Uma implementação de Risk para competição e aprendizado em IA

Abstract

Along with the evolution of Artificial Intelligence (AI), the games have increasingly proven an important training and testing environment in this area. This is because they are controlled environments and easy to repeat. In addition, they have been widely used as an AI education tool. To complement this teaching, competitions are also used, because in addition to generate interest and motivation, they are able to produce content for the next students who are going to study or research the area of AI. With this in mind, this work propose an implementation of the board game Risk, with a simplified architecture and aimed at competitions between agents (AI), where the communication between game and agents will be done through JSON files, allowing agents implementation in different languages with support for reading and writing on this type of file. Risk is a fully observable, stochastic and turn-based game, these characteristics allow the creation of different AI methods, from the simple to the most complex. The goal is to develop an environment with a visual interface for AI competitions, which is simple and easy to implement its agents, where students with different levels of knowledge can use it with great freedom.