A Deep-Reinforcement-Learning-Based Social-Aware Cooperative Offloading in Mobile Edge Computing Networks

Computation offloading primarily addresses the deficiencies of user devices in terms of resource storage, computational performance, and energy efficiency in Mobile Edge Computing (MEC) networks. However, the dynamic network topology, limited resources of devices, and diversity of user requirements in Internet of Things (IoT) pose significant challenges in achieving efficient computational offloading. In this paper, a deep reinforcement learning (DRL) based social-aware cooperative task offloading scheme (DSCTO) is proposed to minimize the system cost, in which the clustering algorithm based on the social relationship is first designed to shrink the decision space of DRL. Then, the DRL is used to make offloading decisions in terms of the offloading devices and offloading rate. Simulation results demonstrate that the performance of the proposed DSCTO outperforms other concerned schemes.