Research on collaborative offloading and resource allocation of Internet of vehicles in ISAC scenarios based on digital twin

The evolution of sixth-generation (6 G) networks toward ultra-low latency and full-domain coverage necessitates dynamic digital twin networks for environmental perception in Internet of Vehicles (IoV) systems. However, the massive volume of sensing and communication data in IoV systems challenges traditional cloud computing with latency and energy overload. While multi-access edge computing (MEC) mitigates latency by moving computation to the network edge, its limited coverage remains a bottleneck. Integrated Sensing and Communication (ISAC) further enhances vehicular awareness by enabling real-time environmental perception. We propose a Sensing-aware Joint service node Auction Offloading (SJAO) scheme that decomposes computation-intensive tasks and dynamically offloads subtasks to the edge server, assisting vehicles, or local units. A virtual control unit orchestrates offloading to minimize overall system cost. To handle the uncertainty of assisting vehicle resources, a multi-attribute reverse auction mechanism is introduced, considering computational capacity, channel state, and pricing. Additionally, a DT-enhanced Twin Delayed Deep Deterministic Policy Gradient Perception Offloading (TD3PO) algorithm is developed, incorporating environment change detection for DT adaptation and fusing DT outputs with policy actions to improve convergence and decision quality. Simulation results confirm that the proposed TD3PO algorithm significantly outperforms benchmark schemes in terms of system cost, efficiency, and adaptability.

© 2014 xxxxxxxx. Hosting by Elsevier B.V. All rights reserved.