Control of Autonomous Multi-Agent Systems

Course objectives

General objectives This course deals with modeling, analysis and control of multi-agent systems, with emphasis on communication/distribution networks and multi-robot systems. Specific objectives Knowledge and understanding: Students will learn basic methods for the modeling, analysis and control of multi-agent systems, with particular attention to distributed control strategies. In the first part of the course, applications relating to communication, energy and health networks/systems will be presented; in the second part, multi-robot systems will be studied, both terrestrial and aerial. Apply knowledge and understanding: Students will be able to analyze and design architectures and algorithms for the control of multi-agent systems in various application fields. Critical and judgment skills: Students will be able to choose the most suitable control methodology for a specific problem and to evaluate the complexity of the proposed solution. Communication skills: The course activities allow the student to be able to communicate/share the main problems concerning networks and systems presented in the course, as well as the possible design choices for the control of such networks/systems. Learning ability: The course aims at giving the students a mindset oriented to the control of complex systems/networks by appropriately combining methodologies coming from the control theory as well as from other engineering disciplines. Furthermore, the course includes the possibility of carrying out application theses on topics related to projects carried out by the research group coordinated by the teachers; as part of these activities, students will acquire the ability to collaborate in groups.

Channel 1
ANDREA CRISTOFARO Lecturers' profile

Program - Frequency - Exams

Course program
Part 1: Examples of multi-agent scenarios in communication, energy, transport, health networks/systems derived from research projects funded by the European Union. Comparison between centralized and distributed architectures. Extension of the methodologies (especially machine learning, reinforcement learning, model predictive control) and problems (concerning the control of communication, energy, transport and health networks/systems, as well as the security of such networks/systems) studied in the context of previous courses (in particular, Control of Communication and Energy Network) to the multi-agent context. Part 2: Examples of applications of multi-robot systems. Centralized vs. decentralized architectures. Elements of graph theory. Connectivity and Consensus. Passivity and Lyapunov stability. Port-Hamiltonian systems. Interconnection of mechanical systems. Application to multi-UAV systems: Formation control with time-varying topology. Formation control with connectivity maintenance. Cooperative exploration of unknown environments: distributed SLAM.
Prerequisites
None
Books
M. Mesbahi, M. Egerstedt, Graph Theoretic Methods in Multi-Agent Networks, Princeton Academic Press
Teaching mode
Class lectures
Frequency
Recommended
Exam mode
The exam consists in the presentation of a group project (2-4 people) or in the individual presentation of a research paper
Bibliography
C. Godsil, G. Royle, Algebraic Graph Theory, Springer
Lesson mode
Class lectures
ANDREA CRISTOFARO Lecturers' profile

Program - Frequency - Exams

Course program
Part 1: Examples of multi-agent scenarios in communication, energy, transport, health networks/systems derived from research projects funded by the European Union. Comparison between centralized and distributed architectures. Extension of the methodologies (especially machine learning, reinforcement learning, model predictive control) and problems (concerning the control of communication, energy, transport and health networks/systems, as well as the security of such networks/systems) studied in the context of previous courses (in particular, Control of Communication and Energy Network) to the multi-agent context. Part 2: Examples of applications of multi-robot systems. Centralized vs. decentralized architectures. Elements of graph theory. Connectivity and Consensus. Passivity and Lyapunov stability. Port-Hamiltonian systems. Interconnection of mechanical systems. Application to multi-UAV systems: Formation control with time-varying topology. Formation control with connectivity maintenance. Cooperative exploration of unknown environments: distributed SLAM.
Prerequisites
None
Books
M. Mesbahi, M. Egerstedt, Graph Theoretic Methods in Multi-Agent Networks, Princeton Academic Press
Teaching mode
Class lectures
Frequency
Recommended
Exam mode
The exam consists in the presentation of a group project (2-4 people) or in the individual presentation of a research paper
Bibliography
C. Godsil, G. Royle, Algebraic Graph Theory, Springer
Lesson mode
Class lectures
FRANCESCO DELLI PRISCOLI Lecturers' profile

