Decentralized Joint Access Point Selection and Power Control in Uplink Cell-Free Massive MIMO

Cell-free massive multiple-input multiple-output (CF-mMIMO) technology is emerging as a highly promising solution for supporting large-scale Internet of Things (IoT) networks, where a large number of spatially distributed and heterogeneous devices operate under stringent power constraints. In these complex and dynamic environments, achieving an optimal balance between spectral efficiency (SE) and energy efficiency (EE) requires intelligent resource allocation and a flexible, adaptive optimization framework capable of responding to fluctuating network conditions, device demands, and energy limitations. This paper presents a novel fully decentralized deep deterministic policy gradient (FD3PG) algorithm, enabling IoT devices (acting as agents) to autonomously optimize joint access point selection and power control (JAPSPC) in uplink CF-mMIMO systems. Our approach, termed Joint D3PG, aims to collaboratively maximize a performance metric known as resource efficiency (RE) to effectively capture the balance between SE and EE, guiding the system toward configurations that best align with its operational demands and resource constraints. Extensive simulation results demonstrate significant performance gains over existing benchmark schemes. Specifically, our approach achieves up to a 79% improvement compared to conventional CF-mMIMO systems and a 12% gain over state-of-the-art methods, highlighting its effectiveness for complex IoT networks.