Multiple Dual-Function UAVs Cooperative Computation Offloading in Hybrid Mobile Edge Computing Systems

The paper proposes multiple dual-function UAVs hybrid enabled mobile edge computing systems scheme, where UAVs serving ground users (GUs) are allowed to dynamically switch their roles to act as edge servers or aerial relays. Due to the limited battery or fuel of UAVs and GUs, a deep reinforcement learning (DRL) and successive convex approximation (SCA) combined algorithm is proposed to minimize the system energy consumption by jointly optimizing service assignment, task splitting, trajectory of UAVs and resource of both UAVs and GUs. Numerical results demonstrate that the proposed algorithm can be fast converged with limited iterations and outperform the single-function UAV schemes and binary task splitting strategies with a better system performance. Meanwhile, the average energy consumption (AEC) can be greatly reduced with the kinetic limitation of UAVs, while maintaining the average task latency (ATL) within an acceptable range.