Distributed reinforcement learning in emergency response simulation