Mitigating xApp conflicts for efficient network slicing in 6G O-RAN: a graph convolutional-based attention network approach

TL;DR

The paper tackles xApp conflicts and communication overhead in O-RAN-enabled network slicing for 6G by introducing a Zero-Touch Management (ZTM) framework that combines Distributed Multi-Agent Reinforcement Learning (MARL) with a Graph Convolutional Network (GCN)–based attention mechanism. This MARL-GCN approach enables xApps (slice-managing agents) to coordinate resource allocation with selective communication, reducing overhead while preserving or enhancing performance as the slice count grows. The authors formalize a system model, develop a GCN-MHA attention module to propagate only relevant information across a learned graph of agents, and validate the method on the COLOSSEUM platform, demonstrating improved efficiency, fairness, and scalability over traditional MARL. The work offers a practical, scalable path toward automated, end-to-end slice management in Open-RAN, with potential broader impact on real-world 6G deployments and RRM automation.

Abstract

O-RAN (Open-Radio Access Network) offers a flexible, open architecture for next-generation wireless networks. Network slicing within O-RAN allows network operators to create customized virtual networks, each tailored to meet the specific needs of a particular application or service. Efficiently managing these slices is crucial for future 6G networks. O-RAN introduces specialized software applications called xApps that manage different network functions. In network slicing, an xApp can be responsible for managing a separate network slice. To optimize resource allocation across numerous network slices, these xApps must coordinate. Traditional methods where all xApps communicate freely can lead to excessive overhead, hindering network performance. In this paper, we address the issue of xApp conflict mitigation by proposing an innovative Zero-Touch Management (ZTM) solution for radio resource management in O-RAN. Our approach leverages Multi-Agent Reinforcement Learning (MARL) to enable xApps to learn and optimize resource allocation without the need for constant manual intervention. We introduce a Graph Convolutional Network (GCN)-based attention mechanism to streamline communication among xApps, reducing overhead and improving overall system efficiency. Our results compare traditional MARL, where all xApps communicate, against our MARL GCN-based attention method. The findings demonstrate the superiority of our approach, especially as the number of xApps increases, ultimately providing a scalable and efficient solution for optimal network slicing management in O-RAN.

Figures (5)

• Figure 1: An Overview of O-RAN Architecture: Components, Control Loops and Interfaces.
• Figure 2: Graph Convolutional Network-Based Attention Modules: in this illustration, each node i.e., xApp collects observations, which are then encoded by an MLP. The output of the MLP is then processed by two successive convolutional layers incorporating multi-head attention (MHA). Finally, the resulting output of these convolutional layers serves as input to the Q network.
• Figure 3: The cumulative reward obtained during the training of MARL-GCN-based attention over 5000 episodes.
• Figure 4: The satisfaction factor of 10 activated slices using MARL-based GCN and the traditional cooperative MARL (the ratio of the number of PRBs allocated to the requested according to throughput).
• Figure 5: The satisfaction factor of 20 activated slices using MARL-based GCN and the traditional cooperative MARL (the ratio of the number of PRBs allocated to the requested according to throughput).