Episodic tasks are the tasks that have a terminal state (end). In RL, episodes are considered agent-environment interactions from initial to final states.

For example, in a car racing video game, you start the game (initial state) and play the game until it is over (final state). This is called an episode. Once the game is over, you start the next episode by restarting the game, and you will begin from the initial state irrespective of the position you were in the previous game. So, each episode is independent of the other.

In a continuous task, there is not a terminal state. Continuous tasks will never end. For example, a personal assistance robot does not have a terminal state.