LEO-Integrated Partial Task Offloading Optimization in Vehicular Networks

Zahir Abbas, Shihe Xu, Xinming Zhang

Abstract: The integration of low-earth orbit (LEO) satellites into vehicular edge computing represents a promising solution for future communication systems, providing global connectivity, extensive coverage, and enhanced computational and communication capabilities for Internet of Vehicles (IoV) applications. However, the mobility of vehicular users, coupled with limited computational and resource allocation capacities with roadside units (RSUs) and integrated UAV schemes, can lead to task processing failures, restricted coverage, and increased processing latency and energy consumption. To overcome these challenges, we propose a partial task offloading and user association scheme that uses LEO satellites to reduce overall consumption for all vehicular users. Our approach employs a partial task offloading scheme based on twin delay deep deterministic policy gradient (TD3) to optimize the average latency and energy consumption of the system. The proposed scheme outperforms four existing methods by significantly minimizing total processing latency and energy consumption.

External IDs:dblp:conf/wasa/AbbasXZ25