20.500.12556/RUL-175114 Razvoj strategij umetne inteligence za igranje namiznega nogometa v pravem okolju Development of artificial intelligence policies for playing table football in a real environment Delo se ukvarja z razvojem oz. učenjem strategij inteligentnih agentov za igranje namiznega nogometa, ki so sposobni igrati proti človeku na pravi avtomatizirani mizi namiznega nogometa. V delu je bil razvit simulator namiznega nogometa, ki je služil učenju in preizkušanju strategij. Za boljši prenos iz simuliranega okolja v pravo okolje je bila izvedena identifikacija igralnih palic, v okviru katere smo na podlagi posnetkov gibanja pravih palic določili parametre simuliranih palic. Vpliv drugih odstopanj, kot so časovne zakasnitve, smo zmanjšali med razvojem posameznih strategij. Pred razvojem strategij je bil izdelan tudi nasprotnik, ki je bil implementiran v obliki končnega avtomata. Nasprotnik je vključeval vse palice nasprotne ekipe, pri čemer je bila vsaka palica svoj agent. Nasprotnik je služil boljši generalizaciji nekaterih strategij za igro proti človeškim nasprotnikom. S spodbujevalnim učenjem so bile razvite strategije za palico vratarja, obrambno palico in palico napadalca. Strategije so bile vse naučene v simulaciji. Dodatnega učenja v pravem okolju nismo izvedli. Strategije so bile najprej preizkušene v simulaciji, kjer smo ovrednotili sposobnost branjenja lastnega gola (strategiji vratarja in obrambe) oz. sposobnost streljanja na nasprotnikov gol (strategija napadalca). Vse strategije skupaj, z dodatkom palice srednje vrste, ki jo je vodil enak agent kot tisti, ki je vodil nasprotnika, so bile na koncu preizkušene v pravem okolju proti več človeškim parom. This thesis explores the development of intelligent agent policies using reinforcement learning, for playing table football against human opponents on a real automated table football table. First, a simulator was developed, whose purpose was both training and testing of policies. For the purposes of simulation-to-real transfer, identification of player rods was performed, where measurements of real rod responses were used to fit simulation parameters. Other discrepancies were compensated during training of individual policy. For better generalization against human opponents, an artificial opponent was developed, which was active during the training process of several policies. The artificial opponent is based on state-machine logic. Three policies were developed using reinforcement learning: a goalkeeper policy, a defense policy and an attacker policy. The midfield rod was made to use the same state-machine logic as the opponent. The developed policies were trained only in simulation, without additional training in the real environment. First, the policies were individually tested in simulation, where we evaluated their ability to defend their own goal (goalkeeper and defense policies) and to score into the opponent's goal (attacker policy). In the end, all policies, with the addition of the midfield state-machine agent, were evaluated against several human opponent pairs in the real environment. spodbujevalno učenje namizni nogomet simulacija prenos v pravo okolje agenti umetna inteligenca. reinforcement learning table football simulation sim-to-real transfer agents artificial intelligence. true false false Slovenski jezik Angleški jezik Magistrsko delo/naloga 2025-10-16 14:30:09 2025-10-16 14:30:20 2025-12-06 03:49:57 0000-00-00 00:00:00 2025 0 0 0000-00-00 NiDoloceno NiDoloceno NiDoloceno 0000-00-00 0000-00-00 0000-00-00 63058 258024963 Hozic_David_-_Razvoj_strategij_umetne_inteligence_za_igranje_namiznega_nogometa_v_pravem_okolju.pdf Hozic_David_-_Razvoj_strategij_umetne_inteligence_za_igranje_namiznega_nogometa_v_pravem_okolju.pdf 1 7FCD1461A450D7810CDFB0339D100431 eb5b6da0ee0d6e4761712c93b66ac982ccfab47c740bee1191a2c9817baf0e02 a09d505c-aa8b-11f0-9328-0050569b8976 https://repozitorij.uni-lj.si/Dokument.php?lang=slv&id=220395 Fakulteta za elektrotehniko 0 0 0