Network Slicing in O-RAN-Enabled Cell-Free Massive MIMO: A DRL-Based Power Control

The advent of 5G networks necessitates more flexible and intelligent architectures, prompting a shift from conventional models to Open Radio Access Networks (0-RAN) augmented with integrated network slicing (NS). The combination of O-RAN with a cell-free architecture improves both coverage and performance, while NS facilitates dynamic resource allocation to support diverse services, such as ultra-reliable low-latency communications (uRLLC) and enhanced mobile broadband (eMBB). This paper introduces a novel NS-enabled, cell-free O-RAN framework designed to optimize resource allocation and power control. In contrast to traditional methods, we adopt a Deep Reinforcement Learning (DRL) approach, leveraging the Soft Actor-Critic (SAC) algorithm to dynamically allocate power and resources across distributed access points (APs), while simultaneously ensuring the efficient management of network slices. The proposed framework aims to maximize the number of admitted slices while minimizing network costs, ensuring optimal data rates for eMBB and low latency for uRLLC services. Through this approach, we demonstrate enhanced flexibility, scalability, and performance in dynamic 5G wireless environments.