Skip to content

Latest commit

 

History

History
31 lines (17 loc) · 1.15 KB

README.md

File metadata and controls

31 lines (17 loc) · 1.15 KB

Q-Learning

The purpose of this exercice is to create an AI dedicated to Dungeon Maps generation. From a 4x4 matrice, the AI will generate a matrice with a starting point, an ending point and a treasure point. Each cell will contain 0 to 4 walls, 0 to 1 starting keypoint, 0 to 1 treasure keypoint and 0 to 1 ending keypoint. A cell can contain only 1 keypoint.

Approach Developped : Our approach consist to generate a a 4x4 matrix and to process a key perfomance indicator ( KPI ) which indicates the difficulty of our grid. https://fr.wikipedia.org/wiki/Matrice_d%27adjacence

To process KPI of difficulty : we transform our 4*4 matrix in an adjacence matrix with size 16x16. It's a non-optimal adapation of Dijkstra : https://fr.wikipedia.org/wiki/Algorithme_de_Dijkstra

We generated a follwing grid :

map

We process/explore (finite combinatory) all possibilities of different size of path :

maze

To run this script :

python 3.11

Pip install numpy