Application of Reinforcement Learning in Games

Lunyuan Hu, Ruike Peng, Junpeng Yang

2024

Abstract

With the advancement of technology, non-player characters (NPCs) can respond to players' behaviors more intelligently, thereby enhancing the fun and challenge of games. In this regard, reinforcement learning is an algorithm suitable for training agents and improving the playability of games. This paper explores the innovation of artificial intelligence in the field of games, focusing on the application of reinforcement learning algorithms such as Q-Learning, policy gradient, and Actor-Critic in game development and deeply analyzes the practical application of reinforcement learning in classic games and the challenges it faces, such as overfitting problems, and explores the prospects for future algorithm improvements. Through reinforcement learning, game content and gameplay can be enriched and optimized, so that characters and environments can present more realistic and natural behavior patterns. In the future, with the continuous improvement of hardware performance and algorithms, Q-learning, policy gradient and Actor-Critic are expected to achieve personalization and dynamic adjustment in more complex game environments, promoting the continuous innovation and development of the game industry.

Paper Citation

in Harvard Style

Hu L., Peng R. and Yang J. (2024). Application of Reinforcement Learning in Games. In Proceedings of the 1st International Conference on Modern Logistics and Supply Chain Management - Volume 1: MLSCM; ISBN 978-989-758-738-2, SciTePress, pages 129-134. DOI: 10.5220/0013234900004558

in Bibtex Style

@conference{mlscm24,
author={Lunyuan Hu and Ruike Peng and Junpeng Yang},
title={Application of Reinforcement Learning in Games},
booktitle={Proceedings of the 1st International Conference on Modern Logistics and Supply Chain Management - Volume 1: MLSCM},
year={2024},
pages={129-134},
publisher={SciTePress},
organization={INSTICC},
doi={10.5220/0013234900004558},
isbn={978-989-758-738-2},
}

in EndNote Style

TY - CONF

JO - Proceedings of the 1st International Conference on Modern Logistics and Supply Chain Management - Volume 1: MLSCM
TI - Application of Reinforcement Learning in Games
SN - 978-989-758-738-2
AU - Hu L.
AU - Peng R.
AU - Yang J.
PY - 2024
SP - 129
EP - 134
DO - 10.5220/0013234900004558
PB - SciTePress