Competitions

The conference has a rich competition program with six major events:

This page will be updated regularly as new competitions are announced.

The 2K BotPrize

The aim of the contest is to see if a computer game playing bot can play like a human. In the contest, bots try to convince a panel of expert judges that they are actually human players.

Computers are superbly fast and accurate at playing games, but can they be programmed to be more fun to play – to play like you and me? People like to play against opponents who are like themselves – opponents with personality, who can surprise, who sometimes make mistakes, yet don’t blindly make the same mistakes over and over. Can a computer be programmed to seem to have personality, falibility and cunning?

The first contest was help in Perth, Western Australia in December 2008. Although none of the competitors was able to fool 4 out of 5 judges and take the major prize, some did fool 2 out of 5! You can read about the 2008 contest here.

As they did in 2008, Game Development Studio 2K Australia is offering a prize of A$7,000 cash plus a trip to their studio in Canberra for anyone who can create a bot to pass this “Turing Test for Bots”.

Further details are available from the 2K BotPrize webpage

Defcon

The aim of the CIG 2009 Defcon AI competition is to see which bot is the best Defcon player.

Format: Each bot plays 30 matches against each other bot, in a series of one-on-one matches with bots playing in each of the possible starting territory configurations. The game will run on a limited-information mode (i.e., bots cannot see units hidden by the fog of war) and the victory timer will start no later than 4 hours into the game (additionally to the usual game ending conditions). For each match, the resulting score (standard rules: 2*kills – own casualties) is recorded for each player. In the end, the player with the highest cumulative score wins.

Screenshots are available at http://www.introversion.co.uk/defcon/about/screenshots.html.

The Defcon API has been developed by Imperial College and Introversion Software to allow easy access to all the functions and data required to write a fully functional AI for Introversion’s thermonuclear war simulation Defcon. Interfaces to C++, Java and LUA are available, and the original implementation of the Defcon AI is provided as a starting point. The API is available for free from
http://www.doc.ic.ac.uk/~rb1006/projects:api,
and it works with the demo version of Defcon, which can be downloaded from http://www.introversion.co.uk/defcon/downloads/ .

Ms Pac-Man versus Ghost-Team Competition

Ms. Pac-Man is a challenging and interesting game that is usually viewed form the perspective of the Ms. Pac-Man agent. The game is also provides an excellent environment for testing multi-agent strategies for controlling the team of ghosts, with the aim either of optimising the playability of the game, or simply to minimise the score obtained by the agent. It is this latter more measurable objective that provides the focus of this competition. For this purpose an efficient simulator of the game has been developed with simple software interfaces with which to connect your controllers. The screenshot below shows a game in progress on the simulator. You can enter an agent controller or a ghost-team controller (or even both). The aim of the Ms. Pac-Man agent is to score as many points as possible. The aim of the ghost team controller is to prevent this!

More details: http://cswww.essex.ac.uk/staff/sml/pacman/kit/AgentVersusGhosts.html

Mario AI competition

This competition is about learning, or otherwise developing, the best controller (agent) for a version of Super Mario Bros.

The controller’s job is to win as many levels (of increasing difficulty) as possible. Each time step (24 per second in simualated time) the controller has to decide what action to take (left, right, jump etc) in response to the environment around Mario.

The competition software is in Java and designed to be as simple as possible to get started with, with no external dependencies. Submissions in Java are encouraged, but any programming language can
be used thanks to a TCP client/server interface (example Python client provided). The organization of the competition is similar to the simulated car racing competitions held in conjunction with recent IEEE
conferences: submission will be by email, and the source code of all submitted entries will be published on the competition web page.

Submission deadline: September 3 2009

Competition web page: http://julian.togelius.com/mariocompetition2009/

Screen-Capture Ms Pac-Man

Ms. Pac-Man is a challenging and interesting game that is usually viewed form the perspective of the Ms. Pac-Man agent. The game also provides an excellent environment for testing multi-agent strategies for controlling the team of ghosts, with the aim either of optimising the playability of the game, or simply to minimise the score obtained by the agent. It is this latter more measurable objective that provides the focus for this competition. For this purpose an efficient simulator of the game has been developed with simple software interfaces with which to connect your controllers. The screenshot below shows a game in progress on the simulator. You can enter an agent controller or a ghost-team controller (or even both). The aim of the Ms. Pac-Man agent is to score as many points as possible. The aim of the ghost team controller is to prevent this!

The mode of interaction is as follows: about 15 times per second your program will be sent a pixel map of the Ms. Pac-Man window, and it then responds with an integer indicating the direction of the joystick.

More details at http://csee.essex.ac.uk/staff/sml/pacman/PacManContest.html

Simulated Car Racing

The simulated car racing competition of CIG-2009 is the final event of the 2009 Simulated Car Racing Championship, an event joining the three competitions held at CEC-2009, GECCO-2009, and CIG-2009.

Important Dates

  • Submission deadline: September 1st 2009
  • Conference: September 7th-10th 2009

Rules and Regulations
The competition involves three Gran Prix on three (unknown) tracks. Each Gran Prix is organized in two stages: the warm-up and the actual race. In the warm-up, each driver will race alone for 10000 game tics, approximatively 3 minutes and 20 seconds of actual game time. The eight drivers that will cover the more distance will qualify for the next stage, the actual race.

In the second stage, the eight drivers will race together. Each race consists of ten trials. The goal of each trial is to complete five laps from a randomly generated starting grid. At the end of each trial, the drivers will be scored using the F1 system: 10 points to the first controller that completed the three laps, 8 points to the second one, 6 to the third one, 5 to the fourth, 4 to the fifth one, 3 to the sixth, 2 to the seveth, and 1 to the eighth. The driver performing the fastest lap in the race will get two additional points. The driver completing the race with the smallest amount of damage will receive two extra points. The final score for each driver in the Grand Prix will be computed as the median of the 10 scores collected during the trials.

Competition Software

Support

Archives

Links