Model-based RL

It is a type of reinforcement learning that uses a model of the environment instead of trial-and-error learning. These models can be learned from data or designed by humans.