Infinite Mario Bross AI using Genetic Algorithm

Abstract

Evolutionary Algorithm (EA) is commonly used to generate optimal Artificial Intelligence (AI) controller. It is a technique used to enhance the performance of generated controller. EA enables the system to evolve, to adapt and learn to give a better output. The implementation of EA into 2D game is not something new. Researchers used gaming platforms to test their own ideology or proposed algorithms. In this paper, a finite state machine which suitable to be used for Infinite Mario Bros game is proposed. The Genetic Algorithm (GA) is used along with the proposed finite state machine to evolve an AI agent that is capable to pass some levels of the game. The experimentation results showed that the finite state machine evolved with GA is able to create a competitive game bot that can pass through at least 3 levels of different game maps. The generated AI controller can guarantee to accomplish the tasks for some levels.