Socially-Aware Energy-Efficient Task Partial Offloading in MEC Networks With D2D Collaboration

Abstract

The future wireless network will face demands of massive connectivity and intensive computation with the increase of mobile devices. Mobile edge computing (MEC) and Device-to-Device (D2D) have emerged as promising technologies to address the above challenges, and implementing social relationships in D2D-MEC networks can improve the reliability of D2D links. Exploiting these benefits, we investigate the energy-efficient task offloading problem in socially-aware D2D-assisted MEC networks, where the user devices can offload tasks to the nearby device or further forward to the MEC server based on social relationships. Specifically, we design a task partial offloading scheme of joint D2D connection selection, transmit power control and task allocation, to maximize the long-term network utility with considering dynamic system status and random task arrival. First, the social relationship among users is quantified into a social trust matrix. As the formulated socially-aware energy-efficient problem is a long-term stochastic optimization problem that is directly intractable, we thus employ the Lyapunov optimization to transform it into a series of short-term problems, each of which can be solved by the Karush-Kuhn-Tucker method and a pricing-based matching algorithm. Finally, we verify the performance optimality and the long-term network stability through numerical simulations as well as theoretical analysis.