Mobility-Aware Resource Allocation in Multi-Access Edge Computing Using Deep Reinforcement Learning

Abstract

Mobile Edge Computing (also known as Multi-access Edge Computing) brings computation and storage resources to the edge of a mobile network, allowing mobile devices (MDs) to offload high demanding tasks while meeting strict delay requirements. In this paper, we study the problem of efficient allocation of computational resources while considering the mobility information using deep reinforcement learning. An optimal policy considering the dynamics of the network is very difficult to achieve. Our objective is to develop an intelligent agent to optimize the decision-making process and the allocation of resources. To address this problem, we have proposed a solution based on the Deep Reinforcement Learning(DRL) method. DRL implements a deep Q-network that can consider a long-term goal and learns from the experience. The proposed method also considers the time-varying workloads of MEC servers and learns a policy to transfer tasks from one MEC server to another which further maximizes the system gain by avoiding unnecessary queue waiting time. The results of the simulation show that our proposed scheme reduces the cost of the system considerably as compared to the other baselines.