Electrotechnical and Computer Engineering
Vol. 37 No. 11 (2022): Proceedings of Faculty of Technical Sciences
PERFORMANCE COMPARISON OF THE STATE VARIABLE FEEDBACK REGULATOR ALGORITHM TO A REINFORCEMENT LEARNING BASED CONTROL ALGORITHM
Abstract
This paper presents one way of utilising Reinforcement Learning (RL) to control a mechanical system. First, the modeling of a double pendulum and the training of an agent that would control the position of the pendulum was done. Then, the results of this algorithm were compared to the result of the state variable feedback controller. Finally, the parameters of the model with which the agent was trained and for which the state variable feedback controller was designed were changed and the results of both control algorithms applied to such a model were compared.
References
[1] R. Hafner / M. Riedmiller, „Reinforcement learning in feedback control,“ Machine Learning, 2011.
[2] M. Riedmiller, T. Gabel, R. Hafner / S. Lange, „Reinforcement learning for robot soccer,“ Autonomous Robots, 2009.
[3] R. S. Sutton / A. G. Barto, Reinforcement Learning: An Introduction, Cambridge, MA: MIT Press, 2018.
[4] C. Watkins / P. Dayan, „Q-Learning,“ Machine Learning, t. 8, 1992.
[5] A. Elbori, „Simulation of Double Pendulum,“ Quest Journals, Journal of Software Engineering and Simulation, 2017.
[6] W. L. Brogan, Modern Control Theory, New Jersey: Prentice Hall, 1991.
[7] H. K. Khalil, Nonlinear Systems, New Jersey: Prentice Hall, 2002.
[8] A. Kuhnle, J.-P. Kaiser, F. Theiß, N. Stricker / G. Lanza, „Designing an adaptive production control system using reinforcement learning,“ Journal of Intelligent Manufacturing, 2020.