Neuroevolution of content layout in the PCG: Angry bots video game

Abstract

Abstract-This paper demonstrates an approach to arranging content within maps of an action-shooter game. Content here refers to any virtual entity that a player will interact with during game-play, including enemies and pick-ups. The content layout for a map is indirectly represented by a Compositional Pattern-Producing Networks (CPPN), which are evolved through the Neuroevolution of Augmenting Topologies (NEAT) algorithm. This representation is utilized within a complete procedural map generation system in the game PCG: Angry Bots. In this game, after a player has experienced a map, a recommender system is used to capture their feedback and construct a player model to evaluate future generations of CPPNs. The result is a content layout scheme that is optimized to the preferences and skill of an individual player. We provide a series of case studies that demonstrate the system as it is being used by various types of players.

Integrated Approach to Personalized Procedural Map Generation Using Evolutionary Algorithms

“…This means that at any map index, some players are having positive experiences and other are having negative ones, leading to a constant median rating of either 'Poor' or 'Fair' at every map index. These fluctuating ratings were observed in the sample player analysis in [12].…”

“…Due to the large solution space of both the geometry and the content quantities, there is a wide variety in maps that were played and no exact map was played by more than one player. As it is hard to establish correlation between such varied maps on a visual basis, readers are referred to [12] for examples of the kind of maps that were evolved for specific players. Table I shows general results of the experiment.…”

“…In [12] we provided an initial high level qualitative examination of the types of maps that were being generated and the experiences that individual players were having. This was done by examining the playing time of three sample players.…”

“…However, Yannakakis and Hallam [35] provide a comparative analysis of post reporting techniques for player modeling and found that a pairwise preference approach introduces less error. The results presented in [12] show that there is indeed an issue associated with the order that maps are experienced, especially as a player's preferences and skill change rapidly in the first few maps. However, it was found that while the classifier may take a sudden penalty for contradicting training data, this is typically quickly negated by further training data that aligns with the new preferences.…”

Video game personalisation techniques: A comprehensive survey

“…Role-playing game maps [5] Difficulty, weapon control, and objectives [6] Platforming levels [7] Personality No current games applications Experience Camera position [8,9] Plot/story points [10] Platforming levels [11,12] Action game levels [13] Performance Platforming levels [14] Enemy type and count [15] Platforming levels and dungeon structure [16] Battle missions [17,18] In-game behaviour Quest structure [19,20] Weapon behaviour [21] player model structure. Bateman et al [33] also points out that without empirical evidence the four types in Bartle's model may or may not be mutually exclusive (i.e., it is possible a single player possesses attributes described by two or more types).…”