A Data Forwarding Mechanism based on Deep Reinforcement Learning for Deterministic Networks

Deterministic networks have been proposed recently, where networks are expected to provide data transmission with bounded latency and very low packet loss rate to applications. In this paper, we analyze the requirements of various applications in deterministic networks, and propose a data forwarding mechanism for applications with bounded latency requirements. The proposed mechanism considers jointly the bounded latency requirement, network states and different resource usages, aiming at optimizing the resource usage of the whole network while satisfying the requirements of as many applications as possible. Deep reinforcement learning is used to connect the transmission latency with network states. Simulation results show that proposed mechanism can adjust effectively the data transmission for deterministic applications according to the resource usage of the networks.