Development of a system to create artifical intelligence with adaptive difficulty adjustment for multiplayer games (german)
Please click here to download my Bachelor Thesis
The usage of artifical intelligence in games and especially video games keeps becoming more important. Projects like AlphaGo show that neural networks are capable of beating even the most skilled players even in complex games. The process for developing video games however presents a few problems that can make the creation and usage of artificial intelligence problematic.
One particularly difficult field is the controlling of agents and adjusting their difficulty for the player without resorting to cheap tricks. Often times in games ai enemies receive stronger weapons to adjust the difficulty, or they receive information that would otherwise be impossible or very hard to get for a normal player so that they can compensate their often times weak logic systems with unfair advantages. The behaviour of these agents is then barely comprehensible and frustrating for the player.
Therefore this thesis is concerned with the development of a system that allows developers to create artifical intelligence within agent systems for video games. These agents shall to the players and external information and adjust their behaviour based on rational input and actions. They make fair judgements about the players’ skill level and their own adjustments.