Reinforcement Learning Toolkits for Gaming: A Comparative Qualitative Analysis  

作　　者：Mehdi Mekni Charitha Sree Jayaramireddy Sree Veera Venkata Sai Saran Naraharisetti Mehdi Mekni;Charitha Sree Jayaramireddy;Sree Veera Venkata Sai Saran Naraharisetti(Tagliatela College of Engineering, University of New Haven, West Haven, USA)

机构地区：[1]Tagliatela College of Engineering, University of New Haven, West Haven, USA

出　　处：《Journal of Software Engineering and Applications》2022年第12期417-435,共19页软件工程与应用（英文）

摘　　要：Historically viewed as a niche economic sector, gaming is now projected to exceed a global annual revenue of $218.7 billion in 2024, taking advantage of recent Artificial Intelligence (AI) advances. In recent years, specific AI techniques namely;Machine Learning (ML) and Reinforcement Learning (RL), have seen impressive progress and popularity. Techniques developed within these two fields are now able to analyze and learn from gameplay experiences enabling more interactive, immersive, and engaging games. While the number of ML and RL algorithms is growing, their implementations through frameworks and toolkits are also extensive too. Moreover, the game design and development community lacks a framework for informed evaluation of available RL toolkits. In this paper, we present a comprehensive survey of RL toolkits for games using a qualitative evaluation methodology.Historically viewed as a niche economic sector, gaming is now projected to exceed a global annual revenue of $218.7 billion in 2024, taking advantage of recent Artificial Intelligence (AI) advances. In recent years, specific AI techniques namely;Machine Learning (ML) and Reinforcement Learning (RL), have seen impressive progress and popularity. Techniques developed within these two fields are now able to analyze and learn from gameplay experiences enabling more interactive, immersive, and engaging games. While the number of ML and RL algorithms is growing, their implementations through frameworks and toolkits are also extensive too. Moreover, the game design and development community lacks a framework for informed evaluation of available RL toolkits. In this paper, we present a comprehensive survey of RL toolkits for games using a qualitative evaluation methodology.

关 键 词：Game Design & Development Machine Learning Reinforcement Learning Deep Learning 

分 类 号：G89[文化科学—体育学]