Cooperative End-Edge-Cloud Computing and Resource Allocation for Digital Twin Enabled 6G Industrial IoT

End-edge-cloud (EEC) collaborative computing is regarded as one of the most promising technologies for the Industrial Internet of Things (IIoT). It offers effective solutions for managing computationally intensive and delay-sensitive tasks efficiently. Indeed, achieving intelligent manufacturing in the context of 6G networks requires the development of efficient resource scheduling schemes. However, improving the quality of service and resource management in the face of challenges like time-varying physical operating environments of IIoT, task heterogeneity, and the coupling of different resource types is undoubtedly a complex task. In this work, we propose a digital twin (DT) assisted EEC collaborative computing scheme, where DT is utilized to monitor the physical operating environment in real-time and determine the optimal strategy, and the potential deviation between the real values and DT estimates is also considered. We aim to minimize the system cost by optimizing device association, offloading mode, bandwidth allocation, and task split ratio. Our optimization is constrained by the maximum tolerable latency of the task while considering both latency and energy consumption. To solve the collaborative computation and resource allocation (CCRA) problem in the EEC, we propose an algorithm with DT based on Multi-Agent Deep Deterministic Policy Gradient (MADDPG), where each user end (UE) in DT operates as an independent agent to determine the optimum offloading decision autonomously. Simulation results demonstrate the effectiveness of the proposed scheme, which can significantly improve the task success rate compared to benchmark schemes, while reducing the latency and energy consumption of task offloading with the assistance of DT.

I. Introduction

The worldwide rollout of 5G technology has given rise to an unprecedented surge in intelligent devices and data traffic [1]. As a result, we are witnessing the emergence of groundbreaking applications and services like the Internet of Vehicles (IoV), Virtual Reality (VR), and Expanded Reality (ER) [2]. These advancements necessitate ultra-high data rates, ultra-low transmission latency, and seamless connectivity on a massive scale [3]. In response to the pressing demands stemming from these breakthroughs, both academia and industry have begun efforts to delve deeper into next-generation networks (6G) [4], [5]. Envisioned as a cornerstone of the future communications infrastructure, 6G is poised to play a crucial role in the forthcoming decade and beyond. In the impending era of 6G, a large number of Industrial Internet of Things (IIoT) end devices will be connected to the network, while the integration of artificial intelligence and 6G networks opens the door to intelligent manufacturing. Nevertheless, end devices in IIoT generate huge amounts of data during operation, but their computational and storage capacities are limited [6], and how to efficiently handle large-scale computationally intensive and delay-sensitive tasks has become an imperative problem.