20.500.12556/DKUM-96000 Razvoj računalniške igre, podprte z umetno inteligenco magistrsko delo Development of an AI-Powered Computer Game V magistrskem delu smo se ukvarjali z izdelavo večplatforme igre, razvite v programskem jeziku Java s pomočjo ogrodja LibGDX. Glavni cilj te naloge je bila implementacija pametnega agenta, ki se uspešno odziva v okolju naše definirane igre. V okviru naloge smo izdelali vso grafiko, načrtali koncept igre, implementirali statične animacije, učinke delcev, dinamične animacije oziroma senčnike ter vzpostavili logiko igre. Za izdelavo statičnih animacij smo uporabili orodje Spine2D; vso grafiko smo narisali s pomočjo orodja Inkscape. Programsko kodo smo razvijali v orodjih Android Studio in IntelliJ IDEA Community Edition. S pomočjo knjižnice tf-agents in programskega jezika Python smo implementirali pametnega agenta, pri čemer smo ga učili s pomočjo simulacije igre in lastno implementiranim lokalnim strežnikom. V produkcijski različici smo agenta izvajali z izvoženim modelom nevronske mreže v programskem jeziku Java. Implementirano računalniško igro in pametnega agenta smo izčrpno preizkusili z več eksperimenti; delovanje agenta smo primerjali s človeškim igralcem. Ugotovili smo, da igra deluje nemoteno; pametni agent v večini primerov uspešno odigra računalniško igro, le malenkostno počasneje in manj učinkovito v primerjavi z izkušenim človeškim igralcem. The master’s thesis addresses the development of a cross-platform computer game created in the Java programming language using the LibGDX framework. The main objective was to implement an intelligent agent capable of responding to a pre-created simulated game environment. As part of this thesis, all required graphic assets, game design, static animations, particle effects, dynamic animations, e.g., shaders, and game logic implementation were created. The static animations were created using Spine2D, and all graphic assets were hand-drawn using the tool Inkscape. The programming code was developed in Android Studio and IntelliJ IDEA Community Edition. An intelligent agent was implemented with the support of the Python tf-agents library. It was trained using a game simulation and a custom implementation of a local server. In the production version, the intelligent agent was exported as a neural network model and deployed in Java. The implemented computer game and the intelligent agent were tested through multiple experiments, and the performance of the agent was compared with that of a human player. In conclusion, the game operates without issues. The intelligent agent successfully completes the game in most cases, though it is slightly slower and less successful than its human player counterpart in the experiments. računalniška igra LibGDX Java okrepitveno učenje pametni agent computer game LibGDX Java RL smart agent true true false [P. Urbanč] Slovenski jezik Angleški jezik Magistrsko delo/naloga 2025-11-19 14:58:57 2025-12-22 11:32:24 2026-02-03 03:14:16 2026-02-04 03:11:02 2025 0 Maribor 0 Maribor 1 spletni vir (1 datoteka PDF (XIII, 61 str.)) 0000-00-00 NiDoloceno NiDoloceno NiDoloceno 0000-00-00 0000-00-00 0000-00-00 004.96:004.8(043.2) 266989059 URN:SI:UM:DK:9ZAUPLWV MAG_Urbanc_Peter_2025.pdf MAG_Urbanc_Peter_2025.pdf 1 C59B90146D76A689E96E0BAE54C6C431 453d0f597f88b433330301e920b76ca03d0d64fde9015603adf199ad3d7b805c d01ff5fb-c576-11f0-865f-00155d000105 https://dk.um.si/Dokument.php?lang=slv&id=200772 MAG_Urbanc_Peter_2025.zip MAG_Urbanc_Peter_2025.zip 1 7B3EA1B489E2CB73ABB490E3EC043BFE 39bfe0828fac37fcfd054ed212510490e0ce36db7b8aa3753eab7ba2abc3d2f2 e0e6c8f7-c56b-11f0-865f-00155d000105 https://dk.um.si/Dokument.php?lang=slv&id=200769 Fakulteta za elektrotehniko, računalništvo in informatiko 0 0 0