Utilization of Supervised and Reinforcement Learning in the Automation of the Classical Atari Game "Pong"
| dc.contributor.advisor | Bhounsule, Pranav A. | |
| dc.contributor.author | Waterreus, Andrew J. | |
| dc.contributor.committeeMember | Huang, Yufei | |
| dc.contributor.committeeMember | Alaeddini, Adel | |
| dc.creator.orcid | https://orcid.org/0000-0002-0897-4200 | |
| dc.date.accessioned | 2024-03-08T17:35:53Z | |
| dc.date.available | 2024-03-08T17:35:53Z | |
| dc.date.issued | 2019 | |
| dc.description | This item is available only to currently enrolled UTSA students, faculty or staff. To download, navigate to Log In in the top right-hand corner of this screen, then select Log in with my UTSA ID. | |
| dc.description.abstract | The classical "Pong" game resembles 2-player table tennis and was developed by Atari in 1972. In this game, each player controls a small paddle to bounce a ball across the rectangular area to defend a small goal on either side of the arena. Pong became extremely popular and is generally considered to be the first commercially successful video game. It has even earned itself a spot in the Smithsonian Institution in Washington, D.C. because of its level of cultural impact. This research project was an attempt to automate Pong through the use of two radically different machine learning methods, supervised learning and reinforcement learning. In supervised learning, an expert provides the training data, which consists of example input-output pairs to be used for the learning. In our case, a human controlled one paddle in response to the ball, while the other paddle moved up and down at a defined rate of 2.64 seconds per cycle. Then, an artificial neural network with four layers was trained from this dataset. In reinforcement learning, by using a reward system developed to encourage the paddle to defend the goal by bouncing the ball and receiving points, a controller agent was trained using Deep Q-Neural Networks (DQN). This method allows the computer to teach itself through trial and error. The supervised learning method generated an automated paddle that was deemed unbeatable by several challengers; while the reinforcement learning method was only capable of producing a controller agent with an average of 3-5 ball bounces per episode. | |
| dc.description.department | Mechanical Engineering | |
| dc.format.extent | 44 pages | |
| dc.format.mimetype | application/pdf | |
| dc.identifier.isbn | 9781088389850 | |
| dc.identifier.uri | https://hdl.handle.net/20.500.12588/6134 | |
| dc.language | en | |
| dc.subject.classification | Computer engineering | |
| dc.subject.classification | Computer science | |
| dc.subject.classification | Engineering | |
| dc.title | Utilization of Supervised and Reinforcement Learning in the Automation of the Classical Atari Game "Pong" | |
| dc.type | Thesis | |
| dc.type.dcmi | Text | |
| dcterms.accessRights | pq_closed | |
| thesis.degree.department | Mechanical Engineering | |
| thesis.degree.grantor | University of Texas at San Antonio | |
| thesis.degree.level | Masters | |
| thesis.degree.name | Master of Science |
Files
Original bundle
1 - 1 of 1
No Thumbnail Available
- Name:
- Waterreus_utsa_1283M_12943.pdf
- Size:
- 1.15 MB
- Format:
- Adobe Portable Document Format