Developing AI Agents for Competitive Board Games Using Deep Reinforcement Learning: Two-Player Azul Case Study

Azul is an award-winning board game that challenges players to create stunning tile mosaics while making strategic decisions and competing for high scores. Most players find the agents to be less strategic, simple and not challenging enough hence easy to win the game, making it less competitive for the player. The implementation of Deep Reinforcement Learning (DRL) methods by designing AI agents that can play the competitive two-player strategy board game can bridge this gap. With a pre-existing simulation model of the game at hand, the intention is to create intelligent agents that can learn and improve their strategies through environmental interaction. Through training these agents using DRL algorithms such as Deep Q-Networks (DQN), the project will observe how AI can be made to create strategies that are close to human-like decision-making for the game.