Course teached as: B028335 - SISTEMI MULTIAGENTE Second Cycle Degree in ELECTRICAL AND AUTOMATION ENGINEERING Curriculum INGEGNERIA ELETTRICA
Teaching Language
Italian
Course Content
The covered topics include: intelligent agents and multi-agent system models, distributed control and data processing algorithms, reinforcement learning, applications (robot navigation, sensor networks, multi-robot systems, formation control, distributed data processing and machine learning, etc. ).
Learning Objectives
The course aims to provide methodologies for modeling, analyzing, and designing intelligent agents able to carry out complex tasks, taking into account:
both physical and virtual agents (autonomous vehicles, robots, sensors, processing units, pieces of software); interconnected, possibly heterogeneous, agents;
decentralized architectures.
Course program
1. INTRODUCTION TO MULTI-AGENT SYSTEMS
What is a multi-agent system? Cooperative vs competitive multi-agent systems. Examples of multi-agent systems in science and engineering. Motivating problems.
2. ELEMENTS OF GRAPH THEORY
Connectivity of a graph. The Laplacian of a graph: properties and applications (graph partinioning and spectral clustering).
3. INTELLIGENT AGENTS
Paradigms of intelligent agents. Planning vs behaviors. Application to robot motion planning and navigation. PRM and RRT algorithms. Mobile robot models. Artificial potential fields.
4. SYNCHRONIZATION AND COORDINATION IN MULTI-AGENT SYSTEMS
Consensus for undirected and directed graphs. Applications (social networks, disrtibuted computing). Artificial potential fields for synchronization and coordination.
5. MULTI-ROBOT SYSTEMS
Coordination algorithms for rendez-vous, formation control, and flocking problems. Connectivity maintenance and collision avoidance. Covering. Mapping and exploration.
6. MULTI-AGENT OPTIMIZATION AND LEARNING
Distributed regression over networks. Distributed linear least squares. Distributed optimization over networks.
7. MULTI-AGENT INFORMATION FUSION
Centralized and distributed information fusion. Application to sensor networks and distributed estimation.
8. REINFORCEMENT LEARNING
Markov Decision Processes. Stochastic dynamic programming. Value and policy iteration. Reinforcement learning: time-difference, Monte-Carlo method, Q-learning, policy gradient. Exploration vs exploitation. Multi-agent reinforcement learning.