Procedural Personas as Critics for Dungeon Generation

Abstract

Abstract. This paper introduces a constrained optimization method which uses procedural personas to evaluate the playability and quality of evolved dungeon levels. Procedural personas represent archetypical player behaviors, and their controllers have been evolved to maximize a specific utility which drives their decisions. A "baseline" persona evaluates whether a level is playable by testing if it can survive in a worst-case scenario of the playthrough. On the other hand, a Monster Killer persona or a Treasure Collector persona evaluates playable levels based on how many monsters it can kill or how many treasures it can collect, respectively. Results show that the implemented two-population genetic algorithm discovers playable levels quickly and reliably, while the different personas affect the layout, difficulty level and tactical depth of the generated dungeons.

Automated Playtesting With Procedural Personas Through MCTS With Evolved Heuristics

“…This paper builds on our previous work on MCTS agents with hand-crafted utility scores for the MiniDungeons 2 game [5], expanding on those concepts by broadening the number and utility of personas (through human design) as well as discovering UCB-like criteria (through evolution) which outperform UCB1. More broadly, this paper enhances our earlier definitions of procedural personas [6]- [8] which were applied for simulation-based level generation [9]. However, the MCTS agents used in this paper are more modular in their utility definitions and afford far faster runtimes when performing automated playtesting.…”

Exploring Apprenticeship Learning for Player Modelling in Interactive Narratives

“…These works, although focused on the IN domain, do not implement AL. As for modelling player behaviour, related work includes persona modelling for procedural content generation, where archetypical models of players are used to evaulate playability of generated game content [8,12]. These works focus on game genres other than INs, and their model uses stricter archetypes than ours.…”

Evolving missions to create game spaces

“…The family of search-based PCG [1] methods attempt to gradually improve a level by applying local changes; most often, artificial evolution is used and the local changes take the form of mutation of tiles in a grid-based map or recombination of the layouts of two parents to create offspring that combine the features of both parents. In search-based PCG, it is common to select the most promising parents to create the next batch of results (generation) based on a quantifiable objective function which evaluates how appropriate a game level is: examples include the length of its paths [9], the combat duration between artificial agents [4] or the distribution of its treasures [5].…”

“…Most often, such generators create the level's layout and then evaluate its spatial characteristics such as its navigable regions [4] or functional characteristics derived from e.g. playtraces of artificial agents running through it [5]. In the case of [6], the generator creates a tile-based layout of a dungeon for a role-playing game adventure module, which is then used to derive a room connectivity graph for placing encounters to follow the progression of a player from the dungeon's entrance.…”