On the Analysis of AI-Optimized Aerial Cell-Free Massive MIMO

This study examines a cell-free massive MIMO architecture for unmanned aerial vehicles (UAVs). It evaluates aerial access points (APs) coverage, performance, and data rate of the aerial cell-free network. It proposes deploying an aerial cell-free massive MIMO architecture to mitigate the effects of path loss and interference in aerial cellular networks. The analysis includes a 2-dimensional multi-armed bandit (MAB) model for beam selection optimized with machine learning and using millimeter-wave technology to analyze an aerial cell-free network that connects a HAPS (CPU/data network) with ground vehicles through UAV-based APs. The multi-armed bandit model incorporates 3GPP blockage stochastics, water-filling power allocation, and optimization of multi-user capacity. The results include the aerial cell-free model's comprehensive geometric and radio link simulation analysis. The simulation outcomes demonstrate that the suggested cell-free network outperforms aerial cellular networks and NLOS terrestrial cell-free networks. Finally, we present a comparative study between our MAB model-based AI technique and a conventional non-AI technique, highlighting the significant performance improvements achieved by our approach.