Optimal Resource Allocation in Mobile Edge Computing using Reinforcement Learning Approach

Abstract

Deep learning is now employed for many purposes, such as reading text on a computer screen and understanding natural language. Deep learning models or real-time deep learning analysis must analyze smartphone data and IoT endpoints. However, deep learning needs a lot of computer power for both inference and training to operate quickly. Edge computing enables edge devices to meet deep learning's high computation and low latency requirements. To achieve this, a thin mesh of computation nodes is positioned close to the edge devices. Other benefits of edge computing are scalability, privacy, and maximizing available bandwidth. We can resolve these problems via edge computing. This research intends to provide an examination of the latest advancements in the intersection of edge computing and deep learning. It will encompass a wide range of subjects. For instance, the research involves the training of different models on various edge devices. To increase service efficiency, the study uses a cognitive agent model and a communication network to allocate resources. To achieve accurate resource allocation among end users within Mobile Edge Computing (MEC), a combination of MOACO algorithms is utilized, making use of cost mapping tables to reach an optimal allocation.