preDQN-Based TAS Traffic Scheduling in Intelligence Endogenous Networks

Abstract

The time-sensitive networking (TSN) working group standardizes time-aware shapes (TAS) to reduce network latency, but the traditional TAS standard lacks adaptability and cannot perform well in dynamically changing environments. The continuously developing artificial intelligence techniques can be combined with TSN to better adapt the dynamically changing environments. Therefore, we utilize deep reinforcement learning (DRL) algorithms to dynamically configure the network to improve its adaptability. Meanwhile, we propose the queuing transmission method in TAS by incorporating transmission windows for different types of traffic, to make resource allocation simpler and more efficient. However, DRL algorithms usually take a lot of time to train, which is contrary to the delay sensitivity of TAS. Hence, we propose an improved DRL algorithm, called preDQN, where the network prediction is used to help the agent explore the environment more efficiently. Experimental validation is carried out in a simulation environment, and the experimental results show that the scheme can significantly improve the resource utilization, reduce the end-to-end delay and packet loss rate.