UAV-Based Intelligent Sensing Over CF-mMIMO Communications: System Design and Experimental Results

This paper presents an unmanned aerial vehicle (UAV)-based intelligent sensing system designed to operate within a cell-free massive multiple-input multiple-output (CF-mMIMO) network, addressing key limitations in performing integrated sensing and communication (ISAC) over UAVs. By exploiting the high-speed maneuverability of UAVs and the distributed and user-centric nature of the CF-mMIMO architecture, the proposed system enhances both sensing accuracy and communication reliability in highly dynamic environments. To perform high-precision target recognition, we introduce the fine-tuning YOLOv5s model with hardware acceleration specifically for the low-altitude UAV scenarios. Moreover, we propose an enhanced target tracking algorithm, which is further integrated with Kalman filter-based trajectory prediction for performance improvement. A prototype system is designed and implemented to validate the proposed framework. Extensive experimental evaluations conducted in realistic environments demonstrate significant improvements in both communication reliability and sensing performance. The results show that the proposed system outperforms traditional UAV-based approaches, achieving notable increases in Success Rate (3.4%) and Precision (1.7%) compared to existing methods.