Open Access

ARTICLE

MATD3-Based End-Edge Collaborative Resource Optimization for NOMA-Assisted Industrial Wireless Networks

Ru Hao^1,2,3, Chi Xu^2,3,*, Jing Liu¹

1 College of Information Engineering, Shenyang University of Chemical Technology, Shenyang, 110142, China
2 State Key Laboratory of Robotics, Shenyang Institute of Automation, Chinese Academy of Sciences, Shenyang, 110016, China
3 Key Laboratory of Networked Control Systems, Chinese Academy of Sciences, Shenyang, 110016, China

* Corresponding Author: Chi Xu. Email:

(This article belongs to the Special Issue: Advanced Communication and Networking Technologies for Internet of Things and Internet of Vehicles)

Computers, Materials & Continua 2025, 82(2), 3203-3222. https://doi.org/10.32604/cmc.2024.059689

Received 14 October 2024; Accepted 05 December 2024; Issue published 17 February 2025

Abstract

Non-orthogonal multiple access (NOMA) technology has recently been widely integrated into multi-access edge computing (MEC) to support task offloading in industrial wireless networks (IWNs) with limited radio resources. This paper minimizes the system overhead regarding task processing delay and energy consumption for the IWN with hybrid NOMA and orthogonal multiple access (OMA) schemes. Specifically, we formulate the system overhead minimization (SOM) problem by considering the limited computation and communication resources and NOMA efficiency. To solve the complex mixed-integer nonconvex problem, we combine the multi-agent twin delayed deep deterministic policy gradient (MATD3) and convex optimization, namely MATD3-CO, for iterative optimization. Specifically, we first decouple SOM into two sub-problems, i.e., joint sub-channel allocation and task offloading sub-problem, and computation resource allocation sub-problem. Then, we propose MATD3 to optimize the sub-channel allocation and task offloading ratio, and employ the convex optimization to allocate the computation resource with a closed-form expression derived by the Karush-Kuhn-Tucker (KKT) conditions. The solution is obtained by iteratively solving these two sub-problems. The experimental results indicate that the MATD3-CO scheme, when compared to the benchmark schemes, significantly decreases system overhead with respect to both delay and energy consumption.

---

Keywords

---