A Novel Optimization Strategy For Computation Offloading in the UAV-assisted Edge Computing

Unmanned aerial vehicles (UAVs) capture real-time aerial data. However, onboard computational resources and battery life in a UAV device is limited. In both academic and industrial sectors, solutions based on the mobile edge computing paradigm have been extensively discussed. In this paper, a group of small UAVs are exposed to a range of computing tasks. Some of these tasks call for the execution of difficult computations and complicated algorithms which are computationally-heavy task, and offloading to a powerful computational device is required. In contrary, some tasks are data-heavy, and offloading these tasks lead to a considerable transmission delay. Thus, the performance of the system depends on whether a task is offloaded or computed locally. The UAVs are having three options for a given task, i.e., locally complete the computation task, transfer the task to a surrogate UAV (medium/Iarge UAV) through the wireless local access network, or transfer to a edge server through the cellular network. To solve this optimization problem, a heuristic approach is purposed where each UAV device takes a decentralized offloading decision based on the nature of the task (computationally-heavy or data-heavy) for minimizing the total overhead, i.e., computation delay, transmission delay and monetary cost. The performance of the proposed approach is compared with three models, i.e., local computation, offloading all the tasks to the surrogate UAV, and offloading all the tasks to the edge server. It is observed that the proposed model achieved on average 25–30% less global overhead.