Co-operative Game for Artificial Intelligence

Platform | PC, Academic Paper
Tools | Unity, ML-Agents
Duration | 6 Weeks
Team Size | 1
Roles | Designer, Programmer, Researcher

A solo research project which aims to study Machine Learning Agents and how to design games for them. I worked on designing and creating an environment to train these Agents to become co-operative with each other and work towards a common goal.


Game Design & Research:
- Adapted to designing for two unique intelligence agents, rather than the human intelligence.
- Prioritized the user experience of the AI in the way it interacts with the game, rather than what is considered typical UX.
- Designed reward conditions for the agents to receive and train behaviour off.
- Tested and iterated on the limitations of how the AI interacts with Unity, translating findings into actionable design.
- Conducted "user tests" with the AI, observing interactions and reverse engineering player behaviour.
- Created process documentation for both internal and external parties, mockups and findings.
- Investigated existing AI and interactions with Unity, designed taking previous experiments into account.
- Documented & reported research to supervisor, conducted technical design reviews.

Final research paper can be found here.