Professor Les Gasser
Graduate School of Library and Information Science
AIMS OF THE COURSE:
1. To gain a picture of contemporary thinking about agents and especially multi-agent systems, including principles and theory, common practices, and application areas, current and emerging.
2. To develop some concrete acccomplishments in the agents/multi-agent systems arena through a project.
3. To gain experience and confidence in understanding a new and rapidly evolving information technology area.
PROCESSES AND PROCEDURES:
The course activities will consist of:
Reading: one textbook, and selected additional papers
Discussion: of readings and presented materials.
Projects: building and/or experimenting with agents and multi-agent systems; documenting research and assessment work.
Presentation: of reading materials and class projects
There is one required textbook for this class, listed below. There will also be readings from online sources (typically, downloadable papers or papers readable online as HTML or PDF).
Intelligent Information Agents
Edited by Matthias Klusch
(Available in some of the usual places; BestBookBuys.com and BookPool.com are online sources.)
Additional readings will be online and/or downloadable, so I'll assume you have both the technical ability and the facilities to access and read web-based materials in HTML, Postscript, PDF, and/or text form as needed, including possibly compressed, zipped, or gzipped versions.
You should read all the assigned material before the class in the week it's listed below, so as to be best prepared for class discussion of that material.
For Week 2, the reading will be heavy---some 200 pages or so of introductory technical, though not mathematical, material. This is because we have to establish some common baselines to build upon. We'll talk over some strategies for dealing with the load in class, and the amount of reading will be much lighter for the rest of the semester.
Participation in discussion is truly important - we're a short-term 'community of scholars.' Uncovering and sharing your own insights as a member of this community are one of the key reasons we're here.
There are three general options for projects in this class: research/writing projects, system-building/experimentation projects, and hybrid projects. We'll discuss project options in more detail in Week 3.
I'd like you to select a project topic and approach, and submit a short proposal for my approval by Week 5 . The short proposal should be less than a page, and should include the proposed subject of your project, a description of the type of the project (writing, system-building, or hybrid), a brief description of what you propose to do or write about, and a very preliminary collection of resources (papers, projects, tools, that you'll use. Please submit proposals electronically to "firstname.lastname@example.org" in plain ascii text format (not an attachment).
Project drafts will be reviewed if they are received by Week 10 (28 March).
Writing project options:
Analyze an existing Agents/MAS concept historically---longitudinally---showing the development of ideas across projects and linking the concepts to particular projects and people in the field.
Analyze a particular existing Agents/MAS project/team (a group of people at some research center) historically, i.e. longitudinally, over some period significant to the project, showing the development of the ideas within the project and linking these ideas to other develpments in the field.
Build a detailed survey review and critique of a set of core ideas in the Agents/MAS world. Such a survey would include a) a set of desiderata or success criteria, b) an analysis of how core idea(s) respond to these criteria, c) an evaluation of the approach, and d) suggestions for how things could be improved in the future (e.g., future research). This could be done progressively, by writing detailed reviews of three or four related articles, collecting them, then writing a section that ties the reviews together.
System-building/experimentation projects options:
Build a (component of an) agent-oriented or multi-agent oriented system and experiment with its behavior with the aim of illustrating some novel functionality and/or characterizing some aspect systematically. Demonstrate your accomplishment to the class, and write about the principles it illustrates in a short paper.
Create a detailed design for a component of an agent-oriented or multi-agent system that has some novel aspect, and provide a written rationale for its structure and functionality, with reference to an existing body of research.
Explore an existing testbed or demonstration project and link it to existing ideas in the literature, analyzing its foundations, strengths, and weaknesses. (DECAF, JATLITE, JAFMAS, JAM, etc.)
If you're interested in pursuing a system-building or hybrid project, please see me ASAP to discuss your ideas. If you are looking for projects of this type, I have a long list of sample project suggestions, including extensions and enhancements of existing MAS simulations and systems, to fit your interests and skills.
There are several terrific online repositories or "portals" containing information, pointers, news, and resources on agents and multi-agent systems. These are:
The UMBC AgentsWeb: http://agents.umbc.edu/
The European AgentLink Network of Excellence
ORGANIZATIONAL, ADMINISTRATIVE, AND GRADING ISSUES:
Regular attendance, reading assigned materials and being well prepared to discuss them, and class discussion are required.
