Research On Intelligent Optimization Of Multi-Dimensional Resources Collaboration In Mobile Edge Networks

With the continuous iteration and update of mobile communication technology,applications(e.g.,extended reality,holography type communication and intelligent interaction)with the requirements of ultrahigh reliability,extremely low delay and high bandwidth communication are gradually emerging.There exists a huge challenge to the traditional cloud computing network architecture caused by the massive terminal data and diversified emerging services.Mobile edge computing(MEC)sinking computing,caching and control functions to the edge of the network can provide users with localized processing and real-time computing capabilities,and reduce the backhaul link bandwidth pressure.In the scenario of mobile edge network,the wireless environment is changeable in real time,the user needs are diverse,and the multidimensional resources such as communication,computing,caching and sensing are coupled with each other.Therefore,it is necessary to design a dynamic adaptive optimization algorithm for specific types of businesses to perform intelligent scheduling and optimization of multi-dimensional resources.In order to solve the above problem,this dissertation proposes a multi-dimensional resource cooperative intelligent optimization scheme for mobile edge network,focusing on how to integrate computing,caching and sensing with current wireless communication.Taking heterogeneous cellular network,Internet of Things and Internet of Vehicles(IoV)as typical wireless communication network scenarios,multi-dimensional resource intelligent optimization scheme combining sensing assistance,computing offloading,edge caching and wireless communication is designed respectively to achieve efficient resource allocation,improve system performance and user service experience.Therefore,this dissertation plans to carry out the following four researches:(1)Research on DRL-based computing offloading and resource allocation strategies for mobile edge networksAiming at the limited battery capacity and computing power of user terminals in mobile edge networks,a multi-user computing task offloading and resource allocation scheme is studied.Firstly,a network slicing agent framework based on blockchain is proposed.As an intermediary agent,the slicing agent receives resource requests and responses from network slicing tenants,dynamically allocates resources,and then schedules physical resources from infrastructure providers according to the intelligent contracts.Secondly,taking computing tasks as an example,a joint optimization problem of computing offloading and resource allocation is defined to minimize delay and price.Finally,considering the dynamic service requests of multiple users and the randomness of edge network environment,the resource allocation scheme based on deep reinforcement learning(DRL)is proposed to solve the optimization decision problem of continuous state space.The simulation results verify the convergence of the DRL algorithm,and the proposed algorithm has obvious advantages compared with the comparison algorithm.(2)Research on task offloading and resource allocation strategies of computation-intensive services in ultra-dense networksAiming at the problem that computation-intensive services in ultradense networks need to decompose and offload tasks to multiple edge nodes for collaborative processing,the task offloading and resource allocation schemes for edge computing are studied.Firstly,a hybrid computing offloading strategy is proposed to decompose the problem that requires huge computing power into multiple subtasks and assign them to different computing nodes for parallel processing.Secondly,the dynamic adaptive strategy based on multi-agent distributed DRL is designed,and the multi-agent deep deterministic policy gradient algorithm is used for intelligent scheduling of edge network resources,to achieve low overhead computing offloading strategy and real-time resource allocation decision,and jointly optimize the system delay and energy consumption.Finally,in order to improve the learning speed of each agent,a DRL-assisted federate learning framework is introduced to improve convergence and protect the privacy of participants.Simulation results verify the superiority of the proposed algorithm in resource optimization performance.(3)Research on dynamic resource allocation strategies for network slicing in mobile edge networksTo solve the problem of diversified service requests of mobile edge network users,a dynamic resource allocation scheme based on network slicing is studied.Firstly,an integrated architecture of communication,computing,and caching has been built.Secondly,considering the diversity of user request services and dynamic wireless channel conditions,a DRLbased dynamic optimization strategy is designed for different task types of computing offloading and content caching,and the deep deterministic policy gradient algorithm is used to conduct joint scheduling of edge network communication,computing and caching resources,in order to maximize the utility of mobile virtual network operators while ensuring the quality of service of users.Finally,DRL is combined with ensemble learning to improve convergence and reduce computing costs.The simulation results verify that the proposed resource allocation scheme is obviously superior to other comparison strategies,and the result output speed is faster and the cost is lower.(4)Research on slice-oriented resource allocation strategies in digital twin edge networksIn order to satisfy the diversified business requirements of the future network,the resource optimization scheme for network slicing under the scenario of digital twin edge network is studied.Firstly,an integrated network virtualization architecture integrating digital twin and network slicing is proposed to realize service-centered and user-centered network management.Secondly,in the scenario of the edge network of the IoV,the environment awareness offloading mechanism based on the integrated sensing and communication system is designed to solve the joint optimization problem of task scheduling and resource allocation.Finally,for inter-slice and intra-slice resource allocation,an improved DRL algorithm based on game theory is used to jointly optimize communication and computing resources,so as to meet the low latency requirements of delay-sensitive and computation-intensive services.Simulation results verify the effectiveness and superiority of the proposed mechanism.In addition,the improved DRL algorithm has higher advantages than other comparison algorithms.