Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Intelligent Hybrid Resource Allocation in MEC-assisted RAN Slicing Network

Chong Zheng, Yongming Huang, Cheng Zhang, Tony Q. S. Quek

TL;DR

The paper tackles efficient hybrid resource allocation for heterogeneous service demands in MEC-assisted RAN slicing by formulating a cooperative multi-node system as a weighted topology graph and solving it with a recurrent graph reinforcement learning approach. The proposed RGRL combines a graph convolutional network with a deterministic policy gradient and introduces a time-recurrent mechanism that incorporates the previous action into the current state, enabling robust adaptation to time-varying environments. Key contributions include the design of a graph-based RAC architecture, the formulation of a joint computing and transmission RA policy, and the demonstration of universal gains across single-node and multi-node use cases with improved SSR and stability and reduced policy complexity. The findings indicate that RGRL achieves higher long-term SLA satisfaction, better resilience to dynamic traffic and topology, and scalability for larger MEC-RAN slicing deployments, offering practical benefits for SLA-driven network management.

Abstract

In this paper, we aim to maximize the SSR for heterogeneous service demands in the cooperative MEC-assisted RAN slicing system by jointly considering the multi-node computing resources cooperation and allocation, the transmission resource blocks (RBs) allocation, and the time-varying dynamicity of the system. To this end, we abstract the system into a weighted undirected topology graph and, then propose a recurrent graph reinforcement learning (RGRL) algorithm to intelligently learn the optimal hybrid RA policy. Therein, the graph neural network (GCN) and the deep deterministic policy gradient (DDPG) is combined to effectively extract spatial features from the equivalent topology graph. Furthermore, a novel time recurrent reinforcement learning framework is designed in the proposed RGRL algorithm by incorporating the action output of the policy network at the previous moment into the state input of the policy network at the subsequent moment, so as to cope with the time-varying and contextual network environment. In addition, we explore two use case scenarios to discuss the universal superiority of the proposed RGRL algorithm. Simulation results demonstrate the superiority of the proposed algorithm in terms of the average SSR, the performance stability, and the network complexity.

Intelligent Hybrid Resource Allocation in MEC-assisted RAN Slicing Network

TL;DR

The paper tackles efficient hybrid resource allocation for heterogeneous service demands in MEC-assisted RAN slicing by formulating a cooperative multi-node system as a weighted topology graph and solving it with a recurrent graph reinforcement learning approach. The proposed RGRL combines a graph convolutional network with a deterministic policy gradient and introduces a time-recurrent mechanism that incorporates the previous action into the current state, enabling robust adaptation to time-varying environments. Key contributions include the design of a graph-based RAC architecture, the formulation of a joint computing and transmission RA policy, and the demonstration of universal gains across single-node and multi-node use cases with improved SSR and stability and reduced policy complexity. The findings indicate that RGRL achieves higher long-term SLA satisfaction, better resilience to dynamic traffic and topology, and scalability for larger MEC-RAN slicing deployments, offering practical benefits for SLA-driven network management.

Abstract

In this paper, we aim to maximize the SSR for heterogeneous service demands in the cooperative MEC-assisted RAN slicing system by jointly considering the multi-node computing resources cooperation and allocation, the transmission resource blocks (RBs) allocation, and the time-varying dynamicity of the system. To this end, we abstract the system into a weighted undirected topology graph and, then propose a recurrent graph reinforcement learning (RGRL) algorithm to intelligently learn the optimal hybrid RA policy. Therein, the graph neural network (GCN) and the deep deterministic policy gradient (DDPG) is combined to effectively extract spatial features from the equivalent topology graph. Furthermore, a novel time recurrent reinforcement learning framework is designed in the proposed RGRL algorithm by incorporating the action output of the policy network at the previous moment into the state input of the policy network at the subsequent moment, so as to cope with the time-varying and contextual network environment. In addition, we explore two use case scenarios to discuss the universal superiority of the proposed RGRL algorithm. Simulation results demonstrate the superiority of the proposed algorithm in terms of the average SSR, the performance stability, and the network complexity.
Paper Structure (29 sections, 28 equations, 17 figures, 1 table, 1 algorithm)

This paper contains 29 sections, 28 equations, 17 figures, 1 table, 1 algorithm.

Table of Contents

  1. INTRODUCTION
  2. SYSTEM MODEL
  3. Task Generation Model
  4. Hybrid Resource Allocation Model
  5. Queue Model
  6. Performance Metric Model
  7. PROBLEM FORMULATION AND METHODOLOGY
  8. Problem Formulation
  9. State
  10. Action
  11. State Transition
  12. Reward
  13. Methodology
  14. USE CASE DISCUSSION AND ANALYSIS
  15. Use Case 1: Single-node MEC-Assisted RAN Slicing Scenario
  16. ...and 14 more sections

Figures (17)

  • Figure 1: System model.
  • Figure 2: Framework of the proposed RGRL algorithm.
  • Figure 3: Use case 1: Single-node MEC-assisted RAN slicing system model.
  • Figure 4: Use case 2: Non-cooperative multi-node MEC-assisted RAN slicing system model.
  • Figure 5: Convergence behaviors of the proposed RGRL algorithm. ($N=8$, $U=650$, $C_{\rm B}=10 \; {\rm GHz}$, $Z_{\rm B}=10$)
  • ...and 12 more figures