Research On Dynamic Offloading Of IoT Based On Fog Computing

The proliferation of the Internet of Things(IoT)has met people’s growing demand for intelligent life and production.However,the explosion of IoT devices has led to problems such as a surge in data volume and limited computing resources.In addition,emerging IoT applications require significant computing resources for real-time processing.Therefore,addressing the massive data generated by IoT devices,providing adequate computing resources for emerging IoT applications,and achieving sustainable development of IoT devices and applications have become critical challenges in modern communication network research.To address these challenges,the IoT-Fog-Cloud network utilizes fog computing to assist resource-limited IoT devices in processing computation-intensive tasks,providing low latency,location-awareness,and high-mobility services for IoT devices by offloading tasks to fog nodes.However,the computing resources of fog nodes are limited,and designing an appropriate offloading strategy and achieving effective computing resource planning in the IoT-Fog-Cloud network are the key research issues.This paper focuses on the dynamic offloading joint resource allocation algorithm in the IoT-Fog-Cloud network,including the dynamic selection of offloading location and the allocation of computing resources.Firstly,this paper proposes a dynamic offloading joint computing resource allocation scheme based on the Genetic Algorithm(GA).An IoT-Fog-Cloud network with a double-layer fog structure is built to consider the mobility of some fog nodes.Based on this network,a new dynamic offloading model is proposed,and the offloading position of the task is determined by the central fog node.A GA-based offloading node selection and computing resource allocation algorithm is then proposed to dynamically select the offloading location of tasks and allocate computing resources on demand.The algorithm can minimize the total computing overhead,and generate the optimal offloading decision through crossover and mutation operations.The simulation results indicate that the algorithm achieves superior performance compared to existing solutions by employing mobile fog nodes to support the central fog node in offloading.It also significantly reduces the total computational cost of the dynamic offloading strategy.Furthermore,this study investigates a dynamic offloading and joint resource allocation scheme based on deep reinforcement learning.Considering the independence of separable tasks among multiple IoT devices,a novel fog offloading framework based on the IoT-fog-cloud network is proposed.In this framework,an edge router is introduced,which determines the task’s offloading position selection and resource allocation strategy.By introducing binary offloading decision variables,the problem is transformed into a binary offloading decision problem for resource allocation.A location selection and resource allocation algorithm based on Double Deep Q Network(DDQN)is proposed,which achieves optimal offloading decisions through the joint optimization of task offloading positions,fog computing resources,and channel resources.Simulation results demonstrate the algorithm’s good convergence and low offloading cost.Moreover,it exhibits strong adaptability to multiple IoT devices and different offloading scenarios.