Systems & Control Lab

Topics

Systems and Control area is devoted to teaching and research in various fields related to control system science and engineering, industrial automation, robotics, systems theory, environmental systems, ecology, and operations research.

Networked Control Systems and Information Fusion

Technological advances in computing, communication and control have paved the way for a new generation of engineering systems, called cyber-physical systems (CPS) where physical and software components are deeply intertwined. Application areas are immense, ranging from autonomous driving to energy systems and customised manufacturing for the Industry 4.0. A main challenge of CPS is that they feature tight interactions between the continuous dynamics of physical systems and the discrete dynamics of cyber components. Our research focuses on modelling, analysis, monitoring and control of CPS, combining tools from control theory with tools that are typical of computer and network science such as automata and graph theory.

Networked control systems (NCS) are systems where monitoring and control tasks are carried out using a communication network. The concept of NCS is extremely appealing for industry, especially for wireless automation, but it raises many theoretical and practical challenges due to the fact that communication networks introduce several forms of uncertainty such as transmission delay and packet dropouts. Our research focuses on the design of monitoring and control systems that are resilient against network imperfections, as well as on energy-aware transmission logics like event- and self-triggered control.

Data fusion aims to combine information from multiple, often heterogeneous, data sources so as to provide better situation awareness that any individual source could give. Fundamental theoretical issues in data fusion are to avoid double counting of information and at the same time allow the useful integration of complementary information. Relevant applications of data fusion that are of interest to our group concern distributed monitoring/surveillance via sensor networks and cooperation of autonomous agents (e.g. self-driving vehicles or mobile robots).

Sensor networks consist of the interconnection of multiple nodes with sensing, communication, processing and, sometimes, motion capabilities. Such networks have nowadays wide application in various domains including environmental/industrial/health monitoring, intelligent transportation systems, smart cities/buildings, ground/air/maritime surveillance, etc.. The main challenge of our work on this topic is to develop distributed estimation algorithms that are scalable with respect to the network size, perform as close as possible to the centralized estimator, and are communication/energy efficient.

Adaptation refers to the ability of a system to adjust to different conditions within its environment.  Modern engineering systems are complex and must be flexible and capable of learning and adapting in the face of varying operating conditions so as to optimize performance. The activity of our research group in the broad area of adaptive and learning systems ranges from application of machine learning techniques in identification, estimation, and control to real-time control reconfiguration and data-driven control design.

 

Control Systems and Nonlinear Dynamics

Nonlinear control systems have been extensively investigated in the literature since real-world control systems usually contain blocks whose nonlinear nature cannot be ignored. How to deal with nonlinear systems strongly depends on their inner structure, the considered phenomenon, and the control goal control. Current research topics: analysis of oscillations and their bifurcations in memristor-based devices, actually seen as basic nanoscale elements which may give rise to a new analogue computational paradigm; analysis and control of aero-servo-elastic systems, with particular attention to the problem of flutter for airplane wings

The interplay of dynamics, interactions and topology in a network of dynamical systems is the central theme of this area of research. Applications are multiple, covering areas such as physics, biology, engineering, and ranging from neural networks up to social webs. Modeling nodes and their functional responses and estimating their mutual interconnections are crucial issues. Current research topics: identification of the hidden structural properties, which links the underlying topology with the emerging dynamical behavior of the network; development of distributed local controllers able enforce the network displaying a desired dynamics.

Model Predictive Control (MPC) aims to optimize the performance of a control algorithm with respect to a set of constraints and benchmarks by using a model to predict the system behavior. Current research topics: design and analysis of Economic MPC to directly and dynamically optimize the profitability of a plant while simultaneously enhancing both the transient and steady-state behavior of the system; application of the MPC approach to self- and event-triggered controllers to optimize the schedule of data transfers in control networks, as required by the Industry 4.0 framework.

Robust stability and optimal performance are issues of major concern in the design of feedback control systems. Current research topics: robust stability in the face of exogenous disturbances for nonlinear systems with special attention on the so called Input-to-State Stability approach; synthesis of a finite family of switching linear time invariant controllers for improving robust performance of control systems with uncertain plants; bolstering controllers designed via feedback linearization techniques, when they face possibly inexact nonlinear cancellation issues.

Autonomous systems display self-sustained behaviors, which depend on configuration and initial conditions, but which are not driven by external inputs. A fundamental issue in the field pertains “bifurcations”, i.e., sudden modifications of the behavior activated by varying the set up. Multi-agent systems are a special subclass of autonomous systems, where the internal structure is naturally divided into interconnected non-autonomous subsystems (the agents). Current research topics: consensus problems, where the agents interconnections are designed to reach an agreement, and formation control, with special focus on car platooning.