Calculating Optimal Jungling Routes in DOTA2 Using Neural Networks and Genetic Algorithms
Calculating Optimal Jungling Routes in DOTA2 Using Neural Networks and Genetic Algorithms
复制标题
使用神经网络和遗传算法计算 DOTA2 中的最佳丛林路线
DOI：
--
复制
发表时间：
2014
期刊：
影响因子：
--
通讯作者：
Tom Batsford
中科院分区：
文献类型：
--
作者：
Tom Batsford
在本文中，提出使用学习算法训练的神经网络来解决多人在线竞技场 (MOBA) 游戏中较难的活动之一；丛林。 实施的目的是找到一条围绕丛林的最佳路线，然后玩家和机器人 AI 在 Dota2 中战斗时都可以使用该路线。 为此，我们实现了 Dota2 丛林的准确表示，并使用前馈 S 形神经网络，结合各种遗传算法技术来进行所有决策。 结果表明，已经达到了向最佳路线的一些收敛，但是在实现职业玩家的准确表现之前，模拟器本身还需要做更多的工作。
In this paper the use of a Neural Network trained using Learning Algorithms is proposed to tackle one of the harder activities within a Multiplayer Online Battle Arena (MOBA) game; Jungling. The aim of the implementation was to find an optimal route around the jungle that can then be used by both players and bot AI whilst fighting in Dota2. To do this an accurate representation of the Dota2 Jungle was implemented and a Feed Forward Sigmoidal Neural Network used, in combination with various Genetic Algorithm techniques, for all decision making. Results show that some convergence toward an optimal route has been reached, however more work needs to go into the simulator itself before an accurate representation of a professional player can be achieved.