Algorithmic collusion with imperfect monitoring

We show that if they are allowed enough time to complete the learning, Q-learning algorithms can learn collude in an environment with imperfect monitoring adapted from Green and Porter (1984), without having been instructed do so, communicating one another. Collusion is sustained by punishments take form of “price wars” triggered observation low prices. The have a finite duration, being harsher initially then gradually fading away. Such both deviations adverse demand shocks.