B-KUL-H02H4A Multi-Agent Systems

Show all details |  Hide all details

General information

Academic year: 2011-2012 Study points: 4 Language: English Difficulty: Advanced Duration: 81.0 hours  Periodicity: Taught in the second semester POC: POC Artificial Intelligence This course cannot be followed within the context of an exam contract

Print version

Taught by

Holvoet Tom

Aims

This course studies the research area of 'software engineering using multi-agent systems'.  Multi-agent systems are particularly interesting for modeling and developing a wide range of distributed applications, including internet applications, distributed control systems, robotics, and self-managing systems.  The course includes discussions of software architecture, models, typical issues, concepts and techniques that relate to software engineering using multi-agent systems.
 
In particular, the course aims to:
 
- provide an introduction and overview of software models and techniques in building multi-agent systems;
- provide a general perspective on the domain of collective and cooperative behaviour, with a focus on reactive and behaviour-based agents;
- provide a conceptual framework for distributed problem solving, based on recent research in this area;
- provide the participants with the knowledge and key issues, challenges and problems for assessing the usefulness of employing multi-agent systems in software engineering. 
 
Aspects that are prominent in this course are: software architectures for describing the behaviour of individual agents (reactive, BDI, behaviour-based), collective and cooperative behaviour, actions and interactions, cooperation / coordination / communication.

Previous knowledge

As the course includes programming exercises and a project, knowledge of an object-oriented programming language (e.g. in Java, C++, …) is required.

Content

The content of the course is based on
- selected chapters of the most-referred books on multi-agent systems
- selected papers from the research domain (core references)
- selected papers on the research of the department of computer science in this area

Content
Agents and multi-agent systems
- agents, multi-agent systems
- decentralized systems vs. distributed experts vs. algorithmic techniques
- application areas
- agents vs. objects, agents vs. expert systems
- historical overview
 
MAS in software engineering
- MAS as software architecture
- MAS in UP (the universal process)
- agent-oriented software engineering (incl. GAiA)
 
Agent behaviour architectures
- deliberative
- reactive and behaviour-based (incl. Brooks’ subsumption architecture, Agent network architecture)
- rational (belief-desire-intention  -  BDI)
- horizontally and vertically layered
 
Agent actions and interactions
- action models
- interactions
- cooperation / coordination
- introduction to game theory
 
Agent communication
- direct vs. indirect communication
- agent communication languages
 
Swarm Intelligence
- emergent behaviour, self-organization
- swarm intelligence
- ant-based algorithms
- dynamic task selection
 
Case Studies
- a selection of cases studies is discussed (e.g. manufacturing control, P2P, AGV control, …), based on literature conducted at the department of computer science
 
MAS Evaluation
- pitfalls of developing multi-agent systems (based on literature by M.Wooldridge and N.Jennings)
 
Software
- a selection of software platforms is discussed (e.g. Jason / RePast / …)

This course is included in

Master of Science in Artificial Intelligence   (Option Engineering and Computer Science (ECS)) (Option Cognitive Science (CS))
Master of Science in Advanced Business Studies  
Master of Science in de ingenieurswetenschappen: computerwetenschappen (geen nieuwe inschrijvingen in 2011-2012)   (Artificiële intelligentie)
Master of Science in de ingenieurswetenschappen: bouwkunde  
Master of Science in de toegepaste economische wetenschappen: handelsingenieur in de beleidsinformatica  
Study Abroad Programme in European Culture and Society (PECS)  
Master of Science in de ingenieurswetenschappen: verkeer, logistiek en intelligente transportsystemen  
Master of Science in Artificial Intelligence   (Option Speech and Language Technology (SLT))
Master of Science in de ingenieurswetenschappen: computerwetenschappen (nieuw programma, start in 2010)   (Hoofdspecialisatie Gedistribueerde systemen) (Verplicht)   (Hoofdspecialisatie Artificiële intelligentie)
Master of Science in de informatica (uitdovend, enkel 2e fase)   (Specialisatie computernetwerken en gedistribueerde systemen) (Specialisatie artificiële intelligentie)

Course Material

Articles and literature
Slides, transparencies, courseware

Prerequisites

This course is a prerequisite for the following courses:
H04G7A:  Capita selecta computerwetenschappen: Gedistribueerde systemen

Activities

		B-KUL-H00I2a Multi-Agent Systems: Exercises
	General information Study points: 0.50 Language: English Category: Exercises Duration: 10.0 hours  Periodicity: Taught in the second semester POC: POC Artificial Intelligence Taught by Holvoet Tom Aims The aim of the exercises of this course is to gain experience with a limited set of software environments that support the development and/or execution of multi-agent systems.
		B-KUL-H02H4a Multi-Agent Systems: Lecture
	General information Study points: 1.50 Language: English Category: Lectures Duration: 16.0 hours  Periodicity: Taught in the second semester POC: POC Artificial Intelligence Taught by Holvoet Tom
		B-KUL-H08M2a Multi-Agent Systems: Project
	General information Study points: 2.00 Language: English Category: Assignments Duration: 55.0 hours  Periodicity: Taught in the second semester POC: POC Artificial Intelligence Taught by Holvoet Tom Aims Typical issues and problems in defining a MAS are further addressed in a project. In this project, a multi-agent system is to be developed with a particular focus on coordination of the autonomous entities and its effect on the agent architectures. The project aims to address typical multi-agent concepts and techniques, and critically assess MAS-based solutions in terms of non-functional qualities (e.g. quality of the solution, scalability, flexibility under dynamic circumstances, ...), based on experimental evaluation (e.g. in a MAS simulation environment).

Evaluation

		B-KUL-H22H4a Evaluation : Multi-Agent Systems
	Evaluation description Examination type:     oral with written preparation When?: interim evaluations plus final examination during examination period Evaluation type:     Report     Presentation Explanation The evaluation of this course is based on a research project. The project typically includes a literature study, limited software development based on a multi-agent software architecture, and a criticial, scientific evaluation of the proposed architecture. Students work on the project in teams (2 students per team). The project description is distributed after the 4th week of the semester. Students have appr. 10-12 weeks for working on the project. The course evaluation is based on the submitted solution of the project and on the oral presentation and discussion.