Samenvatting

This thesis explores how the problem of players wanting to test out and practice the mechanics they were
taught in the tutorial can be solved. The objective of this research was to discover how AI can be used to
increase user engagement in this Naval Warfare game, resulting in more interest in Thales as an employer.
Iterating on this yielded several research questions about Machine Learning and “Scripted” AI. Specifically,
these questions dive into the matter of implementing a form of teamwork in both implementations, as well
as how the agent behaviour can adapt to player performance, and how action variety can be
accomplished in the Scripted AI implementation.
To structure this research, the framework Design Thinking was used, in combination with extensive desk
research and continuous playtesting throughout the research duration. In the initial research weeks, as well
as for each implementation and the eventually chosen AI version, playtests were performed with the target
group.
The conclusion that can be drawn from the playtests is that a form of Scripted AI, specifically a Behaviour
Tree, is the best option for this prototype. It is the most developer-friendly and yet allows for enough
opportunities to increase user engagement. Machine Learning fell short in several areas, including ease of
use and technical limitations in various aspects. That is why, for future teams working on this project, it is
recommended that Behaviour Trees be used as the foundation for NPC behaviour. The game could receive
new features such as air-based NPCs, a greater action variety for all NPCs, sub-objectives in the Endless
Game mode and a harbour that serves as a way to replenish and upgrade the player’s battle ship.