Competitions

Competitions Chair: Simon M. Lucas

IEEE CEC 2009 offers the following competitions with the aim of better understanding the state of the art in each of these areas. Each competition has its own deadline and rules, so please read on for more details.

Authors of leading entries of the game-related competitions are encouraged to write up their work to submit to IEEE Transactions on Computational Intelligence and AI in Games. This journal publishes high quality research in all aspects of games related to CI and AI.

Performance Assessment of Constrained / Bound Constrained Multi-Objective Optimization Algorithms

Optimization for multiple conflicting objectives results in more than one optimal solution known as Pareto-optimal solutions. Although one of these solutions is to be chosen eventually, the recent trend in evolutionary multi-objective optimization studies have focused on approximating the Pareto front by a set of solutions.

Such a set of solutions can collectively provide a good insight to the different trade-off regions on the resulting efficient frontier, thereby aiding a better and more confident decision making.

Evolutionary multi-objective optimization (EMO) methodologies have been suggested since the eighties for this task. Since then a number of performance assessment methods has also been suggested. After more than 20 years of research and development of efficient EMO algorithms, we realize that it is time to evaluate the existing EMO methodologies on carefully chosen test problems which are scalable with respect to the objectives, the decision variables and constraints with complex Pareto shape in the decision space. The comparisons will be made for a limited number of overall evaluations, so that the existing or new algorithms can be evaluated for different functional abilities. This competition will be developed in collaboration with the participants, and is also being run as a special session. The competition deadline is the same as for the special session paper deadline.

For more information see here.

Evolutionary Computation in Dynamic and Uncertain Environments

Many real-world optimization problems are subjected to dynamic and uncertain environments that are often impossible to avoid in practice. For instance, the fitness function is uncertain or noisy as a result of simulation/measurement errors or approximation errors (in the case where surrogates are used in place of the computationally expensive high fidelity fitness function). In addition, the design variables or environmental conditions many also perturb or change over time. For these dynamic and uncertain optimization problems the objective of the evolutionary algorithm is no longer to simply locate the global optimum solution, but to continuously track the optimum in dynamic environments, or to find a robust solution that operates optimally in the presence of uncertainties. This poses serious challenges to conventional evolutionary algorithms. This competition will be developed in collaboration with the participants, and is also being run as a special session. The competition deadline is the same as for the special session paper deadline.

For more information see here.

Unreal Tournament 2004 Deathmatch

The competition task is to create the winning computer controlled bot in an Unreal Tournament 2004 Deathmatch Challenge. Each entry will be played against the other entries and against some bots developed for the competition, using standard deathmatch scoring. This competition was inspired by the http://botprize.org challenge.

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. A particularly easy way to interface to the game is to use the Pogamut library, which is written in Java and is available as a Netbeans plugin.

TORCS-Based Car Racing

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 competition software, an introductory manual, and the competition rules can be found on the main competition web page.

Software Agent Ms Pac-Man

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. 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. For more information, including results, see the IEEE WCCI 2008 competition page.

Ms Pac Man Ghost Team Challenge

This competition is currently under development. The idea is that entrants can either provide a teams of ghosts, or a pac-man agent (or both).

The score for an agent is its average score over the set of ghost-teams (excluding any team submitting by the same entrant), and vice versa for ghosts.

One of the tangential aims of the contest is to develop new ghost teams so that we can play against them as humans, and see whether different (and perhaps more intelligent) ghosts are more or less fun to play against.

The software interfaces for each ghost are currently under discussion.

One issue to be aware of is the rules should avoid boring stalemate situations. For example, the ghosts could limit the Pac-Man's high score by constantly patrolling a small area of the maze making it impossible for the Pac-Man to eat the pills in that area.

Possible ways to avoid this may be to: