Digital Twin-Aided Intelligent Offloading With Edge Selection in Mobile Edge Computing

Abstract

In this paper, we study a mobile edge computing (MEC) architecture with the assistance of digital twin (DT) applied for industrial automation where multiple Internet-of-Things (IoT) devices intelligently offload computing tasks to multiple MEC servers to reduce end-to-end latency. To do so, first we propose and formulate a practical end-to-end latency minimisation problem in the DT-assisted MEC model subject to the constraints of quality-of-services and computation resource at the IoT devices and MEC servers in industrial IoT networks. Then, we solve the proposed latency minimisation problem by iteratively optimising the transmit power of IoT devices, user association, intelligent task offloading, and estimated CPU processing rate of the devices. Finally, simulation results are conducted to prove the effectiveness of the proposed method in terms of the latency performance compared with some conventional methods.

Edge Intelligence-Based Ultra-Reliable and Low-Latency Communications for Digital Twin-Enabled Metaverse

“…The maximum latency requirement for each task is T max m = 10 ms [13]. Following [2], [7], [8] and [5], other parameters Algorithm 1 : AO-based Algorithm for Solving (6).…”

“…The DT service for local processing of the m-th UE is denoted as DT ue m , which can be modelled as DT ue m = (f ue m , f ue m ), where f ue m is the estimated processing rate of the m-th UE and f ue m is the deviation between the estimated value and the real value of the processing rate. The deviation can be positive or negative to model the replicated processing rate in the DT [1], [2], [5]. Consequently, the local processing latency of the m-th for executing a task locally is given by…”

“…On the other hand, when the tasks is not cached, it is normally processed with the task offloading computing model. We note that the results returned from the AP to UEs are typical small (e.g., controlled messages) and the AP transmits the messages with more power than the UEs so that we only consider the uplink transmission latency in this paper [2], [3]. As a result, the latency model with edge caching is expressed as…”

“…, where the constant θ is the computation power parameter for energy consumption of the UEs [2], [8].…”

“…The actor-critic deep reinforcement learning (DRL) has been exploited to solve the optimisation problem. In [2], the mobile edge computing (MEC) architecture with the assistance of DT for IIoT has been investigated. This paper has taken into account various variables including transmit power, user association, offloading portions, and the estimated processing of IIoT devices to minimise the end-to-end latency with an iterative optimisation algorithm.…”