Proceeding of the Faculty of technical Sciences, Novi Sad

Electrotechnical and Computer Engineering

Vol. 34 No. 01 (2019): Proceedings of the Faculty of Technical Sciences

TRAINING A REINFORCEMENT LEARNING AGENT TO PLAY ROAD FIGHTER

  • Ivan Radosavljević
DOI:
https://doi.org/10.24867/01BE47Radosavljevic
Submitted
December 21, 2018
Published
2018-12-21

Abstract

This paper presents the application of reinforcement learning for the purpose of training an agent capable of playing the video game Road Fighter. The agent is implemented as a Deep Q-Network and is trained via the Deep Q-Learning algorithm. The only data used for training the agent were screenshots of the game. After 10000 episodes of training the performance of the agent was comparable to that achieved in related approaches under similar hardware constraints.

References

[1] V. Mnih et al., “Playing Atari with Deep Reinforcement Learning,” p. 9.
[2] D. Silver et al., “Mastering Chess and Shogi by Self-Play with a General Reinforcement Learning Algorithm,” arXiv:1712.01815 [cs], Dec. 2017.
[3] D. Silver et al., “Mastering the game of Go without human knowledge,” Nature, vol. 550, no. 7676, pp. 354–359, Oct. 2017.
[4] “OpenAI Five,” OpenAI Blog, Jun-2018. Dostupno na: https://blog.openai.com/openai-five/.
[5] “Nestopia - NES/Famicom Emulator.” Dostupno na: http://nestopia.sourceforge.net/.
[6] M. Satran, “Programming reference for Windows API.”. Dostupno na: https://docs.microsoft.com/en-us/windows/desktop/api/.
[7] S. Ioffe and C. Szegedy, “Batch Normalization: Accelerating Deep Network Training by Reducing Internal Covariate Shift,” arXiv:1502.03167 [cs], Feb. 2015.
[8] “PyTorch dokumentacija.” Dostupno na: https://pytorch.org/docs/stable/index.html.
[9] H. van Hasselt, A. Guez, and D. Silver, “Deep Reinforcement Learning with Double Q-Learning,” p. 7.
[10] A. Paszke, “Reinforcement Learning (DQN) Tutorial — PyTorch Tutorials.” Dostupno na: https://pytorch.org/tutorials/intermediate/reinforcement_q_learning.html
[11] P. J. Huber, “Robust Estimation of a Location Parameter,” Ann. Math. Statist., vol. 35, no. 1, pp. 73–101, Mar. 1964.
[12] D. P. Kingma and J. Ba, “Adam: A Method for Stochastic Optimization,” arXiv:1412.6980 [cs], Dec. 2014.