Reinforcement Learning Algorithms in the Computer Game Flappy Bird

Author: Bc. Martin Glova
Supervisor: doc. RNDr. Gabriela Andrejkova, CSc.
Co-supervisor: RNDr. Lubomir Antoni, PhD.

Objectives

(1) To process already known results which use reinforcement learning algorithms for the computer game Flappy Bird.

(2) To investigate the already known reinforcement learning algorithms to apply them for the game Flappy Bird. Introduce new algorithms for the game focusing on the applicability of the neural networks.

(3) To implement the introduced algorithms and compare them with the known reinforcement learning algorithms for the game Flappy Bird focusing on the quality and the time complexity of the learning.

Bibliography

(1) N. Appiah and S. Vare. “Playing FlappyBird with Deep Reinforcement Learning”. In: Technical Report (2016).

(2) K. Chen. “Deep Reinforcement Learning for Flappy Bird”. In: Technical Report (2015).

(3) G. A. Rummery and M. Niranjan. “On-Line Q-Learning using connectionist systems”. In: CUED/F-INFENG/TR 166 (1994).

(4) Sarvagya Vaish. Flappy Bird RL. URL : http://sarvagyavaish.github.io/FlappyBirdRL/.

(5) Sourabh Verma. FlapPy Bird. URL : https://github.com/sourabhv/FlapPyBird.

(6) M. H. Hassoun: Fundamentals of artificial neural networks. MIT Press, Cambridge, 1995.

(7) S. Rogers, M. Girolami: A First Course in Machine Learning. Chapman & Hall/CRC, 2012.

Progress

Processing of the related works
Theoretical part
Own proposed solution
Implementation of the solution

To Do List

Topic Chosen
Q-learning Algorithm Analyzed
First Optimal Values Found
First Draft Written
New Algorithms Analyzed
New Code Prepared
Neural Network Analyzed
Second Draft Written
Different Parameter's Values Tested
Neural Network Optimized
New Ideas Implemented
Conclusion and Future Works Written
Printing Thesis
Handing Thesis
Defending Thesis

Email address: