Towards Optimally Decentralized Multi-Robot Collision Avoidance via Deep Reinforcement Learning | IEEE Conference Publication | IEEE Xplore