Grades for the course will be assigned based on the following work:
(30%) Class preparation and participation
(30%) Reaction Papers: Please turn in four reaction papers throughout the semester, discussing some aspect of the readings for the course, or discussing a related paper that you've found. (Such papers should be OK'ed by me first)
(40%) Project and its presentation
Any written work you present will be assessed based on the following:
Have relevant sources been consulted?
Are the main concepts understood?
Is the analysis reasonably complete?
Is the writing articulate and are the ideas presented clearly?
Projects will be evaluated on:
Thoroughness of the approach
Coverage of material and completeness of the work
Depth of insight
Clarity of the material presented
Late work will incur a serious penalty without prior approval (which will be granted only rarely ).
General Issues and Principles
Week 1: 17 January Introduction to Agents and Multi-Agent Systems
Overview of the course, requirements, processes, etc.
Background on the origins of the course
Overview of resources available.
History of the field, motivations, and examples
Overview of basic theory that underpins the semester:
--Systematic representation of states and knowledge
--Heuristic action and reasoning
Class 1 lecture notes
Week 2: 24 January Perspectives on Agents and Agent Applications
Definitions of "Agents" and "Agency" and implications for agent design
Desiderata for agent application
Principal "Common" and emerging agent application areas
Discussions of readings on agent applications and agent architectures.
Klusch book: Preface, Introduction, and Chapter 1 (Papazoglou and van den Heuvel)
H. Van Parunak, "Practical and Industrial Applications of Agent-Based Systems," 1998 Online at: http://agents.umbc.edu/papers/apps98.pdf.
Nicholas R. Jennings and Michael J. Wooldridge, Applications Of Intelligent Agents, in Nicholas R. Jennings and Michael J. Wooldridge (Ed.), Agent Technology Foundations, Applications, and Markets , Springer-Verlag, 1998. Online at: http://agents.umbc.edu/introduction/jennings98.pdf
Stan Franklin and Art Graesser, "Is it an Agent, or just a Program?: A Taxonomy for Autonomous Agents", http://www.msci.memphis.edu/~franklin/AgentProg.html
Hyacinth S. Nwana, Software Agents: An Overview. http://www.cs.umbc.edu/agents/introduction/ao/
L. Gasser, "Agents and Concurrent Objects," IEEE Concurrency, 6(4) pp. 74-77&81, October-December, 1998. (Interviewed by J-P. Briot). (HTML)
UMBC AgentsWeb Applications Page
From Agents to Multi-Agent Systems slides
Week 3: 31 January Agent Architectures: Blackboards; Belief-Desire-Intention (BDI) models; Others
Lecture/Discussion:Search as a fundamental architectural issue Blackboard Architectures BDI Architectures Relationships between BB and BDI architectures
K. Pfleger & B. Hayes-Roth. An Introduction to Blackboard-Style Systems Organization. Knowledge Systems Laboratory, 1998 [Also available here (.pdf)].
BB1/Guardian architecture slide - a Blackboard derivative.
Anand Rao and Michael Georgeff, "BDI Agents: From Theory to Practice", Australian AI Institute Technical Note 56, 1995. Online here in PDF.
"The Evolution of Blackboard Control Architectures," N. Carver and V. Lesser, Expert Systems with Applications, Special Issue on The Blackboard Paradigm and Its Applications, vol. 7, no. 1, 1--30, 1994. You should definitely read the sections included here (pgs 1-19 + conclusions) (in .pdf). The entire paper is here: (in gzipped PostScript): http://www.cs.siu.edu/~carver/ps-files/tr92-71.ps.gz.
Lecture Slides:[To appear]
Week 4: 7 February Inter-Agent Communication and Interaction: KQML, FIPA, FLBC, etc.
Lecture/Discussion:Issues in Agent Communication Languages
Speech Act Theory and Speech Acts
Distributed Semantics problems
Singh's Critiques and solutions
Adaptive Language alternatives and their generalizations
Agent Communication Languages: The Current Landscape, by Labrou, Finin and Peng, IEEE Intelligent Systems, March/April 1999, pp. 45-52. Online at: http://www.csee.umbc.edu/~nicholas/courses/691d/papers/ieeeIntelligentSystems1999.pdf.
Munindar P. Singh, "Agent Communication Languages: Rethinking the Principles". IEEE Computer, volume 31, number 12, December 1998, pages 40-47.
Strongly RecommendedScott A. Moore, "KQML and FLBC: Contrasting Agent Communication Languages", in Proc. 32nd Hawaii International Conference on System Sciences, 1999. Online here in .pdf
Lecture Slides:[To appear]
The remainder will appear shortly...