2K Bot Prize - a Turing Test for Bots
Organiser: Philip Hingston
Sponsored by 2K Australia, the competition task is to create a computer game bot (a.k.a. non-player character or NPC) which is indistinguishable from a human player. Those entries that pass this test will share the major prize of A$7,000 cash, and will also be offered a trip to 2K Australia's studio in Canberra. If the major prize is not won, a minor prize of A$2,000 plus a trip to the studio will be awarded.
The game used for the competition will be based on a modified version of the DeathMatch game type for the First-Person Shooter, Unreal Tournament 2004. This modified version provides a socket-based interface (called Gamebots) that allows control of bots from an external program. In addition, several extra modifications will be made especially for the competition:
- No chatting.
- Some aspects of the game play will be modified in a way not known to competitors.
Chatting will be disabled lest the competition revert to a chat bot competition! Undisclosed modifications will be made to encourage the submission of bots that can learn how best to play, as human players are able to do. Examples of game features that might be modified include weapon effects, physics, map layout, and/or any other modification deemed suitable by the organisers.
TORCS-Based Car Racing
Organisers: Daniele Loiacono, Julian Togelius and Pier Luca Lanzi
The competition is very similar to the WCCI 2008 competition and is based on the TORCS simulation environment. The aim of the competition is to learn (or otherwise develop) a controller that races around a number of laps as fast as possible, alone or in the presence of other drivers. We will score every submitted controller on the distance raced in a fixed amount of time when driving on its own on a set of tracks. At the end of the competition, the best few controllers will race against each other on a different set of tracks, validating that the controllers perform well in the presence of other cars and that their performance generalizes to other tracks than those they were trained for. The winner of the final competitive races will get to present their controller at CIG2008, and will have their registration fee reimbursed.
Previous competitions saw a wide variety of controllers submitted, using different architectures (e.g. neural networks, fuzzy logic, force fields, expression trees) and training methods (e.g. genetic algorithms, evolution strategies, td-learning, direct critics, hand-coding). We hope to see the same variety of controller development strategies compared in this competition. We also hope that we will see the participation of people outside the computational intelligence community, e.g. from the games industry.
Software Agent Ms Pac-Man
Organiser: Simon Lucas
The aim of this competition is to provide the best software controller for the game of Ms Pac-Man. This is a great challenge for computational intelligence, machine learning, and AI in general.
Unlike Pac-Man, Ms. Pac-Man is a non-deterministic game, and rather difficult for most human players. As far as we know, nobody really knows how hard it is to develop an AI player for the game. The world record for a human player (on the original arcade version) currently stands at 921,360 (read more). Can anyone develop a software agent to beat that?
The competition uses screen-capture to get the current state of the game, and provides a fascinating challenge for your algorithms with some inevitable uncertainties arising from variable delays in the screen capture process.