The quest for the perfect perfect-maze

Abstract

In this paper, the quest for the perfect perfect-maze is performed over the search space of perfect mazes using an approach of search-based procedural content generation. Perfect maze construction is rather random with little to no control of the final product. We propose a search-based framework based on attributes or metrics of a constructed maze to provide a foundation for evaluation functions (fitness functions). Since the meaning of “perfect” is subjective and different for every designer, we allow designers to construct their own evaluation function to generate the best maze. We have also analyzed each metric's space on an exhaustive enumeration of small-sized mazes to determine allowable, and perhaps desirable, ranges for each metric. Using these metrics, an evaluation function is constructed to search for the “best” maze.