xDiff: Online Diffusion Model for Collaborative Inter-Cell Interference Management in 5G O-RAN

2025-08-19

AI Summary

To address the challenge of real-time resource allocation optimization for inter-cell interference management (ICIM) in 5G Open Radio Access Networks (O-RAN), this paper proposes an online policy generation framework integrating diffusion models with reinforcement learning. The method employs preference values as differentiable policy representations, enabling cooperative optimization across distributed units (DUs), and achieves near-real-time, user-reward-driven dynamic resource allocation via end-to-end training. Its key contribution lies in being the first to introduce diffusion models into O-RAN ICIM—leveraging their generative capabilities for continuous, gradient-based policy learning under distributed constraints. Evaluated in a three-cell real-world testbed, the approach significantly outperforms state-of-the-art ICIM solutions in terms of interference suppression and spectral efficiency. Results demonstrate the efficacy and practicality of diffusion models for online optimization in wireless networks. The implementation code is publicly available.

Technology Category

• Computer Vision: Diffusion Models for Vision
• Search and Optimization: Distributed Search
• Planning, Routing, and Scheduling: Optimization of Spatio-temporal Systems

Application Category

• Systems and Infrastructure for Web, Mobile and WoT: Virtualization and resource management in Web systems and infrastructures
• User Modeling, Personalization and Recommendation: On-Device user modeling, personalization, and recommendation
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs

Abstract

Open Radio Access Network (O-RAN) is a key architectural paradigm for 5G and beyond cellular networks, enabling the adoption of intelligent and efficient resource management solutions. Meanwhile, diffusion models have demonstrated remarkable capabilities in image and video generation, making them attractive for network optimization tasks. In this paper, we propose xDiff, a diffusion-based reinforcement learning(RL) framework for inter-cell interference management (ICIM) in O-RAN. We first formulate ICIM as a resource allocation optimization problem aimed at maximizing a user-defined reward function and then develop an online learning solution by integrating a diffusion model into an RL framework for near-real-time policy generation. Particularly, we introduce a novel metric, preference values, as the policy representation to enable efficient policy-guided resource allocation within O-RAN distributed units (DUs). We implement xDiff on a 5G testbed consisting of three cells and a set of smartphones in two small-cell scenarios. Experimental results demonstrate that xDiff outperforms state-of-the-art ICIM approaches, highlighting the potential of diffusion models for online optimization of O-RAN. Source code is available on GitHub [1].

Problem

• Manages inter-cell interference in 5G O-RAN networks
• Optimizes resource allocation through diffusion-based reinforcement learning
• Enables near-real-time policy generation for efficient coordination

Innovation

• Diffusion-based reinforcement learning for interference management
• Online learning solution with near-real-time policy generation
• Preference values as policy representation for resource allocation