Automatic Generation And Verification For Tower Defense Game Levels

In recent years,with the increase in the number of game players and the increase in game development costs,a more efficient content production method is urgently needed in the field of game development.Procedural Content Generation(PCG)in Games is a technology that can automatically generate game content.Based on PCG,this thesis proposed an automatic generation scheme of tower defense game levels,and further explored the automatic verification method of level's difficultyFirst,this thesis takes a classic tower defense game KRF as research environment,developed a simulation system for tower defense games,and built the simulation environment required for research.Automatic level generation includes two parts:map generation and monster sequence generation.In the map generation part,this thesis proposed a method to represent game paths,and designed an algorithm for automatically generating game paths based on search.By adjusting some parameters,this algorithm can generate a variety of paths that have different styles.In the part of monster sequence generation,this thesis proposed a compound genetic representation of monster sequences.This representation can enhance the generation effect by reusing some schemes that are designed by humans.Based on this,a monster sequence generation algorithm based on genetic algorithm is designed,and the validity of this algorithm is proved by a Turing test.It proves that the game levels finally generated in this thesis are similar to the original game.This thesis also proposed a quantitative calculation formula for game's difficulty based on the subjective feelings of players in the game.Based on Monte Carlo tree search,two agents are designed which can automatically play games and output game strategies.The agents are used to simulate game players,and the difficulty of the game level is quantitatively measured by the performance of the agents.Quantitative measurement of game level difficulty can also be combined with level generation to further enhance the effect of level generation.