Resource Management using Machine Learning in Mobile Edge Computing: A Survey

Abstract

Mobile Edge Computing (MEC) aims to overcome the limited terminal battery and processing capabilities associated with running applications in the mobile terminal and the high latency introduced by offloading these applications to the cloud. It extends the computing resources of the cloud at the edge of the cellular network closer to the mobile user. Resource management in mobile edge computing is one of the main issues that are studied recently by many researchers. It consists of resource allocation and computation offloading. Allocation of resources involves managing and scheduling the resources to accomplish the requests of the users. It depends on the availability and the capacity of the resources. According to the deadline of each requested task, the service provider will assign each user the sufficient resources. Computation offloading is the transfer of the tasks to be executed at an external platform (edge or cloud server). It depends on the processing capability and the storage capacity of the device. It is difficult to provide an optimal solution for resource management in a dynamic system due to the random variations of tasks required by the users and the mobility of these users, thus machine learning techniques are proposed to solve this optimization problem. In this paper we provide the state-of-the-art for using machine learning to optimize resource management in mobile edge computing. We divide the research into four categories: 1) minimizing the cost, 2) minimizing the energy consumption, 3) minimizing the latency and 4) minimizing both latency and energy consumption. We then classify the system model, the constraints and the types of machine learning techniques that are used in each optimization problem.