Multi-Agent Capture the Flag was a project developed for the Machine Learning curricular unit at Iscte - University Institute of Lisbon. Make sure to check out Marco Rodrigues as he was the co-author for this project.

For a video of the end result, see this link.

Using Machine Learning we developed a system where we can train agents to partake in a simple fetch quest, as well as train opponents to add a challenge to this mechanic.

For a first phase we implemented our own version of the Q-Learning Reinforcement Learning algorithm. We also implemented the same system using Unity's ML Agents to compare to our own implementation.

// TODO