Reinforcement learning s aplikací ve hrách

Abstract

The aim of this thesis is to analyse and validate the methods of approches used in reinforcement learning with application in games. The first part of the thesis describes the theoretical introduction to reinforcement learning, including a description of individual methods that are implemented as part of the practical task. The second part of the thesis is dedicated to the implementation task. Experiments are performed in an application developed in the Python programming language. The PyGame framework is used for the purpose of visualising the game environment. Working with artificial neural networks is handled using the PyTorch framework. The essence of the application is to collect coins generated in a maze-like environment. The conclusion of the thesis summarizes the observations and evaluates the experiments performed with individual methods, the results of which are represented by graphs of the reward obtained by the agent.