Projects
MEDEVAC Wargaming

MEDEVAC Wargaming

MEWI (Medical Evacuation Wargaming Initiative) — a first-of-its-kind, high-fidelity multiplayer simulation of tactical medical evacuation, run in the US Army’s Medical Evacuation Doctrine Course.

Multi-Agent
78%
more confident under time pressure
MEWI participant survey
66%
better grasp of medical regulating
MEWI participant survey

The problem

US military medical-evacuation planning means building a robust network of platforms and facilities to move and treat large numbers of casualties under fire. It is high-stakes and hard to rehearse: doctrine students learn the fundamentals, but had no high-fidelity way to practice the planning and real-time decisions under realistic pressure.

What we built

MEWI is a custom Unity multiplayer simulation of tactical-level medical evacuation across two scenarios — an amphibious assault in the Pacific (Operation Storm Surge) and a land conflict in Eastern Europe (Operation Eastern Crucible). Players coordinate multimodal air and ground evacuation platforms under fog of war, day/night cycles, and stochastic (Poisson) mass-casualty events. An adversarial AI opponent — trained with reinforcement learning (Proximal Policy Optimization) and modeled as an MDP — intelligently places ground obstacles and air-defense rings to delay patient movement, forcing players to adapt in real time.

Deployment & evaluation

MEWI ran across two iterations of the US Army’s Medical Evacuation Doctrine Course at Fort Rucker, Alabama. Impact was measured with a 15-question post-simulation Likert survey (cooperative decision-making, lessons learned, and simulation quality), external observer notes, and in-game performance data.

Results

Participation substantially improved uptake of medical-evacuation lessons and cooperative decision-making. 78.2% of participants agreed the simulation increased their confidence to make critical evacuation decisions under time pressure; 65.6% strongly agreed it helped them understand medical regulating — a task facilitators consider one of the hardest and a known weakness in operational units; and over 95% agreed it deepened their grasp of platform constraints, triage, and air–ground integration. 56.3% were very satisfied and 59.3% strongly agreed they would recommend it.

Stack

C#UnityReinforcement Learning (PPO)Multiplayer