Primjena podržanog učenja za automatsko igranje igara

undergraduate thesis

• Title (english): Reinforcement Learning for Automatic Game Playing
• Author: Daniel Rakovac
• Mentor: Marko Čupić (mentor)
• Committee member: Marko Čupić (predsjednik povjerenstva)
• Committee member: Martin Tutek (član povjerenstva)
• Committee member: Karlo Knežević (član povjerenstva)
• Granter: University of Zagreb; Faculty of Electrical Engineering and Computing; Zagreb
• Defense date and country: 2018-07-13, Croatia
• Scientific / art field, discipline and subdiscipline: TECHNICAL SCIENCES; Computing

Abstract

U ovom je radu obrađeno podržano učenje kao vrsta strojnog učenja čiji je cilj maksimizirati ukupnu nagradu. Objašnjena je uloga neuronskih mreža u podržanom učenju te proces treniranja neuronskih mreža. U ovom se radu primjena podržanog učenja fokusirala na automatskom igranju igara. Za početak je odabrana jednostavna igra s malim brojem stanja, nazvana FrozenLake, gdje je automatsko igranje ostvareno pomoću algoritma SARSA i Q-učenje sa spremanjem Q-vrijednosti u strukturu nazvanu Q-tablica. Kao primjer složenije igre je odabrana igra FlappyBirds s opcijom da su prepreke pomične. Problem je riješen pomoću Q-učenja aproksimacijom Q-vrijednosti pomoću neuronske mreže. Rezultati algoritama su grafički prikazani i uspoređeni.

Abstract (english)

This paper describes reinforcement learning as a type of machine learning aimed
at maximizing total reward. It explains the role of neural networks in reinforcement learning and their training process. To start with, a simple game with a small number of
states was chosen, named FrozenLake, where automatic game playing was accomplished with the help of SARSA algorithm and Q-learning where Q-values were stored in
a structure called a Q-table. As an example of a more complex game, the FlappyBirds
game was selected with the option to have moving obstacles. The problem was solved
with the help of Q-learning where Q-values were approximated using neural network.
The results of algorithms are graphically displayed and discussed.

• Keywords: strojno učenje; podržano učenje; automatsko igranje igara; Markovljev proces odlučivanja; Q-učenje; SARSA; neuronske mreže
• Keywords (english): machine learning; reinforcement learning; automatic game playing; Markov decision process; Q-learning; SARSA; neural networks
• Language: croatian
• URN:NBN: urn:nbn:hr:168:251759
• Study programme: Title: Computing; Study programme type: university; Study level: undergraduate; Academic / professional title: sveučilišni/a prvostupnik/ prvostupnica (baccalaureus/baccalaurea) inženjer/inženjerka računarstva (sveučilišni/a prvostupnik/ prvostupnica (baccalaureus/baccalaurea) inženjer/inženjerka računarstva)
• Type of resource: Text
• File origin: Born digital
• Access conditions: Closed access
• Created on: 2019-03-14 00:21:10