An analysis of multi-agent reinforcement learning for decentralized inventory control systems | Synapse