Monte-Carlo Tree Search for the Simultaneous Move Game Tron
Monte-Carlo Tree Search for the Simultaneous Move Game Tron
复制标题
同步移动游戏 Tron 的蒙特卡洛树搜索
DOI：
--
复制
发表时间：
2012
期刊：
影响因子：
--
通讯作者：
M. Winands
中科院分区：
文献类型：
--
作者：
N. D. Teuling;M. Winands
蒙特-卡罗树搜索（MCTS）已成功应用于许多游戏，特别是围棋。在本文中，我们研究了MCTS在Tron中的性能，这是一个两个玩家同时移动游戏。我们试图通过对MCTS的选择、扩展和播放阶段进行几项增强来增加MCTS程序在Tron游戏中的播放强度。根据实验结果，我们可以得出结论，渐进偏差，改变扩展阶段和播放截止都增加了整体播放强度，但结果不同的板。MCTS-Solver似乎是Tron游戏中MCTS的可靠替代品，并且由于其能够搜索状态空间以证明获胜而优于MCTS。MCTS程序的性能仍然不如最好的αβ程序a1 k 0 n，后者使用了复杂的评估函数，这表明还有相当大的改进空间。
Monte-Carlo Tree Search (MCTS) has been successfully applied to many games, particularly in Go. In this paper, we investigate the performance of MCTS in Tron, which is a two-player simultaneous move game. We try to increase the playing strength of an MCTS program for the game of Tron by applying several enhancements to the selection, expansion and play-out phase of MCTS. Based on the experiments, we may conclude that Progressive Bias, altered expansion phase and play-out cut-off all increase the overall playing strength, but the results differ per board. MCTS-Solver appears to be a reliable replacement for MCTS in the game of Tron, and is preferred over MCTS due to its ability to search the state space for a proven win. The MCTS program is still outperformed by the best αβ program a1k0n, which uses a sophisticated evaluation function, indicating that there is quite some room for improvement.