Code for TUT-29: Machine Learning for MIMO Systems with Large Arrays - https://globecom2021.ieee-globecom.org/program/technical-tutorials-0

(previous version used at IEEE ICC'2021 Tutorial 14 - https://icc2021.ieee-icc.org/tutorials)

Nuria Gonzalez-Prelcic (NCSU), Aldebaro Klautau (UFPA) and Robert W. Heath Jr. (NCSU)

December 7, 2021.

In case you have questions or find problems in the code, feel free to indicate it by creating an issue via github or even submitting a pull request.

You can find here the RL code for a pedagogical beam selection problem in a grid world and information to run it.

The notebooks lstm_time_variant_channel.ipynb and qam_classifiers.ipynb generate themselves the used data. The main dependencies are listed in requirements.txt. You can use, for instance:

pip install -r requirements.txt

They can also be executed in Google's Colab.

The notebook beam_selection.ipynb requires downloading the input and output data for the neural network. The files can be obtained at https://nextcloud.lasseufpa.org/s/mrzEiQXE83YE3kg where you can find folders with the three distinct sets of input parameters: coord, image, lidar. You can use only one, all 3 or any combination. You also need the correct output labels, which is the file beams_output.npz in the folder beam_output. After downloading the datasets, save them in the folder specified in the notebook or change the code to point to your folder. The default folder name is data.

In case you intend to use all three sets of input parameters, you will end up with the files in the following folders:

• Contents of: your_folder\ml4comm\data

  •                                      beam_output
    

  •                                      coord_input
    

  •                                      image_input
    

  •                                      lidar_input
    

• Contents of: your_folder\ml4comm\data\beam_output

•                                        45,850,920 beams_output.npz  
  

•                                        45,850,920 beams_output_no_ori.npz (we will not use this one)
  

• Contents of: your_folder\ml4comm\data\coord_input

•                                        179,380 coord_input.npz
  

• Contents of: your_folder\ml4comm\data\image_input

•                                        43,522,538 img_input_20.npz
  

• Contents of: your_folder\ml4comm\data\lidar_input

•                                        447,760,264 lidar_input.npz