Modeling and analysis of computation offloading in NOMA-based fog radio access networks

The hierarchical local-fog-cloud computing-enabled paradigm applied in fog radio access networks (F-RANs) has been considered as a promising architecture to cope with the surge of mobile intelligent applications. Meanwhile, non-orthogonal multiple access (NOMA) technology with superior spectrum efficiency has been widely utilized to provide transmission services for offloading tasks in F-RANs. A comprehensive understanding of the computation offloading performance of NOMA-based F-RANs is thus essential to provide guidelines for F-RANs planning and resources scheduling. In this paper, we first develop a mathematical framework to analyze the ability of NOMA to support computation offloading in F-RANs. Then, the expression for a lower bound of the average computation offloading probability is provided based on stochastic geometry and order statistics, and the relation between the computation offloading probability and the average task execution latency as well as energy consumption is further derived by queuing theory. Finally, extensive numerical simulations are executed to verify the accuracy of our theoretical results. The numerical results show that the average computation offloading probability is nearly linear with most of key network parameters, such as the coverage radius of cellular networks, the density of fog radio access points (F-APs), the number of users sharing the same spectrum resources, and the task request rate, respectively.