Scalable Reinforcement Learning for Multiagent Networked Systems

Highlighted by success stories like AlphaGo, reinforcement learning (RL) has emerged as a powerful tool for decision making in complex environments. However, the of RL thus far been limited to small-scale or single-agent systems. To apply large-scale networked systems such energy, transportation, and communication networks, critical hurdle is curse dimensionality, because these systems, state action space can be exponentially large number nodes network. This article attempts break this dimensionality designs scalable method, named actor critic (SAC), The key technical contribution exploit network structure derive an exponential decay property, which enables design SAC approach.