Learning to Hybrid Offload in Space-Air-Ground Integrated Mobile Edge Computing for IoT Networks

Recently, space-air-ground integrated network (SAGIN) has garnered considerable interest from both academia and industry due to its broad-coverage and high-reliability features collaborated by low earth orbit (LEO) satellites, unmanned aerial vehicles (UAVs), and ground devices. In the meantime, the integration of SAGIN with other emerging communication technologies is promising for research and applications. Mobile edge computing (MEC) enables the resource limited devices, i.e., the Internet of things (IoTs) devices, to offload their data packet to the data processing center for computing. In this paper, a space-air-ground integrated MEC network is studied, where the UAV and satellite are capable for providing computing services to IoT devices. All IoT devices could split their data packet for local computing and offloading. The IoT devices can communicate with the UAV by active mode and/or passive mode through backscatter communication. The utility efficiency maximization problem that jointly considers the data volume, time latency, and energy consumption is formulated. As the problem is nonconvex and can not be addressed by conventional methods, a deep reinforcement learning based method is proposed to acquire the data offloading policy. Numerous numerical results confirm the effectiveness and robustness of the proposed method compared to other benchmark methods.