A Brief introduction to a massive clustering algorithm for fast radio resource allocation to distributed units in O-RAN networks
G. Amati, S. Angelini, F. Mangiatordi, E. Pallotti, P. SalvoDescription
In this page the authors present will present our research on the development and application of a massive clustering algorithm designed for efficient radio resource allocation in Open Radio Access Networks.
This study used graph modelling and Leiden clustering to analyze radio measurements collected by the Ugo Bordoni Foundation on Italian highways and urban networks.
The dataset is comparable to MDT data from UEs and consists of over 9.9 million sections, each lasting 10ms. This analysis helps identify patterns and optimize 5G network performance.
The Leiden clustering algorithm was applied to a network of 228,692 RU nodes, resulting in the formation of about 2000 clusters, which represent virtualized Distributed Units (DUs). The clustering was performed using the Constant Potts Model with a resolution parameter set to 0.001, ensuring finely-tuned community detection for optimizing network structure and performance.
Leiden’s algorithm proves highly effective in optimizing network processes, particularly in signaling procedures and the allocation of Distributed Unit (DU) functionalities by Central Units (CUs). By fine-tuning the placement of DUs and CUs based on Minimization Driving Tests (MDTs), this method can significantly reduce latency and service interruptions. This ensures more stable and efficient network operations, especially in areas with heavy mobile traffic.
