Learning to Play an Aerial Combat Game with Reinforcement Learning

Wit, J.W. de (2017) Learning to Play an Aerial Combat Game with Reinforcement Learning. Bachelor's Thesis, Artificial Intelligence.

Preview	Text AI_BA_2017_JWDEWIT.pdf - Published Version Download (288kB) | Preview
	Text Toestemming.pdf - Other Restricted to Backend only Download (79kB)

Abstract

Reinforcement learning has been used previously to let agents learn to play games. We have created a game environment with planes, that have to learn to fly from rewards. They were successful in learning this using the Continuous Actor-Critic Learning Automaton (CACLA). We have extended the planes with the ability to shoot each other. The agents are unable to learn this very well, because the reward from shooting is delayed which makes it difficult to credit the actions leading to the reward correctly. Furthermore, we have explored different configurations of CACLA in a multiagent environment. We have tested configurations that we have named Personal, Shared and Team. They are all successful in learning to fly in the busier environment, learning approximately the same strategy for flying.

Item Type:	Thesis (Bachelor's Thesis)
Degree programme:	Artificial Intelligence
Thesis type:	Bachelor's Thesis
Language:	English
Date Deposited:	15 Feb 2018 08:29
Last Modified:	15 Feb 2018 08:29
URI:	https://fse.studenttheses.ub.rug.nl/id/eprint/15401

Actions (login required)

View Item