G22.3033-012

Autonomous Multiagent Systems

Peter Stone

Graduate Division, Computer Science

Tuesdays, 5pm-7pm in WWH 101


 

Overview

There is no generally accepted definition of artificial intelligence "agents." But practitioners know them when they see them. In loose terms, agents are programs that (i) sense their environment, (ii) make decisions about how to act based on these sensations, and (iii) then execute these actions. Autonomous agents do all three of these steps on their own, i.e. without a human in the loop. Multiagent systems are collections of multiple agents that interact with one another.

This course provides a broad introduction to autonomous agents with an emphasis on multiagent systems. Topics include

  • agent architectures
  • inter-agent communication
  • teamwork
  • distributed rational decision making
  • agent modeling
  • multiagent learning
  • It is a programming-intensive course by the end of which, the student will have implemented a full-fledged team of autonomous agents in the RoboCup soccer simulator. At the end of the course, students' teams of soccer-playing agent will compete against each other in a tournament.

    Class Prerequisites

    Some background in artificial intelligence is recommended, but not required.
    Good programming skills, preferably in C++ or Java, are required.

    Class Requirements

    Students will be expected to attend and participate in the class as well as complete assignments in a timely fashion. Grades will be determined based on (proportions are tentative):

  • Written assigments (20%)
  • Class participation (20%)
  • Initial programming assignments (20%)
  • Final programming project (competition) and report (40%)
  • Assignments

    Reading, written, and programming assignments will be updated on the assignments page.
    You can go directly to the final project page.

    Mailing List

    Please subscribe to the class mailing list to receive relevant announcements.

    Office Hour

    You can find the instructor in WWH 401 from 4pm to 5pm on Tuesdays, immediately before class. The phone number there is 998-3081.

    Syllabus

    A preliminary version of the syllabus is available here (postscript).
    The course textbook is Multiagent Systems: A Modern Approach to Distributed Artificial Intelligence, edited by Gerhard Weiss.
    Selected readings from this text will be assigned, along with relevant research papers.

    Slides

    Here are some of the slides used in class.

  • Week 1
  • Week 3
  • Week 4, some more
  • Week 5, some more
  • Week 6
  • Week 7
  • Week 8, some more
  • Week 9, some more
  • Week 10, some more
  • Week 11
  • Week 12, some more, still more
  • Week 13
  • Week 14 (the tournament) and the tournament logfiles (file names match the letters of the teams from the slides).
  • Relevant Links

  • If you would like an overview of artificial intelligence prior to the course, I recommend Artificial Intelligence: A Modern Approach by Stuart Russell and Peter Norvig.
  • RoboCup, the robot soccer world cup.
  • The RoboCup Soccer Server System used as the simulator for the programming assignments in this course. Download the manual from here.

  • [Back to Department Homepage]

    Page maintained by Peter Stone
    Questions? Send me mail