Deep Q-learning agent that learns to navigate an 8x8 treasure maze using experience replay, target networks, and automated tests.