Agents and Distributed Artificial Intelligence

Code:

EIC0033

Acronym:

AIAD

Keywords
Classification	Keyword
OFICIAL	Artificial Intelligence

Instance: 2020/2021 - 1S

Active?	Yes
Responsible unit:	Department of Informatics Engineering
Course/CS Responsible:	Master in Informatics and Computing Engineering

Cycles of Study/Courses

Acronym	No. of Students	Study Plan	Curricular Years	Credits UCN	Credits ECTS	Contact hours	Total Time
MIEIC	168	Syllabus since 2009/2010	4	-	6	56	162

Teaching language

Suitable for English-speaking students

Objectives

This course has an engineering bias and proposes a global perspective on the techniques associated with agent-based computing, exploring, on one hand, agent-based complex systems modeling and simulation, and the development of intelligent agents and multi-agent system applications.

Agent-Oriented Programming and Software Engineering are presented as a new metaphor to describe and program distributed computational systems.

The acquired knowledge is consolidated through the use of appropriate software tools, with which students are incentivized to work on the development of small projects. The main goal is that students are able to specify and implement complex, adaptive, distributed, and decentralized systems using the multi-agent systems paradigm.

Learning outcomes and competences

At the end of the course, the student is expected to be able to:

• Model complex systems through computational agents
• Project distributed and decentralized systems following the multi-agent systems paradigm
• Develop intelligent software agents, using different architectures
• Evaluate the execution of an agent-based simulation or a distributed multi-agent system
• Develop multiagent systems with adaptability mechanisms

Working method

Presencial

Pre-requirements (prior knowledge) and co-requirements (common knowledge)

Knowledge of Artificial Intelligence techniques

Program

1. Agent-based computing

• Motivation and goals
• Agent-based simulation vs multi-agent systems
• Tools for MAS development and agent-based simulation: JADE and Repast

2. Intelligent agents

• Definitions
• Agents and environments
• Basic kinds of agent programs
• Agent architectures. Deductive reasoning agents. Practical reasoning: BDI agents. Reactive agents: the subsumption architecture

3. Multi-agent systems

• Definitions.
• Communication. Speech acts. ACL. FIPA standards.
• Interaction protocols. JADE.
• Agent-Oriented Software Engineering

4. Multi-agent decision making

• Game theory. Utilities and preferences. Dominant strategies. Nash equilibria and Pareto-efficiency. Uncertainty. Cooperative game theory.
• Mechanism design. Social choice theory.
• Auctions. Auction protocols. Combinatorial auctions. Double-sided auctions.
• Negotiation. Negotiation attributes. The alternating offers protocol. Negotiation tactics: time and behavior. Negotiation for task allocation. Deals and utilities. Dominance. Pareto optimal and individual rational deals. The monotonic concession protocol.  Negotiation in resource reallocation.

5. Introduction to agent-based simulation
6. Conversational agents
7. Reinforcement Learning

• Learning agents.
• Reinforcement learning. Policies, rewards, and value functions. Markov decision processes. Episodic and continuing tasks. Discounted returns. Temporal-difference learning. Q-learning.
• Deep reinforcement learning

Mandatory literature

Michael Wooldridge; An introduction to multiagent systems. ISBN: 978-0-470-51946-2

Complementary Bibliography

Shoham Yoav; Multiagent systems. ISBN: 978-0-521-89943-7
S. Russel and P. Norvig; 'Artificial Intelligence: A Modern Approach', Prentice Hall, 2003
Richard S. Sutton, Andrew G. Barto; Reinforcement Learning: An Introduction, A Bradford Book, 2018. ISBN: 978-0262039246

Teaching methods and learning activities

Oral presentation of the themes of the course in theoretical classes, with interaction with students. Tool experimentation and exploration in the practical classes.
Practical classes will be based on the supervision of assignments. Project-oriented learning.

Software

Plataforma de Sistemas Multi-Agente JADE
Ambiente de Simulação REPAST

keywords

Physical sciences > Computer science > Cybernetics > Artificial intelligence
Physical sciences > Computer science > Computer architecture > Distributed computing
Technological sciences > Engineering > Simulation engineering

Evaluation Type

Distributed evaluation with final exam

Assessment Components

Designation	Weight (%)
Exame	50,00
Trabalho laboratorial	50,00
Total:	100,00

Amount of time allocated to each course unit

Designation	Time (hours)
Estudo autónomo	44,00
Frequência das aulas	52,00
Trabalho laboratorial	66,00
Total:	162,00

Eligibility for exams

Enrolled students are admitted to the exam if they do not exceed the allowed number of non-attendance to lab classes (maximum 25% of non-attendance).

Calculation formula of final grade

Final Grade = 50% * Assignment + 50% * Exam
Assignment = 70% * Assigment1 + 30% * Assignment2

In order to obtain approval in the course, the following minimum scores apply:
- Assignment1: 8.0 values out of 20
- Assignment2: 8.0 values out of 20
- Exam: 7.0 values out of 20

Examinations or Special Assignments

Evaluation in special seasons consists of two practical Assignments and a written Exam, where each of these components weights 50% on the final grade. Approval in the course demands for a minimum score of 8.0 out of 20 in each of the practical assignments, and a minimum of 7.0 out of 20 in the written exam.

Special assessment (TE, DA, ...)

All assessment components are required for all students. Students enrolled using special frequency modes, without obligation to attend to the practical classes, must arrange with teachers appropriate consultation and evaluation sessions.

Classification improvement