Distributed User Association for Computation Offloading in Green Fog Radio Access Networks

Featuring edge computing capabilities, fog radio access networks (F-RANs) can help user equipments (UEs) execute computation-intensive tasks, which effectively alleviates their computation burden and energy consumption. However, current UE association optimization for computation offloading in F-RANs with multiple UEs and multiple fog access points (FAPs) can incur high complexity. To overcome this issue, a low complexity optimization approach is proposed, which aims at minimizing system energy consumption under practical constraints like per-FAP computation capability constraints. The core idea of the proposal is to first decouple the primal problem into two subproblems, namely a UE association problem and a computation offloading problem at each FAP, and then swap matching theory is applied to deal with UE association given any computation offloading scheme. For the second subproblem, a greedy algorithm is developed for each FAP to identify which received tasks are locally computed with the remaining tasks forwarded to a cloud. Simulation results show that the greedy computation offloading scheme can achieve near optimal performance and swap matching based association significantly outperforms various baselines.