Using DQN and Double DQN to Play Flappy Bird

DOI

10.2991/978-94-6463-010-7_120How to use a DOI?

Keywords

Deep Reinforcement Learning; Deep Q Learning; Double Q Learning; Artificial Intelligence

Abstract

Reinforcement Learning (RL), which is mainly used to solve sequential decision making problem, is an important branch in machine learning. Deep Learning (DL) also plays a leading role in the field of artificial intelligence. It uses neural network to approximate nonlinear functions. Deep Reinforcement Learning (DRL) is an algorithm framework combining reinforcement learning and deep learning that absorbs both advantages, and this DRL is capable of helping training agent learn how to play video games. Among them, the Deep Q Network (DQN) plays an important role. However, DQN will cause overestimation of values, and Double Deep Q Network (DoubleDQN) comes out and is used to fix this problem. The author presents the study of how DQN and DoubleDQN work and the difference between the result and training loss when these two algorithm were implemented in a video game called Flappy Bird in pycharm by using Keras (a neural network API in Python designed for deep learning problems). The author also adds improvement in DQN model to accelerate the training speed and compares it with the results of other people’s experiments. After experiments, as expected, the modified DQN works better than traditional DQN but not as good as Double DQN. The training loss graph depicts that Double DQN decreases training loss present it is a good way to solve the overestimation problem.

Copyright

© 2023 The Author(s)

Open Access

Open Access This chapter is licensed under the terms of the Creative Commons Attribution-NonCommercial 4.0 International License (http://creativecommons.org/licenses/by-nc/4.0/), which permits any noncommercial use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license and indicate if changes were made.

Download article (PDF)

TY  - CONF
AU  - Kun Yang
PY  - 2022
DA  - 2022/12/02
TI  - Using DQN and Double DQN to Play Flappy Bird
BT  - Proceedings of the 2022 International Conference on Artificial Intelligence, Internet and Digital Economy (ICAID 2022)
PB  - Atlantis Press
SP  - 1166
EP  - 1174
SN  - 2589-4919
UR  - https://doi.org/10.2991/978-94-6463-010-7_120
DO  - 10.2991/978-94-6463-010-7_120
ID  - Yang2022
ER  -