Fighting Fire with AI: Using Reinforcement Learning to Train AI in a Variable Multi-Actor Strategy Game
Modern reinforcement‑learning research often relies on a fixed set of actions. However, many real‑world problems such as robotic rescue, autonomous combat support, and disaster response feature multiple actors coordinated by a single decisionmaker, each with a different role and a distinct set of permissible actions. To bridge this gap and research environments with multi-actors instead of fixed actions, we built a dynamic game simulation where an AI learns to manage multiple units with heterogenous capabilities.
Intern: Emery Que
Mentors: Zachary Goddard (FPS/KVM), Tanner Fowler (FPS/KVM)
_page-0001.jpg)
-[Autosaved].png)
.png)
.png)
