Generic Puzzle Level Generation for Deterministic Transportation Puzzles

Abstract

The focus of this thesis lies on the generic evaluation of levels. We believe this is the most crucial component of a generic level generator. We describe three ’generic’ evaluation methods, of which the breaking-rule approach was the most promising. A heuristic is used that predicts the problems the player has to overcome, to solve the level. It uses a restricted input space and breaks certain rules to estimate the distance from a solution. To test the evaluation method, we generated levels for 5 di↵erent puzzles and tested their diculty label in a small experiment. The levels are deemed hard by the experiment, even after generating an easier level set, when most participants could not solve any levels in an hour of play. The solving time and diculty rating the participants gave to the levels has a high correlation with the evaluation scores of the breaking-rule approach.