Institutional Repository of Vilnius University
Nidavellir stalo žaidimo agento kūrimas naudojantis Monte Karlo medžio paieška /
Abstract
dc:descriptionThis work describes application of the Monte Carlo tree search algorithm to the „Nidavellir“ board game and improvements to the Monte Carlo tree search based agent. „Nidavellir“ is an imperfect information game with elements of randomness and can be played by two to five players. This type of game challenges game agents because it has complex decision spaces and deep strategic depth. The Monte Carlo tree search algorithm was chosen as a result of literature analysis, as this algorithm is known for its success in various board games such as Go and chess. Strategies to improve the Monte Carlo tree search algorithm have also been selected, such as opponent move abstraction and progressive widening. The paper describes the process of agent development, refinement and optimisation and the process results. The developed agents were tested in different groups, with different numbers of players, in an attempt to measure the agents' ability to play the game. Statistical tests are performed to ensure the significance of the results. The tests performed show that the Monte Carlo tree search algorithm is able to play the game competently. It is also shown that the opponent move abstraction strategy is ineffective for this game and degrades agent ability to play. On the other hand, progressive widening strategy is able to improve the number of wins in a multi-player environment.
Degree
thesis:*- Grantor dc:publisher
- Institutional Repository of Vilnius University
- Year dc:date
- 2024
Author and committee
dc:creator, dc:contributor.*- Author dc:creator
-
- Višnevskis, Dominykas,
Rights
dc:rights- Statement dc:rights
-
- info:eu-repo/semantics/openAccess
- Language dc:language
- lit
Identifiers
dc:identifier.*- Repository record dc:identifier
- https://repository.vu.lt/VU:ELABAETD210648302&prefLang=en_US