Institutional Repository of Vilnius University
Application and comparison of reinforcement learning algorithms in doom game agent control /
Abstract
dc:descriptionIn this thesis, the use of reinforcement learning algorithms for controlling a Doom game agent was investigated. The paper describes the theory, main ideas, and methods of reinforcement learning. The Doom game and ViZDoom platform are reviewed. Three algorithms were chosen and compared: DQN, PPO, and A2C. Two custom Doom maps were created for testing the algorithms. Each map had different scenarios with fixed or random goals and with or without enemies. The algorithms were trained and evaluated using success rate, mean reward, and episode length. The results were compared to see how the algorithms performed in different Doom game environments.
Degree
thesis:*- Grantor dc:publisher
- Institutional Repository of Vilnius University
- Year dc:date
- 2026
Author and committee
dc:creator, dc:contributor.*- Author dc:creator
-
- Kavoliūnas, Dovydas,
Subjects
dc:subject × 1Rights
dc:rights- Statement dc:rights
-
- info:eu-repo/semantics/openAccess
- Language dc:language
- eng
Identifiers
dc:identifier.*- Repository record dc:identifier
- https://repository.vu.lt/VU:ELABAETD308337337&prefLang=en_US