Dissertation Stealth ai

Being intrigued about the stealth game genre as a whole, I decided for my project to look into the AI that makes up the guards' behaviour. The primary question being researched was whether the AI should be designed to be smart or dumb to enhance the overall stealth experience. I found that dumb behaviours could be used for a more structured and linear implementation, whereas smart behaviours were more favoured to create a dynamic and more sporadic challenge.

​

The project included two mirroring levels, each with its respective AI. From this project, some of the key things implemented included behaviour trees, AI controllers, occupancy maps, splines, AI perception and EQSs. In the levels I investigated, I examined when and how the implemented systems can be effectively used and learned about the pros and cons of each. 

​

There were many key learning takeaways from this project. Even with the preconceived belief of AI being complex, this task further showed me how much goes into creating great NPCs. I also have a greater understanding of the thought processes and decision-making that go behind tailoring the AI's behaviour to prioritise the best gameplay experience for the player.