Institutional Repository of Vilnius University
Dinaminio programavimo taikymai Tetris žaidimui spręsti /
Abstract
dc:descriptionTetris is a popular computer game which is widely used to benchmark various optimization methods. One of such optimization techniques is dynamic programming. In scientific literature that researches the game of Tetris there are a variety of dynamic programming methods which try to create the best performing agent. It is oſten the case that the results of the proposed method are only compared to the results reported in other research papers without testing them in the same environment. This highlights the problem that the literature lacks comparative type of works which would identify which method properties allow to create better performing agents. For this reason in this paper a literature analysis of dynamic programming applications for the game of Tetris is performed and a hierarchical chart is created which contains the relations of method properties. Using this chart three temporal difference‑based methods are chosen which allow to compare two method properties: batch and stochastic updates and also forward and backward returns. The experiments performed in the Tetris environment suggest that the use of stochastic updates allowed to create better performing agents than batch updates. Also, experiments show that agents using forward returns performed better than the ones that used backward returns. Finally, all three temporal difference‑based methods were parameterized by parameter λ which controls how important are forward and backward returns to the current state. The performed experiments showed that this parameter does not have any significant effect on agent performance.
Degree
thesis:*- Grantor dc:publisher
- Institutional Repository of Vilnius University
- Year dc:date
- 2025
Author and committee
dc:creator, dc:contributor.*- Author dc:creator
-
- Mileika, Justas,
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:ELABAETD239031795&prefLang=en_US