Collaborative Task Offloading with Digital Twin in Multi-Vehicle and Multi-Edge Environments

Authors:

Abstract
In recent years, the effective utilization of edge servers to assist vehicles in handling compute-intensive and latency-sensitive tasks has emerged as a pivotal concern in Vehicular Edge Computing (VEC). In this paper, we adopt a cooperative approach that leverages the collective capabilities of multiple edge servers. This strategy is designed to effectively manage tasks and alleviate the computational burden imposed on these servers. Specifically, Graph Neural Network (GNN) is applied to extract and classify features such as the geographical locations and communication statuses of multiple edge servers, enabling the selection of the most suitable servers for collaborative task execution. We have utilized solar energy for local computing, effectively achieving environmental protection and reducing the local energy burden on vehicles. Moreover, a novel edge attraction formula is defined to refine the rationality of clustering. In addition, Deep Reinforcement Learning (DRL) is employed to make real-time offloading decisions. To ensure experimental accuracy while mitigating costs, we establish a corresponding digital twin environment to acquire experimental data. By conducting a comparative analysis against three other baseline methods, we effectively reduce task completion time and thus meet the stringent demands of time-sensitive tasks.
Funding Information
  • National Natural Science Foundation of China

This publication has 12 references indexed in Scilit: