Placement of UAV-Mounted Edge Servers for Internet of Vehicles

In future Internet of Vehicles (IoV), edge servers will be deployed to the proximity of vehicles on road to offer realtime computation offloading services. However, servers deployed at fixed sites are static, while the traffic load on road is highly dynamic, incurring the mismatch between the demand and supply from time to time. In this paper, we study using multiple Unmanned Aerial Vehicles (UAVs) to act as edge servers for the IoV in a city-wide area. We first build a model on how to dispatch multiple UAVs to multiple hovering locations where they can cover a maximum number of user requests on discretized time slots. As the problem is NP-hard, we design a heuristic algorithm to solve it in polynomial time, namely the UAV pie-eating dispatching algorithm. Next, we propose a matching-based strategy to reduce the total travel distance of the UAVs when they move between the hovering locations, with the purpose of saving energy. We run extensive trace-driven experiments on the trajectory data of the taxicabs in Shanghai. The results verify that our UAV pie-eating algorithm effectively dispatches the UAVs to different locations at different time to adapt to the spatio-temporal demand variation, and our matching-based path planning strategy also reduces the total travel distance significantly compared with other popular strategies.