"PRB-Level Channel Power Estimation Using CNN for CQI Reporting in 5G NR: Code and Simulation Framework"

"This dataset provides the complete source code and simulation framework for reproducing the experimental results of a PRB-level channel power estimation approach using a lightweight CNN (15,697 parameters) for CQI reporting in 5G NR systems. The package includes 53 source files spanning two execution environments: (1) MATLAB scripts for 5G NR link-level simulation using 3GPP-compliant TDL channel models, dataset generation with randomized SNR\/Doppler\/delay-spread conditions, CNN training via a teacher\u2013student supervised regression pipeline (LS input features, LMMSE-derived labels), and evaluation of per-PRB power, SINR, and CQI metrics; (2) Python and shell scripts for the Vitis AI INT8 quantization and FPGA deployment pipeline targeting the Xilinx ZCU102 (Zynq UltraScale+ MPSoC with 3 B4096 DPU cores). The codebase covers the full research workflow: data generation across 5 TDL profiles (A\u2013E), 5-channel engineered feature extraction per Rx antenna, MATLAB-based CNN training, hardware-aware model conversion (Keras to TensorFlow to INT8 xmodel), on-board DPU inference, and post-deployment evaluation including CQI aging analysis under mobility and channel estimation latency benchmarking. The repository preserves the original directory structure to ensure all relative paths function correctly, and includes a 12-page documentation PDF describing the execution sequence, software requirements, library function reference, and dependency graph."