Unlock: Model-Based Reinforcement Learning

Learning a model of the environment and planning with it. Dyna architecture, learned world models, planning as simulated experience, sample efficiency, and the model-error problem.