Ing Ind - Inf (Mag.)(ord. 270) - MI (263) MUSIC AND ACOUSTIC ENGINEERING

*

A

ZZZZ

089169 - AUTONOMOUS AGENTS AND MULTIAGENT SYSTEMS

Ing Ind - Inf (Mag.)(ord. 270) - MI (473) AUTOMATION AND CONTROL ENGINEERING - INGEGNERIA DELL'AUTOMAZIONE

*

A

ZZZZ

089169 - AUTONOMOUS AGENTS AND MULTIAGENT SYSTEMS

Ing Ind - Inf (Mag.)(ord. 270) - MI (481) COMPUTER SCIENCE AND ENGINEERING - INGEGNERIA INFORMATICA

*

A

ZZZZ

089169 - AUTONOMOUS AGENTS AND MULTIAGENT SYSTEMS

Obiettivi dell'insegnamento

Computer systems are increasingly distributed and interconnected. This trend naturally leads to the development of systems composed of autonomous decision-making entities, called autonomous agents, that interact with each other in complex environments. Agent-based systems are an abstraction of specific computing systems deployed in several applications, including electronic commerce, control of industrial processes, logistics, ambient intelligence, web services, robotics, space systems, and modeling of complex systems.

This course aims at presenting general techniques for developing multiagent systems, independently of the applicative domains. In particular, the course will present methods for developing single agents, able to make rational decisions in situations affected by uncertainty, and for developing systems composed of multiple agents, with special emphasis on the interaction mechanisms between the agents. Moreover, some real-world applications of agent systems will be discussed. At the end of the course, students will acquire the ability to design and develop distributed systems based on the agent paradigm.

Risultati di apprendimento attesi

Dublin descriptors

Expected learning outcomes

Knowledge and understanding

Students will learn:

the models that represent distributed systems as collections of computational systems, called autonomous agents, that interact together,

the models that represent individual autonomous agents as rational utility maximizers,

some significant types of interaction mechanisms between multiple autonomous agents (negotiation, voting, auctions, coalition formation, planning, distributed constraint optimization, learning, ...),

some of the relevant algorithms available to implement the interaction mechanisms and their properties (range of application, optimality guarantees, ...).

Applying knowledge and understanding

Given application cases that involve multiple decision-makers, students will be able to:

build agent-based models and systems that are based on the identification of individual decision-making entities (autonomous agents) and on the identification of the interaction mechanisms between them,

analyze and understand existing multiagent systems.

Lifelong learning skills

Students will be able to understand and critically evaluate the modeling based on multiagent systems composed of multiple decision-makers.

Students will be able to understand, learn, and develop new interaction mechanisms for autonomous agents.

Students will be able to apply acquired knowledge when designing complex distributed computational systems.

Argomenti trattati

1. Introduction to the concepts of autonomous agents and of multiagent systems.

2. Autonomous agents as rational decision makers: architecture for intelligent agents, Markov decision processes.

3. Interactions between self-interested agents: short introduction to game theory, negotiations, voting mechanisms, auctions, coalition formation.

5. Multiagent learning: multiagent Markov decision processes and stochastic games, evolutionary game theory.

6. Examples of real-world applications of agent-based systems.

Prerequisiti

Students are required to know basic notions of computer science (programming), algebra, and probability.

Modalità di valutazione

The students' assessment is based on a 1.5-hour closed-book written test at the end of the course, consisting in exercises and questions on all the topics of the course. The test assigns a maximum of 31 points (30 cum laude is assigned when the total score is strictly larger than 30 points). Students can take the written test at any exam session during the year.

Type of assessment

Description

Dublin descriptor

Written test

Solution of numerical problems

Application of algorithms to given problem instances.

Exercises focusing on design aspects

Designing models of autonomous agents and of their interaction mechanisms for simple problems

Theoretical questions on all the course topics with open answer