This Bachelor thesis analyzes evolutionary algorithms in changing environment problems. The purpose of this research was to develop a methodology for creating a self-operating game controller based on the principles of evolutionary algorithms. In this work, the collection of information regarding artificial neural networks, compositional pattern producing networks and evolutionary algorithms: genetic algorithms, neuroevolution and its algorithms NEAT and HyperNEAT is presented, followed by ideas and the implementation of self-operating game controller, along with experiments to determine its effectiveness. The methodology helped to create a game controller, which controls a copy of „Chrome Dino“ game. The choice of the experimental game was influenced by the fact that it met the criteria of changing environment problems. Presented results of experiments confirms the success of this implementation. The effectiveness of the game controller’s player management to successfully play the game shows, that evolutionary algorithms can be very useful when faced with changing environment problems. Furthermore, the provided code base also can act as a tool for creating other projects regarding game or simulation controllers.
Evoliuciniai algoritmai besikeičiančios aplinkos uždaviniams ; Evolutionary algorithms in changing environment problems
2020-06-09
Hochschulschrift
Elektronische Ressource
Lithuanian , Englisch
DDC: | 629 |