MCTS library for unit movement planning in real-time strategy game StarCraft
MCTS knihovna pro plánování pohybu jednotek v real-time strategické hře StarCraft
Authors
Supervisors
Reviewers
Editors
Other contributors
Journal Title
Journal ISSN
Volume Title
Publisher
České vysoké učení technické v Praze
Czech Technical University in Prague
Czech Technical University in Prague
Date
Abstract
Existuje společnost vývojářů umělé inteligence, kteří zkouší své nápady a pilně pracují, aby vytvořili neporazitelného protivníka pro živou strategickou hru Starcraft, což dokázali v Šachách a Go. Tato práce předvádí využití knihovny pro Monte CarloTree Search Considering durations algorytmus, který byl prvně navrhnut Albertem Uriarte a Santagem Ontańon z Drexelské univerzity. Tento algorytmus prokazuje vynikající výsledky v řízení armády v živých strategických hrách. Jako menší náhradu přidáme do knihovny vyhledávání Negamax. Naše využití algorytmu je vypracováno jako statická knihovna S++, která může být připojena k jakémukoli možnému botovi. Instalace je velmi jednoduchá a nenáročná. V průběhu práce vyhodnocujeme algoritmy, porovnáváme je a demonstrujeme jejich využití. Tyto algoritmy jsou založeny a testovány na platformě UAlberta bot.
There is a live community of AI developers that are trying their ideas and putting effort to create an unbeatable rival for real-time strategy game Starcraft, as it was done with Chess and Go. This work presents an implementation of the library for the Monte Carlo Tree Search Considering Durations algorithm, that was recently proposed by Alberto Uriarte and Santiago Onta~n´on from Drexel University. It appears to bring outstanding results in real-time strategy army control. As a smaller substitute, we add a Negamax search to the library. Our implementation of the algorithms is designed as a static C++ library, which could be easily plugged in-to any possible bot. The setup is simple and intuitive. During the course of the work we evaluate the algorithms, compare them and demonstrate their usage. The algorithms are based and tested on UAlberta bot framework.
There is a live community of AI developers that are trying their ideas and putting effort to create an unbeatable rival for real-time strategy game Starcraft, as it was done with Chess and Go. This work presents an implementation of the library for the Monte Carlo Tree Search Considering Durations algorithm, that was recently proposed by Alberto Uriarte and Santiago Onta~n´on from Drexel University. It appears to bring outstanding results in real-time strategy army control. As a smaller substitute, we add a Negamax search to the library. Our implementation of the algorithms is designed as a static C++ library, which could be easily plugged in-to any possible bot. The setup is simple and intuitive. During the course of the work we evaluate the algorithms, compare them and demonstrate their usage. The algorithms are based and tested on UAlberta bot framework.