CONTEXT
Catch the AI is a chess-like game based on Yokaï no Mori. It uses the Monte Carlo Tree Search (MCTS) algorithm to calculate AI strategies during matches.
This project allowed me to understand the various algorithms used for AI. Although I mainly implemented MCTS here, I also explored and understood the principles of the following approaches:
-
Minimax with Alpha-Beta Pruning: Ideal for competitive 1v1 games. Optimization is possible to reduce the number of branches explored via pruning.
-
Q-Learning: Suitable for games with many possibilities and unpredictable reactions (e.g., FIFA).
-
Deep Reinforcement Learning (DRL): Suitable for complex games where AI learns through experience.
-
Heuristic Decision Tree: Used for games with well-defined rules established in advance.
-
Behavior Tree: A simple approach for behaviors based on specific rules defined through conditions ("if").
-
Model-based Planning: Ideal for games with numerous variables but limited response options (e.g., Civilization). This method simplifies the game environment and relies on a mathematical model to predict possibilities.
-
Various Optimization Techniques: Includes state definitions to reduce search, combining methods, anticipation, caching information, using object pools, and more.
Project Creation Date
Developed in 2024 as part of Gaming Campus.
Team
Developed in 1 week in collaboration with another developer, Hugo Marseu.
LINK(s)
Click here to access GitHub
