Controlling a Dynamic System Through Reinforcement Learning

Control theory deals with the problem of finding a control law for a given dynamic system on which one can act by means of a command. The goal is then to bring the system from a given initial state to a certain final state, possibly respecting certain criteria. In this paper, we are interested in the problems of optimal control of dynamical systems, seen under the prism of reinforcement learning (RL). RL is an area of machine learning that enables an agent to learn in an interactive environment by trial and error using feedback from its own actions and experiences. We compare the Q-learning, a model-free reinforcement algorithm, to the feedback controller Linear Quadratic Regulator on the OpenAI Gym Cart-Pole system, a toy example which is quite famous in both the control engineering and RL community.