A Novel Offloading Mechanism Leveraging Fuzzy Logic and Deep Reinforcement Learning to Improve IoT Application Performance in a Three-Layer Architecture Within the Fog-Cloud Environment

Abstract

This paper presents a novel offloading technique designed to enhance the efficiency of Internet of Things (IoT) applications within a sophisticated three-layer architecture situated in a fog computing environment. The IoT layer contains various intelligent IoT devices that generate a large number of tasks, each characterized by distinct specifications such as size, computational demand, communication requirements, and latency constraints. owing to the limited storage and computing capacity of resource-constrained IoT devices, it is essential to offload these tasks to different layers to ensure effective processing while satisfying the required Quality of Service (QoS) goals. To address this challenge, a fuzzy logic-based task scheduler is employed to make informed offloading decisions, considering task attributes and determining the most suitable processing layers - whether locally at the IoT layer, on collaborative fog nodes, or in the cloud. Furthermore, the study leverages the Deep Q Network (DQN) method, a form of deep reinforcement learning, to identify the optimal fog node for offloading tasks and to maintain a balanced workload distribution across collaborative fog nodes. The experimental findings demonstrate that the proposed scheme outperforms state-of-the-art solutions in terms of latency, power consumption, network usage, throughput, and offloading rate in comparison with the Non-offload, First-Fit, GASDEO, and NAFITO-FLA methods.