the current network topology using dynamical data; ii) how many measurement devices are needed and at which locations in the network they should be deployed; and iii) which topologies and dynamics can render detection easier and how this can be integrated into existing infrastructures. Answering these questions may further our understanding of the behavior of such complex systems, and, in the long run, lead to the development of numerically efficient algorithms to be integrated into supervisory control and data acquisition systems.

Data-driven control design

Modern engineering systems are becoming more large and complex. In some cases, modeling these systems in an accurate way using first principles or identification can be difficult and time-consuming. Data-driven control design is a branch of control theory addressing the question: “How can we develop direct data-to-controller algorithms”? I am interested in both theoretical and practical aspects of this problem. In particular, my research focuses on computationally efficient optimization algorithms that can provide stability guarantees without requiring multiple datasets (experiments). One fascinating application of data-driven control design is in the area of Adaptive Optics for ground-based telescopes. In collaboration with the National Institute for Astrophysics, Italy, I work on the use of data-driven control for the design of Adaptive Secondary Mirrors position control systems.