Games are commonly used as playground for AI research, specifically in the field of Reinforcement Learning (RL). RL has shown promising results in developing intelligent agents to play a multitude of games. Previous work have explored how RL agents can be used in the process of playtesting in game development. This thesis investigates different aspects of the RL algorithm Deep Q-Network (DQN) that learns to play the match-three game Candy Crush Friends Saga (CCFS). This thesis also investigates two of the challenges in applying RL in the context of CCFS. First, different sampling strategies are explored to speed up the training of a DQN-based agent. With inspiration from Imitation Learning (IL), demonstrations of game play are incorporated in the DQN algorithm to speed up the training. Another challenge when doing research in RL is volatility in target metrics during the training phase. In this thesis, we investigate how different factors contribute to this volatility. We propose three metrics to assess the agent during the training phase. Our results show that some of these factors contribute more than others. We also show how incorporating demonstrations into the DQN algorithm speeds up the training when learning to play the game CCFS. Based on these findings, this work highlights approaches to speed-up the training of a DQNbased agent. In addition, we propose a few recommendations that enable more reliable and reproducible research to apply RL in match-three games. ; Spel är en vanligt förekommande lekplats för AI forskning, speciellt inom fältet Förstärkande Lärning (RL). RL har visat lovande resultat i att utveckla intelligenta agenter som spelar en mångfald av olika spel. Tidigare arbeten har utforskat hur RL agenter kan användas i speltestningsprocessen i spelutveckling. Den här uppsatsen undersöker olika aspekter av RL algoritmen Djupt Q-Nätverk (DQN) som lär sig att spela match-tre spelet Candy Crush Friends Saga (CCFS). Två utmaningar undersöks i användandet av RL i kontext till CCFS. Först utforskar ...
Scaling Reinforcement Learning Solutions For Game Playtesting ; Skalbar förstärkningsinlärning för speltestning
2020-01-01
Theses
Electronic Resource
English
DDC: | 629 |
Game Theory and Reinforcement Learning based Smart Lane Change Strategies
SAE Technical Papers | 2022
|