CS 4850 Senior Research Project:

Reinforcement Learning for AI Game-Playing Agents

Full disclosure: my primary goal in supporting CS 4850 research projects is to identify and develop strong candidates for the PhD program.

Research Project Title

Reinforcement Learning for AI Game-Playing Agents

Project Overview

Write a reinforcement learning-based game playing agent that improves upon existing work and enter it in a competition such as AI Areana using DeepMind's PySC2, or something like the new Kaggle Game Arena (which is pretty new and sparse at the moment), or work towards a publishable result on a benchmark such as Starcraft Multi-Agent Challenge V2.

Depending on student interests and schedule, we may look for another competition, like the ones at AAMAS. For example, see AAMAS 2024 Competitions. Or just try something novel with one or more of the games in the Arcade Learning Environment.

Preferred prerequisite knowledge:

Python, Numpy, Linux, AI, machine learning

Skills/Knowledge that students will learn:

Python, Numpy, Linux, JAX, PyTorch, Keras, AI, machine learning, reinforcement learning, research, and writing.

Major Work, Milestones and Expected Outcome

• Code that implements a game-playing agent
• Code that collects performance metrics
• Analyses of performance metrics
• Research paper suitable for submission

Ideas For Final Systems / Deliverables To Develop:

• Game-playing agent
• Performance analyses
• Submission-quality research paper

Recommended Team Size

Two

Maximum Number of Teams Accepted For This Topic

One

References/Tutorials For Project

TBD

Project Publication and Procedure

1. Is this request in support of a research project? If yes, will the students get credit when the research is published?

Yes, and you will be an author! I will help you write the paper and, most importantly, teach you how to write it.

2. Is a Non-Disclosure Agreement (NDA) required?

No.

3. Are you available to meet with the students on a regular basis (schedule to be determined by you and the team) and provide feedback on student progress?

Yes, and regular meetings will be required. Ideally you will join my research lab's weekly meeting. My ulterior motive is to identify strong candidates for the PhD program.

4. Do you want the student team to sign up for your project or would you prefer to interview them and select the team?

I prefer to interview candidates.

5. May students apply to, and present this project to, C-Day at the end of the semester?

Yes! I will help you with the poster and demonstration, if applicable.

Contact Information

Christopher.Simpkins@kennesaw.edu