From classification to optimization: Slicing and resource management with TRACTOR

Abstract

5G and beyond cellular networks promise remarkable advancements in bandwidth, latency, and connectivity. The Open Radio Access Network (O-RAN) framework offers significant flexibility to reach these competing goals through network slicing and closed-loop RAN control. Central to this evolution is the integration of machine learning (ML) for dynamic network control. This paper presents a comprehensive framework designed to optimize the operation of O-RAN systems. First, we build and share a robust, O-RAN dataset generated from real-world traffic captured across diverse locations and mobility scenarios, replicated within a full-stack srsRAN-based O-RAN system using the Colosseum RF emulator. This dataset serves as a foundation for training and deploying ML models. We then introduce a traffic classification approach that leverages a variety of ML models, showcasing the ability to rapidly train, test, and refine models to improve classification accuracy. With up to 99% offline average accuracy and 92% online accuracy for specific slices, the results highlight the flexibility and efficiency of the framework in adapting to different models and network conditions. Finally, we present a physical resource block (PRB) assignment optimization strategy within our framework, which employs reinforcement learning to iteratively refine resource allocation. Our learned policy achieves a higher mean performance score (0.631) than both a manually configured expert policy (0.609) and a random allocation baseline (0.588), demonstrating improved PRB utilization. More importantly, our approach exhibits significantly lower variability, with the Coefficient of Variation (CV) reduced by up to an order of magnitude in three out of four cases, ensuring more consistent and robust performance across varying user configurations. Our contributions, including open-source access to the tools and datasets, enable accelerated research in O-RAN systems and ML-driven network control.