Advanced AI for enemy management in digital games

Abstract

The video game industry suffers from increasing costs of production. From multimillion dollar projects to single developers, new techniques for increasing game value are needed. In this Thesis we focus on the challenge of players becoming too powerful in the final stages of a game, which decreases late game playability and player satisfaction. A set of rules implemented in the design stage of the game can be monitored by an AI controller and, if implemented correctly, can boost the game playability duration without financial overheads. A game should emphasize that every item or power that a user can have, has an equal amount of advantages and disadvantages. This would allow the AI to learn about the player’s build through the interactions of the player with the enemies generated at the area which the AI is monitoring. An Artificial Intelligence Algorithm which we have named Score Based AI was developed that uses a formula for generating the composition of the AI adversaries based on how hard it was for the player to defeat them in the last wave. For instance, if wave 2 has 1 more enemy of type A and the user’s score is lower than in wave 1, then we need more enemies of type A. The project was successful in implementing a simplistic yet effective method which decides the optimal path ahead based on the 2 previous waves of enemies.