Lösning av Sudoku med mänskliga strategier

Abstract: Sudoku is a puzzle that has become popular during the last decade. A great number of algorithms have been implemented to solve the problem but many use approaches that do not correspond to how a human solves a sudoku puzzle. The purpose of this project was to evaluate how a strategy based algorithm which utilizes a more human approach compares to a so called Dancing Links-algorithm and an exhaustive search algorithm with respect to solving time and correctness. A qualitative study was conducted, with four respondents, which together with a literature study lay the foundation for the strategies in the implemented algorithm. Furthermore, a comparison was performed with 725 sudoku puzzles of which the strategy-based algorithm only managed to solve sudoku puzzles of a limited difficulty where the others solved all. The result showed though, that the strategy based algorithm was the fastest for the difficulties that it solved, but at the same time it showed a trend hinting at increased solving time with increasing difficulty. Finally, it could be determined that Dancing Links is a generally faster algorithm than the others. A strategy based algorithm could however be a fast alternative when solving sudoku puzzles of the difficulties that often appear in daily papers.