Using Genetic Algorithm to Solve Puzzle Games: A Review

Abstract

Puzzles have been recognized for their development as a popular form of entertainment due to their ability to intricately challenge the mind while engendering creativity in the player. The development of puzzle games has given rise to a new generation of puzzle games characterized by diverse sequences and different image variations. With the rapid development of puzzle games, we looked at solving approaches using Genetic Algorithms (GA). In this paper, we try to analyze several puzzle games such as Sliding Blocks, Sudoku, Tic-Tac-Toe, and Jigsaw that can be solved using GA. We found that 120 papers have examined the use of GA for puzzle games, and eliminated into 14 papers. We evaluated these 14 papers for each puzzle game we selected by comparing the chromosome representation, GA operator, GA parameters, and the results. Based on the discussion, the application of GA to solve puzzle games can be effectively executed with a high degree of accuracy. Puzzle games that use measurement methods such as Sliding Block, Sudoku, and Jigsaw run in a similar pattern. What is common to all of them is that the chromosomes are represented as matrices or arrays in all cases, and standard genetic operators such as selection, crossover, and mutation are used. The population size is large, often 1000 chromosomes, and parameters such as mutation rate are kept low, around 5%. On the other hand, the performance of GA for solving Tetris and Tic-Tac-Toe from each publication cannot be compared due to different measurement methods and metrics.