Traffic-Aware Cloud-Edge Collaborative Offloading for Vehicular Tasks At Complex Intersections

Cloud computing, edge computing, and vehicle-to-vehicle (V2V) communication are essential for intelligent vehicular task processing, offering powerful computation and cooperative capabilities. However, dynamic traffic conditions, limited edge server coverage, and high cloud latency pose significant challenges for task offloading—especially in complex intersections prone to communication disruptions. To address these issues, we propose a cloud-edge collaborative offloading framework that takes into account real-world traffic environments and regulations. We model the problem as a distributed online service optimization problem and prove it NP-hard. To address this issue, we design predictive models for vehicle steering and passing time, and introduce the Traffic-Aware Distributed Online Offloading (TDOO) algorithm. Experiments using real traffic data show that TDOO significantly reduces task latency and outperforms existing methods.