Mobility-aware task offloading in UAV-MEC-assisted IoV: A two-stage approach

With the development of 5G technology, Internet of Things technology has been integrated into vehicles, resulting in the concept of Internet of Vehicles (IoV). Nonetheless, IoV entails unique requirements and challenges. It is not wise to put all the tasks in the vehicle for computing. Fortunately, mobile edge computing (MEC) can improve the computing performance of IoV tasks by offloading some or all tasks to servers deployed at edge nodes to assist in computing. However, edge servers have limited resources, and some tasks generated by vehicles have strict processing time constraints, and the high mobility of vehicles introduces many uncertainties in task offloading. In view of the above issues, we study the problem of computation offloading for delay-sensitive applications in MEC-enabled high-mobility scenarios of IoV. Technically, we propose a two-stage mobility prediction and computation offloading method. More specifically, in the first stage, we propose a Transformer-based algorithm to predict the mobility of vehicles. Then, the predicted future positions of the vehicles are used as the input for the second stage, and computation offloading decision is made using Multi-Agent Proximal Policy Optimization algorithm. Extensive experiments are conducted to demonstrate the effectiveness and superiority of our proposed solutions in different situations. • A three-tier architecture of vehicles, UAVs, and edge servers enables flexible offloading. • In the first stage, Transformer-based position prediction reduces handover delay. • In the second stage, a MAPPO-based method leverages prediction results for delay-energy optimization.