Research On Offloading Strategy Of Intelligent IoT Mobile Edge Computing Based On Deep Reinforcement Learning

In the past decade,with the advancement in science and technology,communication technology has been growing rapidly.In particular,the fifth generation(5G)has become a key development direction,and the status of mobile devices has become extremely important.Mobile devices have evolved from a single communication device to an intelligent terminal with considerable computational power and the ability to handle a variety of tasks.However,due to the size limitation of mobile devices and stagnation in battery technology,not only are their computational power limited,so too is the number of tasks they can perform.In order to solve this problem,people begin to consider offloading computing tasks to the cloud server to solve the problem of insufficient local computational power,and reduce energy consumption of the equipment.Due to the long wireless transmission link from mobile devices to the cloud,the latency and energy consumption inevitably increase.In order to solve problems such as latency and energy consumption,industry and academia put forward the concept of mobile edge computing(MEC).Placing cloud computing nodes around users not only reduce the latency,but also decrease the energy consumption of mobile devices.With the wide application of cloud computing,more and more people begin to pay attention to the problem of cloud computing billing.How to get the most computational power in the cloud with the lowest price has become the focus of users.In this paper,we study the task offloading problem of MEC network in intelligent internet of things(Io T).Many users perform some computational tasks with the help of multiple computing access points(CAPs).By transferring some tasks to CAPs,the performance of the system can be improved by reducing latency and energy consumption.The latency and energy consumption are two important optimization indices in MEC network.This paper presents the offloading strategy intelligently by the deep reinforcement learning algorithm,and designs the offloading system.The algorithm uses Deep-Q Network(DQN)to learn the offloading decision automatically to optimize the system performance,train the neural network to predict the offloading behavior,and the training data is generated by the interaction between the agent and the environment.In addition,we use bandwidth allocation to optimize the wireless spectrum of the link between the user and CAP,and propose three bandwidth allocation schemes.Further,we use CAP selection to select the best CAP to assist users in the computational task.In view of the problem that UAV threatens the security of data transmission,this paper also proposes a secure green mobile edge computing network optimization framework based on deep reinforcement learning.In order to reduce the local computational pressure,some computational tasks can be transferred to the CAP,at the cost of price,transmission latency and energy consumption.By jointly reducing pricing,latency and energy consumption,a secure mobile edge computing network optimization framework based on deep reinforcement learning is proposed.Specifically,we first adopt four optimization criteria,including criterion I to minimize the linear combination of pricing,latency and energy consumption,criterion II to minimize the pricing of latency and energy consumption,criterion ? to minimize the latency of pricing and energy consumption,and criterion ? minimizes energy consumption in the case of pricing and latency restriction.For each criterion,we propose an optimization framework,which can dynamically adjust the task offloading rate and bandwidth allocation rate,and propose a new feature extraction network to improve the training effect.The simulation results verify the effectiveness of the optimization framework.