Maximize the cumulative reward

Minimize the immediate reward

Maximize the state-action value

Achieve the optimal state transition