Procedural dungeon generation for a 2D top-down game
Michael Weeks, Jonathon Davis · 2022
This work examines solvability, difficulty, and engagement of procedurally generated content for a 2D, top-down game. We adapt the A* algorithm to find a solution when there are keys and locks between the starting position and the goal. It requires iteration to find a solution, since analysis of points in the dungeon alone are computationally expensive. If no solution exists, the program can detect this by showing no progress made in an iteration. We ran this on 30 dungeons, and classified their content according to the metrics of solvability, difficulty, and engagement.