Reinforcement Learning Based Offloading Framework for Computation Service in the Edge Cloud and Core Cloud

Abstract

Smartphones have increasingly become indispensable tools of our everyday lives, with extensive applications beyond communications, from utilitarian to entertaining. As such, demands on the technology remain stringent, leading to a limitation in performance and battery lifetime, requiring approaches such as computation offloading to improve the user experience for computation-intensive tasks such as gaming. This work presents the use of Reinforcement Learning in an offloading framework that provides smartphones with the ability to decide whether to perform computations on the smartphone or on the remote Cloud (Edge and Core) to minimize process. Several scenarios have been used to produce simulations that demonstrate that the proposed algorithm can operate efficiently in a dynamic Cloud computing and networking environment.