Obtaining Robust Control and Navigation Policies for Multi-Robot Navigation via Deep Reinforcement Learning