Welcome! This (in-progress) project implements a custom capture-the-flag environment for multi-domain (ground + air) swarms. It furthermore includes a shared situational awareness layer per swarm, enabling individual agents within the swarm to have a global understanding of the environment. This project also supports multi-agent reinforcement learning, with baseline policies and reward functions implemented.
A high-level overview can be found here: report.pdf. Demo videos can be found in the /demos folder.
- Implement and release simulation code
- Release RL baseline demos + training code