Lightweight Offloading with Task Dependency Awareness in ISCC-Enabled Vehicular Networks

This paper proposes a latency-minimizationoriented lightweight task offloading scheme (DM-LTOS) for Integrated Sensing, Communication, and Computing (ISCC)-enabled vehicular networks, aiming to minimize the total application completion latency of the system while ensuring sensing accuracy. Specifically, a spatially coupled task deduplication and merging model is designed to eliminate redundant tasks with spatiotemporal overlap under the constraint of task dependencies, thereby reducing computational load and improving task processing efficiency. Furthermore, a wireless resource pre-allocation method based on the Cramér-Rao Bound (CRB) is proposed to guarantee sensing accuracy and enhance task transmission rates. On this basis, a hybrid offloading algorithm combining Graph Convolutional Network (GCN) and Deep Reinforcement Learning (DRL) is constructed for dependent applications, enabling accurate capture of task dependencies and network dynamics to determine the optimal offloading strategy. Simulation results show that the proposed scheme reduces the total application completion latency of the system by 60.71 % to 71.79 % across various network scenarios by minimizing redundant task processing and improving resource utilization, effectively enhancing the system's real-time responsiveness and operational efficiency.