Deep Reinforcement Learning for Multi-agent Simulation using a partial floor field cutout | IEEE Conference Publication | IEEE Xplore