LLM-Based Intent Processing and Network Optimization Using Attention-Based Hierarchical Reinforcement Learning

Md Arafat Habib,Pedro Enrique Iturria Rivera,Yigit Ozcan,Medhat Elsayed,Majid Bavand,Raimundus Gaigalas,Melike Erol-Kantarci

2024-06-10

Abstract:Intent-based network automation is a promising tool to enable easier network management however certain challenges need to be effectively addressed. These are: 1) processing intents, i.e., identification of logic and necessary parameters to fulfill an intent, 2) validating an intent to align it with current network status, and 3) satisfying intents via network optimizing functions like xApps and rApps in O-RAN. This paper addresses these points via a three-fold strategy to introduce intent-based automation for O-RAN. First, intents are processed via a lightweight Large Language Model (LLM). Secondly, once an intent is processed, it is validated against future incoming traffic volume profiles (high or low). Finally, a series of network optimization applications (rApps and xApps) have been developed. With their machine learning-based functionalities, they can improve certain key performance indicators such as throughput, delay, and energy efficiency. In this final stage, using an attention-based hierarchical reinforcement learning algorithm, these applications are optimally initiated to satisfy the intent of an operator. Our simulations show that the proposed method can achieve at least 12% increase in throughput, 17.1% increase in energy efficiency, and 26.5% decrease in network delay compared to the baseline algorithms.

Networking and Internet Architecture

What problem does this paper attempt to address?

The paper aims to address several key challenges in intent-driven network automation, particularly the complexity of achieving automation in Open Radio Access Networks (O-RAN). Specifically, the paper mainly addresses the following three issues:

1. **Intent Processing**: Identifying the logic and necessary parameters to achieve a certain intent. This involves understanding the natural language descriptions of Mobile Network Operators (MNOs) and extracting specific performance metrics and optimization goals from them.

2. **Intent Verification**: Ensuring that the proposed intent is consistent with the current network state to avoid performance degradation due to changes in network conditions. For example, activating energy-saving applications during low traffic periods and activating throughput-enhancing applications during high traffic periods.

3. **Intent Realization**: Meeting the operator's intent through optimization functions (such as xApps and rApps). These applications use attention mechanism-based hierarchical reinforcement learning algorithms for optimal scheduling, thereby improving key performance indicators (such as throughput, latency, and energy efficiency).

The paper proposes a three-stage strategy to address the above issues: first, using a lightweight large-scale language model (LLM) for intent processing; second, utilizing a Transformer-based time series predictor to forecast future network traffic to verify the feasibility of the intent; and finally, adopting an attention mechanism-based hierarchical reinforcement learning algorithm to select and schedule appropriate network optimization applications to achieve the operator's goals. Experimental results show that this method improves throughput, energy efficiency, and latency by 12.02%, 17.1%, and 26.5% respectively compared to baseline algorithms.