An RL Approach for Radio Resource Management in the O-RAN Architecture

Abstract

The new generation mobile network requires flexibility and efficiency in radio link management (RRM), in order to support a wide range of services and applications with diverse target KPI values. In this perspective, the O-RAN Alliance introduces a flexible, intelligent and virtualized RAN architecture (O-RAN), which integrates artificial intelligence models for effective network and radio resource management (RRM). This work leverages an O-RAN platform to develop and assess the performance of an RRM solution based on Reinforcement Learning (RL) and deployed as xApp in the O-RAN ecosystem. The framework receives periodic reports from the O-RAN Distributed Unit (DU) about the network status and dynamically adapts the per-flow resource allocation as well as the modulation and coding scheme to meet the traffic flow KPI requirements.