To improve the quality of computation experience for mobile devices, mobile-edge computing (MEC) is a promising paradigm by providing computing capabilities in close proximity within a sliced radio access network (RAN), which supports both traditional communication and MEC services. Nevertheless, the design of computation offloading policies for a virtual MEC system remains challenging.Specifically, whether to execute a computation task at the mobile device or to offload it for MEC server execution should adapt to the time-varying network dynamics. In this paper, we consider MEC for a representative mobile user in an ultra-dense sliced RAN, where multiple base stations (BSs) are available to be selected for computation offloading. The problem of solving an optimal computation offloading policy is modelled as a Markov decision process, where our objective is to maximize the long-term utility performance whereby an offloading decision is made based on the task queue state, the energy queue state as well as the channel qualities between MU and BSs. To break the curse of high dimensionality in state space, we first propose a double deep Q-network (DQN) based strategic computation offloading algorithm to learn the optimal policy without knowing a priori knowledge of network dynamics. Then motivated by the additive structure of the utility function, a Q-function decomposition technique is combined with the double DQN, which leads to novel learning algorithm for the solving of stochastic X. Chen is with the ). M. Bennis is with the Centre for Wireless Communications, University of Oulu, Finland (email: bennis@ee.oulu.fi). 2 computation offloading. Numerical experiments show that our proposed learning algorithms achieve a significant improvement in computation offloading performance compared with the baseline policies. Index Terms Network slicing, radio access networks, network virtualization, mobile-edge computing, Markov decision process, deep reinforcement learning, Q-function decomposition. I. INTRODUCTION With the proliferation of smart mobile devices, a multitude of mobile applications are emerging and gaining popularity, such as location-based virtual/augmented reality and online gaming [1]. However, mobile devices are in general resource-constrained, for example, the battery capacity and the local CPU computation power are limited. When executed at the mobile devices, the performance and Quality-of-Experience (QoE) of computation-intensive applications are significantly affected by the devices' limited computation capabilities. The tension between computation-intensive applications and resource-constrained mobile devices creates a bottleneck for having a satisfactory Quality-of-Service (QoS) and QoE, and is hence driving a revolution in computing infrastructure [2]. In contrast to cloud computing, mobile-edge computing (MEC) is envisioned as a promising paradigm, which provides computing capabilities within the radio access networks (RANs) in close proximity to mobile users (MUs) [3]. By offloading computation ta...

No abstract

No abstract

5G is the next generation cellular network that aspires to achieve substantial improvement on quality of service, such as higher throughput and lower latency. Edge computing is an emerging technology that enables the evolution to 5G by bringing cloud capabilities near to the end users (or user equipment, UEs) in order to overcome the intrinsic problems of the traditional cloud, such as high latency and the lack of security. In this paper, we establish a taxonomy of edge computing in 5G, which gives an overview of existing state-of-the-art solutions of edge computing in 5G on the basis of objectives, computational platforms, attributes, 5G functions, performance measures, and roles. We also present other important aspects, including the key requirements for its successful deployment in 5G and the applications of edge computing in 5G. Then, we explore, highlight, and categorize recent advancements in edge computing for 5G. By doing so, we reveal the salient features of different edge computing paradigms for 5G. Finally, open research issues are outlined.

This work was supported in part by ROIS NII Open Collaborative Research 2020-20S0502 and in part by JSPS KAKENHI under Grants 18KK0279 and 19H04093, Japan.