Top-down 2D 5v5 bomb-defuse multi-agent RL simulator with live match viewer. Recurrent MAPPO + TarMAC emergent communication, no scripted strategies.

An intelligent strategic advisor for simulated battlefield environments. Automates the OODA loop using Gymnasium, LSTM-based intent prediction, and Monte Carlo Tree Search (MCTS) to recommend optimal Courses of Action (COA).

把传统自动战斗升级为自然语言代理指挥：玩家一句话接管关键战术决策。