In simple words: RL learns action choices from reward signals.
The parts
In Reinforcement Learning (RL), an agent observes a state, chooses an action, and receives a reward signal. It learns a policy for choosing actions that produce better long-term reward. Reward is a number defined by designers; a machine does not feel happy.
Use with care
A robot vacuum might receive positive reward for cleaning floor space and negative reward for collisions or wasted time. The reward must match the real objective, or the agent may learn an unwanted shortcut.
See the flow
state→action→reward→updated policy
Simple example
A vacuum tries a route, bumps into a chair, gets a negative score, and adjusts future route choices.
Remember: RL learns action choices from reward signals.