Offloading of Tasks With Tight Delay Requirements via Combined Half and Full Duplex UAV Relays

In this paper, we focus on offloading of computationally demanding tasks with tight processing delay requirements (hundreds of ms) from user equipments (UEs) to multi-access edge computing servers while exploiting unmanned aerial vehicles (UAVs) as relays. Traditionally addressed problem jointly optimizing an offloading decision and an association of the UEs to the UAVs is, in our work, enriched by optimization of the relaying via UAVs through selection between half-duplex (HD) and full-duplex (FD) under a constraint on the maximum processing delay of individual tasks. To solve this complex problem, we also optimize the length of relaying time slots in HD and the transmission power allocation for the UEs and the UAVs in FD. Then, we transform the problem to one-to-many matching and propose a low-complexity greedy algorithm for the joint offloading decision and UEs association solved together with the duplex selection. We show that our proposal reduces the sum energy consumed by the UEs and UAVs for the tasks offloading by up to 40% compared to related works.