DEEP LEARNING-BASED TASK OFFLOADING AND SCHEDULING ALGORITHMS IN MOBILE EDGE COMPUTING: A COMPREHENSIVE REVIEW

Abstract

The introduction of IoT, 5G, and latency-sensitive applications has created a huge demand for efficient computing and resource management. Mobile edge computing (MEC) has been suggested as a viable framework by which cloud computing functionalities can be extended to the edge of the network and improve the QoS by reducing latency. Task offloading and scheduling are two very important problems in MEC because of varying network states, availability of resources, and the demands posed by the application. Traditional optimization methods are efficient, but they exhibit very high computational complexity and lack real-time adaptivity. The development of advanced deep learning approaches has led to intelligent and adaptive methods for addressing task offloading and resource scheduling in MEC. Deep learning includes several methods, such as Deep Neural Networks (DNN), Convolutional Neural Network (CNN), Recurrent Neural Networks (RNN), LSTM networks, and DRL. They exhibit very good optimization capabilities for latency, energy efficiency, execution time, and resource utilization. This paper provides a thorough review of the use of deep learning for task offloading and scheduling in Mobile Edge Computing . The different methods are described and categorized based on architecture and optimization criteria among other factors. Also, performance evaluation and future direction of research are covered. The survey will help researchers understand more about the intelligent MEC paradigm and how future wireless technologies can benefit from it.