A Survey of Model-Free and Model-Based Reinforcement Learning for the Acrobot

Abstract

This survey selects several model-free and model-based techniques and evaluates their relative strengths and weaknesses on the Acrobot task. While this is not indicative of these algorithms’ performance on continuous state-space control problems in general or exhaustive of the techniques available, the ultimate goal is to produce a finding about the importance of learning a model.

Publication
Department of Electrical & Computer Engineering, Cornell University
Date
Links