基于卷积神经网络的《黑暗之魂》游戏AI设计

近年来,电子游戏成为受广大年轻人喜爱的娱乐方式,而游戏开发者们也试图将人工智能,机器学习的研究成果应用于游戏产业。具体就体现在游戏人物的行为逻辑、游戏难度曲线自适应等方面。让虚拟角色的行为看起来更加接近于智慧生物。从而为玩家带来更好的代入感和成就感。对于游戏中画面数据的训练,我们使用了对于图像数据效果好的卷积神经网络模型,并改进了数据科学领域广泛应用的AlexNet。从而实现了模型训练效果的高效准确。通过FROM SOFTWARE开发的广受好评的动作角色扮演游戏《黑暗之魂》为例,展示如何利用深度学习技术让角色实现自动化和自适应的攻击与防御,塑造更加接近于人类行为的虚拟角色。

AI Design of Dark Soul Game Based on Convolutional Neural Network

In recent years,electronic games have become a favorite way of entertainment for young people,and game developers also try to apply the research results of artificial intelligence and machine learning to the game industry.It is embodied in the behavior logic of game characters,the adaptability of game difficulty curve and so on.Make the behavior of virtual characters look closer to intelligent creatures.So as to bring players a better sense of substitution and achievement.For the training of picture data in the game,we use the convolution neural network model with good effect on image data,and improve the Alex net widely used in the field of data science.Thus,the model training effect is efficient and accurate.This paper attempts to take the widely praised action role-playing game"soul of darkness"developed by from software as an example to show how to use deep learning technology to make the role realize automatic and adaptive attack and defense,and shape a virtual role closer to human behavior.