Downlink MIMO Channel Estimation from Bits: Recoverability and Algorithm

This is the official implementation repository for "Downlink MIMO Channel Estimation from bits: Recoverability and Algorithm", published in IEEE Transactions on Signal Processing. A shorter conference version of this paper is published in Asilomar 2025 that can be found here.

Channel Estimation Problem Problem:

Limited feedback scenario considered in this work.

Results

An alternating direction method of multipliers (ADMM) algorithm, called REDEEM, is proposed for two different feedback schemes.

Performance in antenna setup with 32 transmitter antennas, 24 receiver antennas, 8 channel paths and 3-bit quantization.

Usage

The script can be run in randomly generated geometric channel models and the channels generated using DeepMIMO simulator

Main Function

main(method_type, data_type, N_trials, Nr, Nt, K_true,R, quant_bits,K_model, test_name, data_file_path)

% Main script to run the experiments
% method_type: 'admm', 'newadmm', 'all'
% data_type: 'random', 'from_file'
% N_trials: number of trials to run
% Nr: number of receiver antennas
% Nt: number of transmitter antennas
% K_true: number of paths in the true channel
% R: number of measurements to be sent back
% quant_bits: number of bits to quantize each measurement
% K_model: number of paths in the model used for recovery
% test_name: name of the test to save data
% data_file_path: path to the data file if data_type is 'from_file' [only required if data_type is 'from_file']

Trial run with random channels

matlab -nodisplay -nosplash -nodesktop -r "main('all','random',10,16,32,6,300,2,6,'test');"

Run with DeepMIMO datasets.
DeepMIMO datasets can be downloaded from here.

matlab -nodisplay -nosplash -nodesktop -r "main('all','from_file',10,16,32,6,300,2,6,'test',./deepmimo-datasets/user_channel_data_Boston5G_3p5_BS1_32x16antennas-deepmimov2.mat);"

Citations

Please cite these on using this repository.

@ARTICLE{11106918,
  author={Shrestha, Rajesh and Shao, Mingjie and Hong, Mingyi and Ma, Wing-Kin and Fu, Xiao},
  journal={IEEE Transactions on Signal Processing}, 
  title={Downlink MIMO Channel Estimation from Bits: Recoverability and Algorithm}, 
  year={2025},
  volume={},
  number={},
  pages={1-15},
  keywords={Downlink;Channel estimation;Maximum likelihood estimation;Channel models;Dictionaries;Antenna arrays;Vectors;US Government;Science - general;Training;Channel estimation;compression;quantization;limited feedback;recoverability},
  doi={10.1109/TSP.2025.3593414}}



@INPROCEEDINGS{10052071,
  author={Shao, Mingjie and Fu, Xiao},
  booktitle={2022 56th Asilomar Conference on Signals, Systems, and Computers}, 
  title={Massive MIMO Channel Estimation via Compressed and Quantized Feedback}, 
  year={2022},
  volume={},
  number={},
  pages={1016-1020},
  keywords={Maximum likelihood estimation;Quantization (signal);Channel estimation;Massive MIMO;Downlink;Harmonic analysis;Frequency conversion;channel state information;matrix/tensor recovery;quantization;compression},
  doi={10.1109/IEEECONF56349.2022.10052071}}

@misc{alkhateeb2019deepmimo,
      title={DeepMIMO: A Generic Deep Learning Dataset for Millimeter Wave and Massive MIMO Applications}, 
      author={Ahmed Alkhateeb},
      year={2019},
      eprint={1902.06435},
      archivePrefix={arXiv},
      primaryClass={cs.IT},
      url={https://arxiv.org/abs/1902.06435}, 
}