Apply one TD(0) state-value update using immediate reward, discounted next-state value, and learning rate.
