Distributed Computation Offloading Based on Stochastic Game in Multi-server Mobile Edge Computing Networks

With the popularity of the Internet of things (IoT), 5G technology needs to meet various requirements of a large number of IoT applications. Mobile edge computing (MEC) is a promising approach in 5G scenario, which can solve the problems of resource shortage and high latency. Computation offloading is a key technology to reduce latency and energy consumption in MEC. In this paper, we consider a scenario with a dense distribution of edge nodes, namely multi-edge server distribution, and focus on the offloading problem in the overlapping coverage area of service scope. We build a two-step game model using the stochastic game theory. We pay attention to the relevance of state transition, that is, the cost of the next state is taken into account when making decisions in current state. In addition, we prove the existence of Nash Equilibrium (NE) by the concept of exact potential game, then propose the best response based on stochastic game (BRSG) algorithm to solve the problem. Numerical results illustrate that our algorithm can reach a NE through finite number of iterations, and an equilibrium strategy can be obtained. Besides, considering the state correlation makes the equilibrium cost significantly lower.

I. Introduction

Driven by 5th-Generation (5G) communication, mobile computing model has shifted from centralized cloud computing to edge cloud computing in recent years [1],[2]. Different from traditional cloud computing, edge computing migrates data computing tasks to the network “edge”, near the terminal user [3]. Thus, many edge nodes distributed on the network can relieve the pressure of centralized cloud computing centers [4]. There are currently a myriad of diverse IoT applications for many different environments. The diversity of these applications means that there is no one solution that fits all, as each application has different requirements in latency and data rate [5]–[7]. Offloading computing to the edge of the network can significantly reduce latency and improve quantity of service for a variety of computationally intensive applications [8],[9].