20.500.12556/DKUM-97650 Učenje agentov in primerjava obstoječih rešitev za igro MicroRTS na platformi GIANT magistrsko delo Agent training and comparison of existing solutions for the MicroRTS game on the GIANT platform Magistrsko delo obravnava razvoj in vrednotenje inteligentnih agentov za realnočasovno strateško igro MicroRTS, ki predstavlja poenostavljeno, a raziskovalno bogato okolje za testiranje metod umetne inteligence. Igro smo implementirali v pogonu Unity in jo integrirali v platformo GIANT, kar omogoča modularno izvajanje agentov, ponovljivo eksperimentiranje ter medsebojno primerjavo strategij. Glavni cilj naloge je bil razviti štiri nove agente, zasnovane z vedenjskimi drevesi, ter jih primerjati z izbranimi obstoječimi rešitvami. Učinkovitost agentov smo testirali na različnih mapah MicroRTS in jih ocenili z rangirnim sistemom TrueSkill. Rezultati kažejo, da vedenjska drevesa omogočajo pregledno, modularno in učinkovito zasnovo agentov, ki dosežejo konkurenčne rezultate v primerjavi z referenčnimi rešitvami. Integrirano okolje MicroRTS predstavlja uporabno raziskovalno osnovo za nadaljnji razvoj agentov in širitev eksperimentalnih scenarijev. Kljub omejitvam poenostavljenega okolja naloga ponuja celovit prispevek k razvoju razložljivih inteligentnih agentov v realnočasovnih strateških igrah. This master's thesis addresses the development and evaluation of intelligent agents for the real-time strategy game MicroRTS, which represents a simplified yet sufficiently rich research environment for testing artificial intelligence methods. The game was reimplemented in the Unity engine and integrated into the GIANT platform, enabling modular agent execution, reproducible experimentation, and comparison of strategies. The main objective of the thesis was to develop four new agents based on behavior trees and compare them with selected existing solutions. The agents’ performance was tested on different MicroRTS maps and evaluated using the TrueSkill rating system. The results show that behavior trees enable a transparent, modular, and effective design of agents that achieve competitive results compared to reference solutions. The integrated MicroRTS environment represents a useful research basis for further agent development and the expansion of experimental scenarios. Despite the limitations of the simplified environment, the thesis offers a comprehensive contribution to the development of explainable intelligent agents in real-time strategy games. MicroRTS Unity platforma GIANT vedenjska drevesa inteligentni agenti TrueSkill MicroRTS Unity GIANT platform behavior trees intelligent agents TrueSkill true true false [A. Omerzu] Slovenski jezik Angleški jezik Magistrsko delo/naloga 2026-03-29 18:14:46 2026-05-29 13:17:28 2026-06-17 03:07:47 0000-00-00 00:00:00 2026 0 Maribor 0 Maribor 1 spletni vir (1 datoteka PDF (IX, 50 str.)) 0000-00-00 NiDoloceno NiDoloceno NiDoloceno 0000-00-00 0000-00-00 0000-00-00 004.388.4:004.8(043.2) 281709571 MAG_Omerzu_Anze_2026.pdf MAG_Omerzu_Anze_2026.pdf 1 92D82A77271AD806F67A182EE098A271 c28908779d0ba2a1bd2e5a502a532b6ab49e065db4c5d13558d9611dbf53568e 82af44b9-2b99-11f1-be47-00155d0001a1 https://dk.um.si/Dokument.php?lang=slv&id=204878 Fakulteta za elektrotehniko, računalništvo in informatiko 0 0 0