Optimal Policy

A policy is a mapping from states to actions. It will tell the Agent what Action to do in each state that it can be in.

An optimal policy maximizes the expected sum of rewards.