Rogueinabox: a Rogue environment for AI learning. Framework development and Agents design.

In this thesis we introduce Rogueinabox: a higly modular learning environment built around the videogame Rogue, the father of the roguelike genre. It offers easy ways to interact with the game and a whole framework to build, customize and run learning agents. We discuss the interest and challengies of this game for machine learning and deep learning, and discuss our initial experiments of training. We show the userfulness and convenience of Rogueinabox employing it in combination with QLearning tecniques to build an agent that explores the dungeon.