Cohesive Hierarchical AI At Multiple Levels Of Abstraction
Complex games that operate on multiple levels of domains or abstractions - for example macro and micro, economy, warfare, and scientific progression - can be hard to create satisfactory AI for. On its own, each domain may be within the realms of possibility, but tied together, the complexity of the task increases greatly, to the point where even turn-based games can take enough time to calculate next actions that it directly impacts the pace atwhich the user is able to play.
My research aims to address this by building that complex whole out of composing simpler, more easily-trained modules, allowing for flexibility in the kinds of AI modules involved, tightly-focused training, and efficient usage of CPU time through selection of modules. In the same way that a soldier knows very well how to prevail physically, and a general where to move said soldier for maximum effect. Were either to have tried learning both at the same time, it would be much harder for either to have reached the same level of proficiency, but separated and communicating, they can leverage both skills, with focused training.
Having at the time of writing recently acquired a BSc in Computer Games, I’ve always had the interest of solving problems from the ground up, in as elegant and usable a way as possible. This drove many of my projects in university, from a custom Entity-Component game framework, to a library to support incredibly complex and interacting rulesets that change throughout play. Even in college, I’d used libGDX to procedurally generate flowers on Android, as part of a children's game I was doing for coursework.
I’ve had experience programming an automated testing tool for a medical web service, and have also worked as a Peer-Assisted Learning (PAL) mentor, essentially a lab assistant helping undergraduates in both Object-Oriented Programming and Mathematics for Computing.
Home institution: Essex
Supervisor: Professor Richard Bartle
Ready to apply?
Once you have identified your potential supervisor, we would encourage you to contact them to discuss your research proposal.Learn More