Under demanding operational conditions such as traffic surges, coverage issues, and low latency requirements, terrestrial networks may become inadequate to provide the expected service levels to users and applications. Moreover, when natural disasters or physical calamities occur, the existing network infrastructure may collapse, leading to formidable challenges for emergency communications in the area served. In order to provide wireless connectivity as well as facilitate a capacity boost under transient high service load situations, a substitute or auxiliary fast-deployable network is needed. Unmanned Aerial Vehicle (UAV) networks are well suited for such needs thanks to their high mobility and flexibility. In this work, we consider an edge network consisting of UAVs equipped with wireless access points. These software-defined network nodes serve a latency-sensitive workload of mobile users in an edge-to-cloud continuum setting. We investigate prioritization-based task offloading to support prioritized services in this on-demand aerial network. To serve this end, we construct an offloading management optimization model to minimize the overall penalty due to priority-weighted delay against task deadlines. Since the defined assignment problem is NP-hard, we also propose three heuristic algorithms as well as a branch and bound style quasi-optimal task offloading algorithm and investigate how the system performs under different operating conditions by conducting simulation-based experiments. Moreover, we made an open-source contribution to Mininet-WiFi to have independent Wi-Fi mediums, which were compulsory for simultaneous packet transfers on different Wi-Fi mediums.
Keywords: SDN; UAV networks; edge computing; emergency computing; task offloading; vehicular communications.