Dynamic programming methods use a bootstrapping process to update the values to make them self-consistent under the current policy. Monte Carlo methods sample the MDP to acquire information. Temporal difference (TD) methods combine both bootstrapping and sampling. However, unlike Monte Carlo methods, they update the values and policy while the agent traverses the states of the MDP instead of afterwards.