16.3 Policies

A policy is the agent's strategy: a rule that maps each state to an action (or to probabilities over actions). A deterministic policy always picks the same action in a given state, while a stochastic policy chooses actions with some randomness. The aim of RL is to find an optimal policy that maximizes expected long-term reward.

A farmer's planting policy is a good analogy: 'if the rains have started and soil is moist, plant teff; if it's still dry, wait.' That is a mapping from the state of the land to the action to take.