CS353 - Artificial Intelligence, Project 4 (Final Project)

Colby College, Fall 1998, Randolph M. Jones

Soccer Agent

Initial report due Monday, Nov. 16

Program implementation due Friday, Dec. 4

Final report, code and documentation due Monday, Dec. 7

In this project, you will create a team of four intelligent agents to play a simulated soccer game. The main points of the exercise are to:

Create intelligent agents for a somewhat complex, interactive environment
Build a team of agents that share information and goals
Focus on agents that must deal with other cooperative and antagonistic agents
Implement methods for making informed guesses about the state of the world and choosing the best course of action given limited and imprecise information
Translate what you have learned in this course into a relatively sophisticated computer program
Kick some polar bear butt

The activities you will engage in

Play the game with some example agents, study the game rules, and brainstorm about potential game-winning strategies, goals, and methods for the agents
Create an initial report, describing some of the strengths and weaknesses you found in the example agents, and a general description (as technical as possible) of the strategies, goals, and operators you would like to include in your agents
Create a well documented team of soccer agents, and try to win a competition that includes the agents developed by this class and agents developed by the AI class at Bowdoin College.
Write a technical document that presents the design of your agents at very general and very detailed levels
Write a final report that discusses your team of agents as AI systems, and particularly relates how you used what you have learned about AI to design and build your system

What you will turn in

Due at the beginning of class on Monday, November 16
- An initial report that describes your experiences playing with the example soccer agents, thinking about how the game works, and initial design thoughts for your team of agents
  - Discuss the strengths and weaknesses of the example agents, and how you will use those to influence the design of your own team.
  - Describe the features of the game that seem to be key to scoring goals, defending your own goal, and winning the game in general
  - Work out as much as possible the general game strategies your team will adopt
  - Work out the different states you anticipate your team's agents to be in, and the goals and operators they will use in each state
  - Work out how your agents will cooperate with each other, how they will attempt to defeat the opponents' goals, and in what ways (if any) your agents will learn
  - Describe all of these design decisions at a high level, and also (where appropriate) in as much detail as you can
Due on Friday, December 4, in time for the contest
- Your working agent code
Due on Monday, December 7, at 4:00 PM
- Your well commented agent code (which should not be substantially different from the code you used in the contest)
- Any enlightening documentation of your code (including descriptions of anything you changed after the contest)
- A final technical report with the same format as the initial report, but reflecting the design of your final agent code
- A final report that discusses your team of agents as AI systems
  - Focus on the concepts we have covered in class, including search, goals, operators, pattern-matching, knowledge representation, reactivity, planning, learning, constraints, etc.
  - Also describe in detail how you used what you have learned about AI (both in this class and anywhere else) to design and build your system

Grading

100 points will be divided as follows:
- 25 points for the initial report
- 20 points for the commented code and supporting documentation
- 20 points for the final technical report
- 25 points for the final report
- 10 points to reward your performance in the contest (the better your agents do, the more you will get)
You will be graded on the quality of writing in your reports, as well as the quality of the insights and analysis you provide. Bland reports that shallowly summarize what you did will not receive full points.
The code must be clear and documented. The harder it is for me to figure out what you have done, the less likely you are to get full points.

Technical details

You can write your agents in either Tcl or C, using the example agents as a template. I recommend using Tcl unless you have access to a machine running Linux. However, if your agents are going to be doing quite a bit of information processing, you might need to use C for speed.
There will be an executable version of the game and example agents in the Tcl Soccer folder on my Macintosh server, in the same place as the other project code for this class.
As always, if you want to set the engine up on some other machine, talk to me and we will figure out how.

The soccer package consists of the following files. Put the soccer.tcl file into the folder or directory from which you will run the program. You will also need to create a teams folder or directory to hold the code for the example teams and your own team. The folder/directory structure should look like this:

   tclsoccer/ (main Tcl Soccer folder)
         TclSoccerForMac (Copy of wish executable to run Soccer under MacOS)
         soccer.tcl (Main soccer game code in Tcl/Tk)
         doc/ 
            INTERFACE-FOR-C
            INTERFACE-FOR-TCL
            RULES
         include/ 
            east.h (header file for compiling east team in C)
            west.h (header file for compiling east team in C)
            tcl.h (header file for compiling tcl-compatible code in C)
            tclMac.h (header file for compiling tcl-compatible code in C under MacOS)
            interface.c (interface code for linking teams written in C to Tcl Soccer)
         lib/ 
            eastwinlink.txt (linking directives to compile east team in CodeWarrior under Windows)
            westwinlink.txt (linking directives to compile west team in CodeWarrior under Windows)
            Tcl8.0.shlb (Tcl library for linking C code under MacOS)
            tcl80vc.lib (Tcl library for linking C code under Windows)
         teams/ 
            cexample/
               Makefile (for compiling under Unix)
               CWProject.mcp (for compiling with CodeWarrior under MacOS or Windows)
               main.c
               east.so (binary for Linux)
               west.so (binary for Linux)
               east.shlb (binary for MacOS)
               west.shlb (binary for MacOS)
               east.dll (binary for Windows)
               west.dll (binary for Windows)
            crollers/
               Makefile (for compiling under Unix)
               CWProject.mcp (for compiling with CodeWarrior under MacOS or Windows)
               main.c
               east.so (binary for Linux)
               west.so (binary for Linux)
               east.shlb (binary for MacOS)
               west.shlb (binary for MacOS)
               east.dll (binary for Windows)
               west.dll (binary for Windows)
            defenders/
               main.tcl
            example/
               main.tcl
            rollers/
               main.tcl

You can also download tclsoccer.zip, a zipped archive of all of the above files
You must have already installed Tcl and Tk 8.0 for any of this to work.
Note also that the C agents currently only work on Unix machines, but I am working on that (I welcome help).

Administrative statement

This is a group project. I encourage you to talk to others about the general nature of the project and ideas about how to pursue it. However, the technical work, the writing, and the inspiration behind these must belong substantially to the two group members. If any person besides the group members has contributed in any significant way to the project, you must credit their work in your writeups and your code, so I know which parts to grade you on.

Randolph M. Jones (rjones@colby.edu)

This page is designed to be viewed with any browser