Imitation-based Artificial Player for Pic-A-Boo

Year:

2020

Phase:

Finished

Authors:

Elio Samuel Granados Freitas

Advisors:

Abstract

Pic-a-boo is multiplayer game about taking pictures of other players in a dark room. Not all information is available at all time to the players, so some guessing is required to win. In this work, we developed artificial agents that would be able to imitate human players and serve as their substitute when needed. This was achieved through an artificial neural network iteratively trained by playing agains a human player, in the attempt of capturing and replicating what makes his or her gameplay distinct and identifiable as a player. Although the agents were able to navigate the map believably, other actions such as taking photos were produced randomly without any discernible human pattern.