SPARCS - SPace, Aerial, and gRound Control Systems
Founder and leader of SPARCS - SPace, Aerial, and gRound Control Systems group.
The group activity lies at the nexus of Robotics, Computational Vision, and Learning: building on the fundamentals of Dynamical Systems and Networked Systems Theory, it focuses on mobile robotics, autonomous driving, and visual sensor networks, developing both methodological and experimental research.
The group currently includes 2 fixed-term researchers, 1 post-doc, 6 PhD students, 1 research fellow, and affiliates per year 1 PhD + 1 research fellow + 4-5 Master Students.
Research interests: Multiagent and networked control systems, Rigidity theory, Structural properties of robotics systems with mobile and/or manipulation capabilities, Active sensing and camera networks, ICT and Hyperautomation for the Industry.
Main methodological aspects include: theory of pose representation, distance/bearing/hybrid rigidity theory, dis- tributed optimization over manifolds and over graphs, nonlinear modeling and advanced control architectures, cooperative and coordinated control of multi-agent formations, time-sensitive networking.
Experimental activities involve: rapid prototyping and ROS-based architectures, co-simulation framework for robotics and industrial scenarios.
Research topics
Networked control systems
Networked control systems are multiagent, multitask networks with limited resources. They employ cooperation and the distributed coordination of sensors/actuators to fulfill complex tasks that are not possible to a single agent. Applications of interest include:
- localization and tracking problem for mobile nodes in multiagent networks
- coverage control and information search in robotic networks
- synchronization and swarm control in large scale multiagent systems
- sensor networks for environmental monitoring
- smart environments: smart autonomous greenhouse, domotics
- industrial wireless sensor networks and communication networks
Rigidity control & mobile robotic systems
Rigidity theory has emerged as an efficient tool in the control field of coordinated multi-agent systems, such as multi-robot formations and UAVs swarms characterized by sensing, com- munication and movement capabilities. Interesting connections are found with the theory of graphical models, on the one side, and with that of parallel robotics, on the other. This research is applied to the study of methodologies and the development of algorithms for estimation and control of autonomous aerial and ground vehicles (UAVs-UGVs) and satellites. Applications of interest include:
- SE(2)-SE(3) rigidity theory
- UAV attitude estimation and control
- formation control in multiagent systems
- navigation and docking control for aerial and space systems
- heterogeneous vehicles formation management and control
- autonomous guidance and navigation for aerial and ground systems
Visual Sensor Networks
Sparse camera networks: Automated (Pan-Tilt-Zoom) cameras and fixed cameras cooperate with other sensing devices in distributed networks to perform coordinated tasks of area patrolling, event detection, and event tracking. The system is autonomic: the sensors are smart agents able to coordinate to maximize monitoring and detection performance, manage complex tasks, accommodate for communication/device losses (self-healing), and perform active sensing policies. The information-rich visual sensors act in this sense in conjunction with the other more constrained sensing devices. The outcome of (semantic) scene understanding is a balanced trade-off of the heterogeneous network constraints and requirements.
Dense camera networks: 3D reconstruction in motion capture systems shows critical issues when scaling with the number of cameras or the complexity of the scene. In this context a distributed approach is proposed to solve the multicamera reconstruction problem in large scale motion capture systems.
ICT and Hyperautomation for the Industry
The employment of modern methodologies in the industry is enabling a deep transformation of the traditional manufacturing process, where tools from the communication and information technologies are key for better system performance, improved process efficiency, and increased quality of service/product. In this context, we study how to allow hyperautomation by seating at the nexus among computational vision, learning techniques, cyber-physical system modeling, time-sensitive network communication, under the wide control umbrella. Typical applications include
- fault detection, isolation and mitigation in industrial operation
- process and product quality control
- safety and security in manufacturing systems
- real-time motion and process control
- industrial network traffic control
- industrial Internet of Things
Fusion plasmas modeling and control
Applications of interest include control, diagnostics and modeling in the field of computational electromagnetics:
- sensor selection through convex optimization for global parameter estimation
- digital-twin and model-order reduction for large scale systems
- magnetic map reconstruction techniques from sparse heterogeneous sensors
- plasma modeling and regulator design, for the real time control of plasma current, position and shape
- study of linear and nonlinear models for simulation of the whole plasma-plant system
Scientific seminars and invited talks
(conference talks are omitted):
- On Formation Control for Multi-Agent Systems
"Rigidity Theory for Multi-Agent Systems meets Parallel Robotics" Workshop, Nantes (France) - November 30, 2018 - A distributed approach to Visual Sensor Networks
Dept. of Industrial Engineering, University of Trento, Trento (Italy) - November 18, 2016 - Distributed approach to dense and sparse camera networks[pdf (10MB)]
Max Planck Gesellschaft, Tubingen (Germany) - June 5, 2013 - Reti di sensori applicate al monitoraggio e al controllo di edifici [pdf (14MB)]
WISEWAI Workshop, Padova (Italy) - May 30, 2011
- Turbulence in Adaptive Optics Systems: Modeling, Estimation and Compensation
UCLA - Applied Math Colloquium, Los Angeles (CA-USA) - May 17, 2010
Caltech - CDS seminar, Los Angeles (CA-USA) - May 19, 2010
- Multi-Agent Systems: Modeling, Estimation and Control Issues - Case Study: Camera Networks [pdf (8MB)]
UCB - BCCI seminar, Berkeley (CA-USA) - May 3, 2010
UCSB - Control Group seminar, Santa Barbara (CA-USA) - May 12, 2010
- Deformable Objects: Shape Analysis and Deformation Control
KTH seminar, Stockholm (Sweden) - January 30, 2009
- Deformable Shape Theory: from Controlled Active Vision to
Optimal Shape Control of Deformable Objects to the Analysis of
Collective Behavior
DTG seminar, Vicenza (Italy) - April 7, 2006
- Deformable Models: dalla Ricostruzione di Forma alla Simulazione e al Controllo di Sistemi Dinamici
DEI colloquia, Padova (Italy) - November 11, 2004
- Development of a boundary tracing routine from an active vision technique
General Atomics, San Diego (CA-USA) – September 3, 2004
- Active Contours Techniques for Plasma Boundary Reconstruction and Control
General Atomics, San Diego (CA-USA) – August 25, 2004