Improving Game AI Design Using Adversarial Agents
This project proposes a novel method for improving the quality of behaviour of human authored agents by pitting them against trained agents and observing what bad behaviours/exploits the trained agents reveal. Authored agents refer to AI agents whose actions are explicitly designed by programmers using traditional techniques such as Utility functions, Behaviour Trees and state machines; trained agents refer to agents whose behaviour is learned by playing many games against the authored agents.
After graduating with a MEng in Computer Science from the University of Bristol, Nathan joined the games industry as a programmer, working for Climax Studios and Freejam games, before moving to developing indie games.
Home institution: Queen Mary
Supervisor: Dr Jeremy Gow
Ready to apply?
Once you have identified your potential supervisor, we would encourage you to contact them to discuss your research proposal.Learn More