Prisoner's Dilemma player

Pablo Gomez Esteban, Ulysses Bernardet, Sergi Bermudez i Badia

Two different decision making algorithms have been combined in an Aisoy 1 robot that plays the prisoner's dilemma game. One of these algorithms is concerned on the individual aspect of the game and the other tries to take into account the outcomes of the adversary in order to play more cooperatively. Both algorithms are weighted by a lambda value which depends directly on the simulated emotion, which depends on the result of the last game. So, as more happy the robot is, more cooperative it will be, and vice versa, as more angry it is, more selfish it will behave in the next game. It includes a q-learning process to punish its last action when necessary.