Implementace jednoduché hry metodami reinforcement learning

This bachelor’s thesis focuses on the design, implementation, and experimental validation of an agent based on the Deep Q-Network (DQN) algorithm for the simulated LunarLander-v3 environment from the Gymnasium library. The introductory section summarizes the fundamentals of reinforcement learning and explains the key principles of DQN. The thesis then describes the environment, the landing procedure, the proposed neural-network architecture, and the methodology for defining success metrics. The practical chapter details the hyperparameter tuning process and presents the results of the most successful model. The work is accompanied by publicly available source code that enables straightforward reproducibility and provides a foundation for future extensions.