Computation offloading for mobile edge computing: A deep learning approach

Abstract

Computation offloading has already shown itself to be successful for enabling resource-intensive applications on mobile devices. Moreover, in view of mobile edge computing (MEC) system, mobile devices can offload compute-intensive tasks to a nearby cloudlet, so as to save the energy and enhance the processing speed. However, due to the varying network conditions and limited computation resources of cloudlets, the offloading actions taken by a mobile user may not achieve the lowest cost. In this paper, we develop a dynamic offloading framework for mobile users, considering the local overhead in the mobile terminal side, as well as the limited communication and computation resources in the network side. We formulate the offloading decision problem as a multi-label classification problem and develop the Deep Supervised Learning (DSL) method to minimize the computation and offloading overhead. Simulation results show that our proposal can reduce system cost up to 49.24%, 23.87%, 15.69%, and 11.18% compared to the “no offloading” scheme, “random offloading” scheme, “total offloading” scheme and “multi-label linear classifier-based offloading” scheme, respectively.