Research On Resource Allocation In 5G Radio Access Network Slicing

5G technology has revolutionized communication networks by offering faster network speeds,lower latency,and greater bandwidth capacity,enabling a wide range of new applications and services.To cater to diverse quality of service requirements,5G network slicing has been introduced,which partitions the physical network into multiple logical networks that provide varying latency,throughput,security,and bandwidth characteristics.Mobile Edge Computing(MEC)has emerged as a key technology in 5G communication,leveraging the advantages of proximity to data terminals and the integration of communication,computing,and storage resources in mobile edge networks.However,these advancements bring about challenges in network resource management.This dissertation addressed several critical challenges and presented specific research plans to tackle them.Firstly,the wireless network environment in the edge network was dynamic,with randomly arriving services having distinct quality of service requirements.However,the available resources are unevenly distributed.Existing research predominantly considers complete system models,assuming perfect knowledge of the system state.In reality,the system state dynamically evolves over time,necessitating appropriate decision-making in mobile edge networks to optimize resource utilization.Secondly,MEC network slicing provides edge network resources for various services,demanding user service quality and accommodating different service types.Nevertheless,current research on network slicing resource allocation primarily focuses on static resource allocation,such as bandwidth and power.To enhance edge network slicing performance,it becomes essential to study dynamic resource scheduling and sharing among different services.Lastly,wireless network resources need to be allocated among 5G radio access network(RAN)slices to meet the quality of service requirements for each slice,thereby optimizing long-term resource utilization.This dissertation proposed specific research plans to address the aforementioned challenges,encompassing the following key points:I.Research on multi-access MEC dynamic resource sharing intelligent network slicing frameworkThis research investigated approaches to reduce latency and energy consumption for multiple users connected to the same edge computing node.It focused on computing offloading and resource allocation strategies to address challenges arising from wireless network dynamics,evolving user demands,and uneven resource allocation.To tackle these challenges,an optimization algorithm based on a reinforcement learning model was proposed.The model adapted to the dynamic nature of the network environment and learned from interactions with the environment.Specifically,the model utilized environmental feedback to get rewards,evaluated behavior effectiveness using a value function and selected the lowest-cost computational offloading and resource allocation based on the current state.To realize this dynamic and adaptive resource optimization strategy,a deep reinforcement learning method,Deep Deterministic Policy Network(DDPG),was employed.This approach enabled the model to learn from extensive datasets,adapting to multidimensional resource allocation problems and evolving network environments over time.By comparing the performance of the benchmark algorithm under different conditions,the proposed DDPG-based resource allocation and computation offloading strategies were validated.Simulation results demonstrated the superiority of the DDPGbased method in reducing latency and energy consumption for multiple users in single-edge computing node scenarios.II.Research on dynamic resource allocation of radio access network slicing based on deep deterministic policy gradientNetwork slicing technology has emerged as a promising solution to cater to diverse applications with varying performance requirements by dividing the physical network into logical networks.Radio access network(RAN)slicing,a specific form of network slicing,enabled the allocation of physical resources to different services while ensuring optimal quality of service.However,the dynamic nature of slicing requirements and environmental data posed challenges in determining the most effective strategy for 5G-RAN slicing.Traditional methods often fell short in achieving intelligent resource allocation.In this study,I proposed a novel approach called the Energy-Efficient Deep Deterministic Policy Gradient Resource Allocation(EE-DDPG-RA)method,which aimed to optimize resource allocation while prioritizing energy efficiency.By eliminating unnecessary actions and reducing the available operating space,our method achieved faster convergence while ensuring the quality of service requirements for Beyond 5G(B5G)systems.Extensive numerical simulations demonstrated that our proposed method not only enhanced throughput but also effectively managed resources.III.Research on Priority-based 5G network slicing radio resource allocationThis study addressed the resource allocation problem in 5G Radio Access Network(RAN)slices by proposing a priority-based approach.The objective was to distribute resources among multiple slices with different priorities while ensuring the Quality of Service(QoS)for each slice.The allocation of resources took into account the type and importance of the services provided.To achieve this,machine learning-based models were employed to learn from past resource allocation decisions and adjust future decisions accordingly.Additionally,a network slice controller monitored the performance of each slice in real-time and dynamically adjusted resource allocation based on observed metrics.Through simulations,it was demonstrated that the proposed approach effectively allocated resources,ensuring that critical services received the necessary resources for optimal functionality while improving overall resource utilization.This dissertation presented a study that focused on the intelligent optimization of multidimensional resources in mobile edge network slicing.The primary’ objective was to develop a dynamic adaptive optimization strategy to enhance the wireless communication environment of the edge network.The proposed strategy aimed to achieve rational allocation and efficient utilization of resources while catering to the diverse requirements of multi-service scenarios.This research held significant value in exploring the convergence of wireless communication and artificial intelligence,opening up new avenues for advancements in both fields.