Deep Multiagent Reinforcement Learning Methods Addressing the Scalability Challenge

Motivated to solve complex demand-capacity imbalance problems in air traffic management at the pre-tactical stage of operations, with thousands of agents (flights) daily, even in a restricted airspace, in this paper, we review deep multiagent reinforcement learning methods under the prism of their ability to scale toward solving problems with large populations of heterogeneous agents, where agents have to unavoidably decide on their joint policy, without communication constraints