CS116 -- Artificial Intelligence
Fall, 2006
Autonomous MazeWar Controller
(last updated 11/9/2006)

Design and implement a program that controls a rat in a maze.  The rat you control will be interacting with a rational reconstruction of the historic (and first!) multi-player first-person shooter game.  You can find external information on MazeWar via the link here and more specific information about the local implementation here.

NOTE: make sure your code will compile and run on my system.  If you have any doubts, check with me in advance.

Similar to your first assignment where you implemented a stimulus-response agent in the MÆDEN world, you will write a controller for an rat in a maze.  However, in the MazeWar environment, there are multiple agents (some controlled by AI and some by human players), and the world changes asynchronously.  That is, your agents do not have unlimited time to think about what to do next.  Fortunately (or unfortunately), you are not required to use an S-R architecture for this controller.  You may use any of the techniques we have covered in class; if you want to use yet another technique we have not discussed, talk it over with me in office hours.

Initialization.  Upon connection to the MazeWar game, your agent will receive the global maze information and the controlled rat's specific information.  Then your program will take over and send movement commands and receive sensory information.  A complete game-server communication protocol will be provided on 11/16.  You may assume that you will be given the complete maze specification (number of rows and columns and effectively a two-dimensional array of chars where '1' signifies a wall and '0' an open cell).  You will also be given your rat's row and column location and a heading (one of the four points of the compass). 

Sensory Information.  When another rat is visible in the hall into which your rat is facing, you will receive that information.  When your rat is tagged, you will be notified and relocated to another place in the maze (of which you will be informed).  Rats will be identifiable by name and by controller type.

Again, your agent will receive a message from the environment whenever another rat is visible in the maze based on your rat's current heading.  However, just to make things interesting, from time to time the autonomously controlled rats from CS116 will receive smell information about the human-controlled rats.  This smell information will be quite specific, giving your rat exact relative location information (row and column deltas) to each of the human-controlled rats.  But of course, you should not design your agent to rely on this information.  This information may come in gusts, intermittently, constantly, or not at all.  And smell will not provide any information about other AI rats.

Effector Commands.  Your rat can move about in the maze and tag other rats using the following character commands:
'w': move forward one space
's': move back one space
'a': turnleft
'd': turnright
' ' (space): shoot
Note, when you fire a projectile to tag another rat, your projectile travels down the corridor over time, approximately one cell per 100 milliseconds.  Also, rats may only fire one projectile per second.

You should include some display capability of your rat's situation in the maze; at a minimum, I will expect a curses-based overhead representation depicting your rat's location and orientation.  However, since displays are not the point of the exercise, I will provide Java code that displays a first-person perspective view of the maze and an overhead view.  Other rats are visible in the perspective view but only your rat is visible in the overhead view.  (You are not required to use this code nor to use Java, but you are welcome to take advantage of this so that you can focus on the AI part of the project.)

In a similar vein, I will provide Java code that interacts with the game server.  This code will not be available until 11/16.  Again, you do not need to use this code or the Java language.  You can receive and send all your information over sockets using your favorite language according to the protocol that will be fixed on 11/16.

The goals of this assignment are:
  1. wrap up the semester with a fun game project
  2. experience the challenge of designing an AI in an asynchronous environment
  3. build a controller for an agent in a multi-agent system
  4. have fun and and try to beat the CS030 players.
Submission and grading
Submit your completed search agents to Eureka in the usual way (zip or tar file).  Include a readme file that explains how to start your rat controller, etc.  I will grade this project based on (a) code completeness and correctness, (b) coding style including documentation, and (c) classification accuracy.