16.4 Q-Learning

Q-Learning is a classic RL algorithm that learns a Q-value for each state-action pair: an estimate of the total future reward expected if the agent takes that action in that state and then behaves optimally. Once Q-values are known, the best policy is simply to pick the action with the highest Q-value in each state.

It is model-free: the agent does not need a map of the environment in advance. Like a new minibus driver in Addis who, after many trips, learns from experience which turns at each junction tend to save time, without ever seeing a full city plan.