Roads to What We Want: A Game Generator based on Reverse Design
Zixuan Deng, Yanping Xiang · 2021
It is difficult for existing approaches of procedural content generation (PCG) to balance automatically and purposely generating ideal game levels. Inspired by the idea of reverse design, we give ways of modelling the optimal solutions or the rough ones into our framework. We build the framework based on a multi-objective genetic algorithm NSGA-II, taking the interacts with different game elements in a game as the objectives for maintaining diversity during generation. Experiments demonstrate the effectiveness of our approach in producing ideal game levels purposely and the ability of being a general game generator that generates interesting levels which require the players to think in depth.