Design and Dimensioning of a UAV Set Covering in High-Traffic IoT–Fog Environments

The fog-computing paradigm provides low-latency processing and storage between Internet of Things (IoT) applications and cloud data centers. Normal IoT activity may produce infrequent yet substantial spikes in user traffic, which would require a large static fog infrastructure to service. Instead, we consider the viability of having a smaller static fog infrastructure, and supplementing additional traffic spikes with fog-enabled uncrewed aerial vehicles (fog-UAVs). This article formulates the optimal design and dimensioning of fog-UAVs and fog-UAV charging/deployment stations as a probabilistic location set covering problem (PLSCP). We model the fog-UAV PLSCP as a mixed-integer linear program (MILP), from which we derive several relaxed models, including one based on the Benders decomposition technique. Finally, we simulate our models over a set of city-wide IoT hotspots with various traffic percentile thresholds, and evaluate our results.