Research On Heterogeneous Multi-Dimensional Resource Management For Device-Collaborative Networks

With the rapid development of mobile communications,Internet of Things,and other technologies,mobile devices are flooding into the network explosively.The massive access technology of the 5th generation mobile communication technology(5G)has ushered in a new era "Internet of Every thing".At the same time,new applications of mobile devices are also flourishing,such as face recognition,natural language processing,virtual reality,electronic medical care,smart city,etc.,which have been integrated into our lives.Largescale and multi-category mobile applications put forward higher requirements for network service performance,requiring the network to provide high-speed data rates and providing abundant computing and cache resources.To meet the needs of multiple types of mobile services,mobile edge computing(MEC)technology distributes services by deploying multi-dimensional resources around devices,improves the utilization efficiency of network resources,and meets the service quality requirements of multiple types of services.Stable,efficient,and secure multi-dimensional resource management directly impacts network latency,energy efficiency,and network overhead.It is an essential research issue in 5G and the 6th generation mobile communication technology(6G).Mobile devices have also become more intelligent,and communication,computing,and caching(3C)capabilities have been greatly improved.The Device-to-Device(D2D)technology effectively supports the device-collaborative network.This thesis studies the link enhancement technology of device-collaborative networks,device-collaborative computation offloading,communication and cache integration mechanism,and optimizes the system throughput,service delay,network energy efficiency and system stability.The main work and innovation points are as follows.First,to improve the communication performance in the device-collaborative network,the Non-Orthogonal Multiple Access(NOMA)technology and unlicensed access technology are applied on the D2D link to enhance the throughput of the device-collaborative network further.At the same time,considering that a device may establish connections with multiple devices simultaneously,NOMA technology is introduced on multiple D2D links of the same transmitter device.First,a swap-match-based licensed channel allocation scheme is proposed.Both centralized and distributed power control schemes are proposed to maximize the throughput of all D2D links while guaranteeing the date rate of each cellular link.In addition,considering that the transmit power of the D2D link is low,it can bring less interference to the WiFi network.An unlicensed channel access technology based on the Stackelberg game model is proposed.While ensuring the throughput performance of the WiFi network,the throughput on the D2D link is further enhanced to improve the service performance of the network.The channel capacity of the D2D link is maximized by adjusting the transmit power on the licensed and unlicensed channels.Compared with the existing unlicensed channel access scheme on the D2D link or the D2D link resource optimization scheme based on NOMA technology,the proposed scheme can improve resource utilization efficiency and improve network throughput.Second,a distributed device-collaborative computation offloading scheme is proposed for the joint optimization problem of multi-user and multi-task communication and computing in the device-collaborative network.That is,computation tasks are distributed among multi-users.Considering the average latency of computation tasks as a performance evaluation metric,this work formulates the problem of minimizing the average latency of multi-tasks by optimizing computation task scheduling,spectrum and computing resources.By applying convex optimization theory,the original problem is transformed into a convex problem,and a joint task scheduling and multi-dimensional resource allocation scheme based on the Alternating Direction Method Of Multipliers(ADMM)Algorithm is proposed.Via simulation analysis,compared with the joint MECcloud offloading scheme,the proposed scheme can effectively reduce the computation offloading latency.On the other hand,for the multi-user computation offloading scenario assisted by device collaboration,physical layer security is introduced,considering the security of computation data transmission.A joint optimization scheme of computation offloading decision,power control and computing resource allocation is proposed to minimize the weighted sum of latency and energy consumption and ensure the secure capacity of the transmission link.The proposed scheme is a two-layer optimization framework based on Asynchronous Advantage Actor-Critic(A3C)algorithm,which can make fast decisions and obtain near-optimal solutions,reducing network energy consumption and service latency,effectively ensuring the security of computation offloading.Third,for the long-term stable computation offloading problem in the wireless power transfer(WPT)-assisted device-collaborative network,considering the energy coordination between devices,the device can help other devices forward computation tasks or perform them.The computing energy efficiency is introduced as an evaluation metric of network performance,defined as the amount of computational data completed per unit of energy consumption.In this study,the optimal computing energy efficiency is achieved by optimizing the computation offloading strategy,the transmission power,and the MEC server’s computing resources.The formulated problem is a long-term optimization problem and is limited by the length of the cached data queue.By applying Lyapunov theory,a long-term optimization problem is transformed into a deterministic problem at each moment.Then the convex optimization method is used to solve these deterministic problems and finally obtain a long-term nearoptimal solution to the problem.The simulation results show that the proposed scheme can effectively improve computing energy efficiency and reduce the length of stable computation task data queues.Fourth,for the joint optimization of communication,computing,and cache in the device-collaborative network,cloud,MEC and devices cooperate to achieve joint computation offloading,caching,wireless access control,and corresponding 3 C resource management.Two types of services are mainly considered,i.e.,the computation offloading service and the content delivery service.An average latency minimization problem is formulated by optimizing the computation task scheduling and corresponding spectrum and CPU frequency allocations.Based on the achieved results,a latency-related caching utility maximization problem is formulated by optimizing the cache size allocation.Then the problem is reformulated as Markov Decision Process(MDP),aiming at solving the latency minimization problem.A task scheduling decision and spectrum and computing resource allocation scheme is proposed based on the Multi-agent Deep Deterministic Policy gradient(MADDPG)algorithm.A cache size optimization scheme based on MADDPG is proposed to solve the caching utility maximization problem.The simulation results show that the proposed scheme has advantages in effectively reducing the average weighted latency of computation offloading and content delivery tasks and improving cache utility.This thesis investigates the transmission enhancement for the D2D link,computation offloading,and 3C integration in device-collaborative networks.The proposed scheme effectively improves qualities of user service experiences,energy efficiency,stability,and performance and supports 5GB and 6G networks.The device-collaborative network still needs to continue to explore in terms of supporting digital twins,ubiquitous intelligence,and network security.