Designing Team-based Agents2005HT
|
|
Course plan
No of lectures
8 (16 hours)
Recommended for
Doctoral and master-level students in computer science or artificial
intelligence
The course was last given
New course
Goals
- Learn about the hopes and promise for team-based agents --
and current and potential areas of application
- Learn about the motivation for, and status of, current R&D in the
area of team-based software and hardware agents
- Learn about the current "hard problems" in the field
- Learn various approaches, techniques, and algorithms being used to
address those problems
- Apply this knowledge in the design and development of a small
project that involves implementing a system of team-based agents
Prerequisites
Programming experience (note: extensive background in artificial
intelligence is not a requirement)
Organization
The course is organized as a series of discussions, design sessions,
and small, weekly "deliverables." The course will meet once a week for
8 weeks -- and once at the end of the quarter for final presentations
of student projects.
Contents
This is a project-oriented course for computer scientists interested
in exploring, understanding, and developing team-based agent
implementations.
Team-based AI is an emerging area of research and development. One of
the most visible examples of this is the annual RoboCup tournament,
that highlights progress towards the goal to develop a team of fully
autonomous humanoid robots that can win against the human world soccer
champion team. Other examples of mult-agent collaboration arise in
distributed-agent systems for such things as distributed
problem-solving, simulation, and control.
In addition to the various technical challenges, there are interesting
problems related to the modeling and implementation of agent
coordination and cooperation. These include such things as
agent-autonomy; centralized versus distributed management; adaptive
versus planned responses; chains of communication versus chains of
command, and the like.
This course will examine relevant literature, implementation
techniques and example systems. Topics will include:
- Different design and implementation algorithms, techniques,
and perspectives
- Insights from work on distributed/coordination systems,
concurrent programming-language design, and Artificial Life
- Studies of human team-work and proposals for team-work "best
practices"
- Issues that arise in particular domains, activities, and types
of teams
Students will be asked to form teams and develop working
implementations that demonstrate the application of key concepts,
techniques, and mechanisms covered in the course.
Literature
Readings will be short and distributed as needed. Note that
"readings" for this course may also include "using a number of
illustrative software systems."
Lecturers
Kevin McGee
Examiner
Kevin McGee
Examination
Active participation, regular progress on final project (weekly
deliverables), and a public presentation of a completed final project.
Credit
5 points
Comments
Course size is limited to 16 participants. Course language is English.
For final projects, students are free to choose any programming
language -- Scheme, Lisp, C, C++, Java, Ruby, C#, Haskel, Oz,
Assembler, etc. -- but a project must be an implementation that works
"for real" in some significant sense. It cannot be a "mock up" or
"powerpoint proposal."
Page responsible: Director of Graduate Studies