Model-Based RL

RL methods that learn a model of environment dynamics and plan using it. More sample-efficient than model-free methods but requires accurate world models.