Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Open RAN LSTM Traffic Prediction and Slice Management using Deep Reinforcement Learning

Fatemeh Lotfi, Fatemeh Afghah

TL;DR

The paper tackles dynamic QoS-aware network slicing in ORAN under non-stationary traffic by coupling an LSTM-based traffic-prediction rApp with a distributed DRL-based xApp. The rApp provides forecasts of per-DU traffic loads to inform the SAC-based xApp running across multiple DUs, forming a distributed DRL framework (DDRL) embedded in the ORAN RIC architecture. The approach models slicing as an MDP and uses a multi-agent SAC with a centralized critic to allocate resources while minimizing SLA violations, achieving clear gains over non-predictive baselines. Simulation results in a three-slice ORAN scenario show up to 7.7% improvement in final return and better per-user throughput with reduced QoS-violation variance, demonstrating the practical impact of integrating predictive rApps with DDRL for slice management.

Abstract

With emerging applications such as autonomous driving, smart cities, and smart factories, network slicing has become an essential component of 5G and beyond networks as a means of catering to a service-aware network. However, managing different network slices while maintaining quality of services (QoS) is a challenge in a dynamic environment. To address this issue, this paper leverages the heterogeneous experiences of distributed units (DUs) in ORAN systems and introduces a novel approach to ORAN slicing xApp using distributed deep reinforcement learning (DDRL). Additionally, to enhance the decision-making performance of the RL agent, a prediction rApp based on long short-term memory (LSTM) is incorporated to provide additional information from the dynamic environment to the xApp. Simulation results demonstrate significant improvements in network performance, particularly in reducing QoS violations. This emphasizes the importance of using the prediction rApp and distributed actors' information jointly as part of a dynamic xApp.

Open RAN LSTM Traffic Prediction and Slice Management using Deep Reinforcement Learning

TL;DR

The paper tackles dynamic QoS-aware network slicing in ORAN under non-stationary traffic by coupling an LSTM-based traffic-prediction rApp with a distributed DRL-based xApp. The rApp provides forecasts of per-DU traffic loads to inform the SAC-based xApp running across multiple DUs, forming a distributed DRL framework (DDRL) embedded in the ORAN RIC architecture. The approach models slicing as an MDP and uses a multi-agent SAC with a centralized critic to allocate resources while minimizing SLA violations, achieving clear gains over non-predictive baselines. Simulation results in a three-slice ORAN scenario show up to 7.7% improvement in final return and better per-user throughput with reduced QoS-violation variance, demonstrating the practical impact of integrating predictive rApps with DDRL for slice management.

Abstract

With emerging applications such as autonomous driving, smart cities, and smart factories, network slicing has become an essential component of 5G and beyond networks as a means of catering to a service-aware network. However, managing different network slices while maintaining quality of services (QoS) is a challenge in a dynamic environment. To address this issue, this paper leverages the heterogeneous experiences of distributed units (DUs) in ORAN systems and introduces a novel approach to ORAN slicing xApp using distributed deep reinforcement learning (DDRL). Additionally, to enhance the decision-making performance of the RL agent, a prediction rApp based on long short-term memory (LSTM) is incorporated to provide additional information from the dynamic environment to the xApp. Simulation results demonstrate significant improvements in network performance, particularly in reducing QoS violations. This emphasizes the importance of using the prediction rApp and distributed actors' information jointly as part of a dynamic xApp.
Paper Structure (13 sections, 6 equations, 4 figures, 1 algorithm)

This paper contains 13 sections, 6 equations, 4 figures, 1 algorithm.

Table of Contents

  1. Introduction
  2. System model
  3. Wireless communication model
  4. Optimization problem
  5. Proposed Solution
  6. LSTM for traffic prediction rApp
  7. DRL for autonomous slicing xApp
  8. State
  9. Action
  10. Reward
  11. Proposed Algorithm
  12. Simulation Results
  13. Conclusion

Figures (4)

  • Figure 1: Implementing an LSTM traffic load prediction over ORAN network structure.
  • Figure 2: Convergence analysis and performance comparison of distributed deep RL in different scenarios.
  • Figure 3: achieved throughput per users through distributed deep RL.
  • Figure 4: QoS violation std of each slice for distribution deep RL in different scenarios