Work

A 5G-TSN joint resource scheduling algorithm based on optimized deep reinforcement learning model for industrial networks

Year: 2025

Type: article

Abstract: As the rapid advancement of the Industrial Internet of Things (IIoT), numerous devices are now being connected to industrial networks through both wired and wireless methods. Massive amounts of data are collected and transmitted, necessitating network communications that ensure bounded latency, low jitter, and maximum throughput. Addressing the scheduling requirements for diverse intelligent services has made the collaboration between fifth-generation (5G) mobile cellular networks and Time-Sensitive Networking (TSN) a prominent research topic. Recently, Deep Reinforcement Learning (DRL) algorithms are commonly employed to tackle scheduling issues within the 5G-TSN architecture. However, there has been insufficient focus on maximizing throughput in multi-user scenarios and on the impact of channel quality indication (CQI) on resource allocation. In this article, we proposed a 5G-TSN joint resource scheduling solution based on an optimized Deep Deterministic Policy Gradient (DDPG) algorithm. This solution aims to provide latency guarantees for time-triggered applications, throughput guarantees for video applications, and to maximize total throughput across 5G and TSN. Additionally, we introduce the modulation and coding scheme (MCS) to efficiently utilize the non-Markov CQI. A dynamic MCS-based priority scheme is incorporated into DDPG algorithm, approaching the optimal solution of the objective function with lower computional complexity and higher accuracy. Compared to conventional DDPG-based scheduling methods, the proposed solution shows superior performance in simulations. Results demonstrate that our approach achieves 100% latency reliability for time-triggered flows and improves total throughput by at least 9.16% under varying user ratios and different Gate Control List (GCL) configurations within the TSN domain.

Source: Ad Hoc Networks

DOI: 10.1016/j.adhoc.2025.103783

Authors: Yang Zhang, Lei Sun, Zhangchao Ma, Jianquan Wang, Meixia Fu, Jinoo Joung

Institutions: University of Science and Technology Beijing, Sangmyung University

Language: English

Topic: Network Time Synchronization Technologies

Subfield: Computer Networks and Communications

Field: Computer Science

Domain: Physical Sciences

SDG: Industry, innovation and infrastructure

Open Access status: hybrid

Estimated APC paid: $3,040