A Study on Game Character Learning Based on Player Behavior
1
조회
0
다운로드
0
인용
0
참고문헌
키워드
초록
The autoplay system is a system that automatically plays the game with the one click of a button. Currently, almost all mobile games use this system, and some PC games also use this system. However, this autoplay system tends to play games inefficiently. In this paper, we propose artificial intelligence based on the player's behavior pattern to improve the disadvantages of autoplay system. The artificial intelligence model proposed in this paper stores game data and input button values when a player plays a game as learning data. This stored data was learned using DNN (Deep Neural Network) model. In the game, because the player repeatedly presses another button at the same time, we proceeded to use the multiple Output Layers. We loaded the learned data back into Unity3D and applied it to the character to check the result. It was found that the lap time and the movement path were different each time it was executed. In this paper, we record the results by using the artificial intelligence model proposed by 20 experimenters. Only the data of the player with the constant track without crashing against the wall was learned properly, and the data of the player who did hit the wall could not get the result because the character did not move properly. We also made a simple arcade game to compare reinforcement learning with our AI model. The performance was not as good as reinforcement learning, but the learning speed was about 10 times faster.
Abstract
The autoplay system is a system that automatically plays the game with the one click of a button. Currently, almost all mobile games use this system, and some PC games also use this system. However, this autoplay system tends to play games inefficiently. In this paper, we propose artificial intelligence based on the player's behavior pattern to improve the disadvantages of autoplay system. The artificial intelligence model proposed in this paper stores game data and input button values when a player plays a game as learning data. This stored data was learned using DNN (Deep Neural Network) model. In the game, because the player repeatedly presses another button at the same time, we proceeded to use the multiple Output Layers. We loaded the learned data back into Unity3D and applied it to the character to check the result. It was found that the lap time and the movement path were different each time it was executed. In this paper, we record the results by using the artificial intelligence model proposed by 20 experimenters. Only the data of the player with the constant track without crashing against the wall was learned properly, and the data of the player who did hit the wall could not get the result because the character did not move properly. We also made a simple arcade game to compare reinforcement learning with our AI model. The performance was not as good as reinforcement learning, but the learning speed was about 10 times faster.
참고문헌 (14)
- Rosenblatt, Frank, "The perceptron: a probabilistic model for information storage and organization in the brain", Psychological review, 65(6) (1958)
- Hinton, Geoffrey E, "A fast learning algorithm for deep belief nets", Neural computation, 18(7), pp.1527~1554 (2006)
- Mnih, Volodymyr, "Human-level control through deep reinforcement learning", Nature, 518(7540) (2015)
- 선민주, "‘Auto Play’ 를 통해 바라본 게임의 내러티브 변화 연구", 디지털콘텐츠학회논문지, 19(10), pp.1927~1936 (2018), 한국디지털콘텐츠학회[DOI]
- Patro S, "Normalization : A preprocessing stage", arXiv : 1503. 06462, ~ (2015)
- Karsoliya, Saurabh, "Approximating number of hidden layer neurons in multiple hidden layer BPNN architecture", International Journal of Engineering Trends and Technology, 3(6), pp.714~717 (2012)
- Juliani, Arthur, "Unity: A general platform for intelligent agents", arXiv:pp.1809.02627, ~ (2018)
- Artificial Intelligence and Creative Deep Learning GAN and Reinforcement Learning Technology Trends, "HA YEON", HA YEON, ~ (2017)
- Sutton, Richard, "Introduction to reinforcement learning, Vol. 135" (1998), MIT press