Skip to main content
Please use this identifier to cite or link to this item:

Issue Date: 1-Aug-2019
Title: Perpetually playing physics
Authors: Beeler, Chris
Publisher : University of Ontario Institute of Technology
Degree : Master of Science (MSc)
Department : Modelling and Computational Science
Supervisor : van Veen, Lennaert
Tamblyn, Isaac
Keywords: Reinforcement learning
Machine learning
Abstract: Here we discuss ideas of reinforcement learning and the importance of various aspects of it. We show how reinforcement learning methods based on genetic algorithms can be used to reproduce thermodynamic cycles without prior knowledge of physics. To show this, we introduce an environment that models a simple heat engine. With this, we are able to optimize a neural network based policy to maximize the thermal efficiency for different cases. Using a series of restricted action sets in this environment, our policy was able to reproduce three known thermodynamic cycles. We also introduce an irreversible action, creating an unknown thermodynamic cycle that the agent helps discover, showing how reinforcement learning can find solutions to new problems. We also discuss shortcomings of the method used, the importance of understanding the class of problem being handled, and why some methods can only be used for certain classes of problems.
Appears in Collections:Electronic Theses and Dissertations (Public)
Faculty of Science - Master Theses

Files in This Item:

File Description SizeFormat
Beeler_Chris.pdf2.83 MBAdobe PDFView/Open

Items in e-scholar@UOIT are protected by copyright, with all rights reserved, unless otherwise indicated.