Cary, David Fall 20071
Generational Play

Abstract

In many games, the player is linked to a given character or object within the game (often referred to as the player's avatar). In most cases, this character represents the player and is their primary way of interacting with the game world. As such, the abilities of these characters play a major role in shaping the player's experience. Some games take advantage of this by providing a number of different character abilities. This allows for a wider range of experience, which in turn may allow the players to play closer to their preferred style. It can also encourage replayability as a player may want to try the same game again from a different angle. However, this does leave the question of how each character is designed.

A simple approach is to have the developers provide a premade set of characters. However, this means the choices and styles are limited to that set. As such, the replay options and ability to personalize the character are limited. A common response to this is to let the players customize the character's properties and abilities. The idea is to let the player pick and prioritize the features they want within the game's limitations. This helps give the player the kind of gameplay they're looking for while leaving a good variety of replay options. However, each customization option means one more choice the user has to make. Since these choices usually have long term effects, it's possible for users to get locked up in "analysis paralysis" trying to make the "right" choices. This drive to make a strong set of choices can also lead to "cookie cutter" characters as superior combinations are discovered. So how can we provide a wide range of characters while avoiding these problems?

In this project, we'll see how interactive genetic algorithms can provide this variety. Evolutionary computation has already been applied to art and music as fresh sources of dynamic content. If the same techniques are applied to game characters, they should provide a similarly rich range of possible characters. By creating a sample program that uses these techniques, we'll be able to explore how evolutionary computation can generate a variety of viable and playable characters.

Images

bunny by Asa Tse turtle by Asa Tse flowers by Asa Tse psp guy by Asa Tse blow fish by Asa Tse shapes to cat by Asa Tse dinosaur by Asa Tse cute girl by Asa Tse bee by Asa Tse frog by Asa Tse dog by Asa Tse cat by Asa Tse canvas face by Asa Tse duck face by Asa Tse