Program - Frequency - Exams

Course program
Part I (about 30 hours) Examples of multi-agent scenarios in communication, energy, transport, health networks/systems derived from research projects funded by the European Union and/or by ESA and/or by MIMIT or MUR. Comparison between centralized and distributed architectures. Extension of the methodologies (especially machine learning, reinforcement learning, model predictive control) and problems (concerning the control of communication, energy, transport and health networks/systems, as well as the security of such networks/systems) studied in the context of previous courses (in particular, Control of Communication and Energy Network) to the multi-agent context. Part II (about 30 hours) Introduction: Examples of application of multi-robot systems. Centralized vs. decentralized architectures. Mathematical tools: Adjacency graph and matrix; Laplacian; Connectivity and Consensus; Passivity and Lyapunov stability; Interconnection of mechanical systems. Application to multi-UAV systems: Formation control with time-varying topology; Formation control with connectivity maintenance; Steady-state behaviors; Bearing-based formation control. Application to multi-UGV systems: Cooperative exploration of unknown environments; Mutual localization with anonymous measurements; Target localization and encircling.
Prerequisites
No pre-requirement is necessary. There are no pre-requisit exams.
Books
Lecture notes and papers mainly derived from Deliverables produced by up-to-date EU research projects (in particular, NANCY (FP9), CADUCEO (MIMIT), ATENA (UE FP8), SESAME (UE FP8), 5G ALLSTAR (UE FP8), 5G SOLUTIONS (UE FP8), ARIES (ESA), VADUS (ESA) projects). I. Goodfellow, Y. Bengio, A. Courville, "Deep Learning", MIT Press, 2016. M. Vidal, “Fundamentals of Multiagent Systems,” 2011. M. Mesbahi and M. Egerstedt, “Graph Theoretic Methods in Multiagent Systems,” Princeton University Press, 2010.
Teaching mode
Traditional with possible project. The course is taught by using the blackboard and/or slides depending on the topic. If it is not possible to carry out the lessons in the classroom with all the students due to the pandemic, the lessons are carried out in mixed mode (in the classroom with streaming for students who use remotely) or exclusively in streaming if no student can attend. The instructions for the first part of the course are indicated on the course page https://corsidilaurea.uniroma1.it/it/users/francescodellipriscoliuniroma1it
Frequency
Attendance at the lessons is not mandatory, but is strongly recommended
Exam mode
For the first part of the course the student must carry out an in-depth project ("tesina") (with personal contribution required), to be carried out at home, of one of the seminars held during the course.
Lesson mode
Traditional with possible project. The course is taught by using the blackboard and/or slides depending on the topic. If it is not possible to carry out the lessons in the classroom with all the students due to the pandemic, the lessons are carried out in mixed mode (in the classroom with streaming for students who use remotely) or exclusively in streaming if no student can attend. The instructions for the first part of the course are indicated on the course page https://corsidilaurea.uniroma1.it/it/users/francescodellipriscoliuniroma1it
FRANCESCO DELLI PRISCOLI Lecturers' profile

Program - Frequency - Exams

Course program
Part I (about 30 hours) Examples of multi-agent scenarios in communication, energy, transport, health networks/systems derived from research projects funded by the European Union and/or by ESA and/or by MIMIT or MUR. Comparison between centralized and distributed architectures. Extension of the methodologies (especially machine learning, reinforcement learning, model predictive control) and problems (concerning the control of communication, energy, transport and health networks/systems, as well as the security of such networks/systems) studied in the context of previous courses (in particular, Control of Communication and Energy Network) to the multi-agent context. Part II (about 30 hours) Introduction: Examples of application of multi-robot systems. Centralized vs. decentralized architectures. Mathematical tools: Adjacency graph and matrix; Laplacian; Connectivity and Consensus; Passivity and Lyapunov stability; Interconnection of mechanical systems. Application to multi-UAV systems: Formation control with time-varying topology; Formation control with connectivity maintenance; Steady-state behaviors; Bearing-based formation control. Application to multi-UGV systems: Cooperative exploration of unknown environments; Mutual localization with anonymous measurements; Target localization and encircling.
Prerequisites
No pre-requirement is necessary. There are no pre-requisit exams.
Books
Lecture notes and papers mainly derived from Deliverables produced by up-to-date EU research projects (in particular, NANCY (FP9), CADUCEO (MIMIT), ATENA (UE FP8), SESAME (UE FP8), 5G ALLSTAR (UE FP8), 5G SOLUTIONS (UE FP8), ARIES (ESA), VADUS (ESA) projects). I. Goodfellow, Y. Bengio, A. Courville, "Deep Learning", MIT Press, 2016. M. Vidal, “Fundamentals of Multiagent Systems,” 2011. M. Mesbahi and M. Egerstedt, “Graph Theoretic Methods in Multiagent Systems,” Princeton University Press, 2010.
Teaching mode
Traditional with possible project. The course is taught by using the blackboard and/or slides depending on the topic. If it is not possible to carry out the lessons in the classroom with all the students due to the pandemic, the lessons are carried out in mixed mode (in the classroom with streaming for students who use remotely) or exclusively in streaming if no student can attend. The instructions for the first part of the course are indicated on the course page https://corsidilaurea.uniroma1.it/it/users/francescodellipriscoliuniroma1it
Frequency
Attendance at the lessons is not mandatory, but is strongly recommended
Exam mode
For the first part of the course the student must carry out an in-depth project ("tesina") (with personal contribution required), to be carried out at home, of one of the seminars held during the course.
Lesson mode
Traditional with possible project. The course is taught by using the blackboard and/or slides depending on the topic. If it is not possible to carry out the lessons in the classroom with all the students due to the pandemic, the lessons are carried out in mixed mode (in the classroom with streaming for students who use remotely) or exclusively in streaming if no student can attend. The instructions for the first part of the course are indicated on the course page https://corsidilaurea.uniroma1.it/it/users/francescodellipriscoliuniroma1it
  • Lesson code1041427
  • Academic year2025/2026
  • CourseControl Engineering
  • CurriculumSingle curriculum
  • Year2nd year
  • Semester2nd semester
  • SSDING-INF/04
  • CFU6