Dynamic Scheduling of IoV Edge Cloud Service Functions under NFV: A Multi-agent Reinforcement Learning Approach

Document Type


Publication Title

IEEE Transactions on Vehicular Technology


In recent years, the mobile Internet and communications industry has been growing rapidly, with the Internet of Vehicles (IoVs) serving as a representative example. The emergence of software function virtualization and network function virtualization has expanded the development of IoV beyond the limitations of the traditional network. However, the increasing number of incoming nodes and the explosion of service requests pose significant challenges for data collection, transmission, and fast processing. The scheduling of service functions is a crucial issue that constrains the quality of service. While existing studies on service function scheduling in IoV are scarce, those available extensively employ linear programming and heuristic algorithms with large solution spaces, limited to offline scheduling. In this paper, we model the service function scheduling problem as a flexible shop floor scheduling problem. We integrate consideration of vehicle speed and signal strength based on the network characteristics in IoV, leverage reinforcement learning to learn high-quality scheduling rules, and employ graph neural networks to capture the complex relationship between operations and vehicle nodes, thereby realizing online scheduling. Experimental results effectively demonstrate the proposed scheduling algorithm's effectiveness and its good capability in future complex network situations.

First Page


Last Page




Publication Date



Processor scheduling, Job shop scheduling, Reinforcement learning, Heuristic algorithms, Dynamic scheduling, Petroleum, Vehicle-to-infrastructure


IR conditions: non-described