MobiPower: Scheduling mobile charging stations for UAV-mounted edge servers in Internet of Vehicles

Recently, Unmanned Aerial Vehicles (UAVs) have been proposed as the carriers of mobile edge servers. Such UAV-mounted edge servers can offer computation offloading service to users more efficiently due to their fast movement. However, the limited battery endurance of the UAVs places a real challenge. In this paper, we propose MobiPower: a two-layer UAV-based edge computing and charging scheme for Internet of Vehicles (IoVs). The UAV-mounted edge servers fly in the air to serve the IoV nodes, while the Mobile Charging Stations (MCSs) move on ground to charge these UAVs. As the mobile charging stations are low-speed trucks which are difficult to respond to the realtime traffic change, we predict the future traffic based on the historical data to facilitate obtaining the offline scheduling strategies in advance. Moreover, we use hierarchical clustering to analyze the traffic patterns and classify the regions with same pattern into the same cluster. Mobile charging stations will only be scheduled between different clusters to reduce the complexity. Finally, we design an auction-based mobile charging station scheduling algorithm to maximize the amount of offloaded tasks and reduce the moving distance of the mobile charging stations. We conduct experiments on a popular trace, i.e., the TaxiBJ trace, and the experimental results show that our heuristic strategies effectively schedule the mobile charging stations over the entire map and demonstrate superior performance compared with existing popular methods, e.g., up to \(36.46\%\) improvement in terms of task offloading ratio and \(39.06\%\) less scheduling distance of the mobile charging stations under typical scenarios.