Conception of an AI for the game Stratego

Details

Supervisors

Schaus, Pierre

Faculty

Ecole polytechnique de Louvain

Degree label

Master [120] en sciences informatiques, à finalité spécialisée

Abstract

This thesis resolves into the conception of an AI for the complex game of Stratego, a strategic and challenging zero-sum game opposing two players on a board, this game is characterized by imperfect information and vast state space. The Monte-Carlo Tree Search algorithm technique was used and tuned with several other methods te better fit the challenges. My custom agent succeeds in achieving a 72% win rate in my experiment against other AIs. In this work, I present the game, its rules, strategies, why are the difficulties in it, and also the state-of-the-art agents before presenting the pros and cons of using such a technique of states' exploration. Finally, I show the result of the games against other agents and provide some insight for future works for the enhancement of my custom AI.