The first goal of my thesis, was the design and the implementation of the basics of a round based strategy game. The goal of this framework, is to provide an environment, for artificial agent testing. For this purpose, I built a dynamic system, which is easily extendable by adding new units, terrain objects, players. For the implementation I used the Java platform, because of my experience with this, and the usage of its 1.8 version’s innovation I was able to effectively accomplish the task.
The second goal is to build competing agents in this system, whom can fight against each other. The agents can be grouped into two larger groups and there is a smaller agent. The latter agent is a randomized one, which behaves fully random. The first larger group of agents are simpler, goal-based agents that observe their environment, and choose their actions according to those observations. The difference between these agents are their priorities. The other larger group of agents utilize Bayesian network based decision making. They perform inferences according to their networks, in order to choose their next action. The difference between these agents are the way they interpret their output.
In this thesis these agents are compared in different aspects. During the game several details are recorded such as the player pairs, i.e. who play against each other, the winner, game length, i.e. how long did it take the winner to win, and the average reaction time of the agents. In the thesis these data are used for the comparison.