Multi-User MIMO dataset with inter-user channel correlation

Codebase for "Massive MIMO Channels with Inter-User Angle Correlation: Open-Access Dataset, Analysis and Measurement-Based Validation", which is published on IEEE Transactions on Vehicular Technology 2021.

We used the Argos platform to measure the mobile channels to a 64-antenna massive MIMO base station in outdoor environments during this measurement. We measured user channels in 4 Line-of-Sight and 5 Non-Line-of-Sight clusters in near-stable environments. In each cluster, we measured >25 locations' channels. In total, the measurement covers > 225 locations. And the measured channel at each location covers all 52 data sub-carriers and up to >300 frames. The measurement is done on 2.4-GHz ISM band.

These datasets were collected by Xu Du and Xing Zhang at the Rice University campus. For any questions about this dataset, please contact the paper authors (xdurice@gmail.com) or the RENEW team. For details of the experiment, please refer to Section III-A in the above paper.

In each cluster, we sample > 25 locations. And the collected channel at each cluster can be used as

• a user randomly moving with >3800/7900 (NLOS/LOS) frames over all sub-carriers
• or 25 user channels at each location. Each location will have ~150/~300 frame over all sub-carriers Please see Get started Section for quick-start.

Code Structure

  aoa_estimator.py # for Fig. 5, MUSIC based measurement Angle-of-Arrival estimation.
  numerical.py # for Fig. 7-10, Neumerical Simulation.
  measurement.py # for Fig. 3-4, Ovet-The-Air finding plots.
  demo_access_user_channels.py # example for how to get started to access your channels.
  data
  └── intermediate compute results for faster fig plot fine tuning.
  lib
  ├── argos_lib.py # for raw trace processing
  ├── argos_lts.py # dependencies for estimating channels based on Argos 2
  ├── dsp_lib.py # for signal processing/simulation
  └── ota_lib.py # for OTA measurement setups
  plots
  └── figures generated.
  raw_trace
  └── channel data set collected with Argos platform and used in the paper.
  README.md
  LICENSE

Requirement

• 30 G storage

• = python 3.7

• = matplotlib 1.3

• = numpy 1.8

Reproduce Paper figures

Run the following command to reproduce the measurement findings figures in plots/

python3 measurement.py

Run the following command to reproduce the numerical findings in plots/

python3 numerical.py

Run the following command to produce the AoA estimation based on our measurement:

python3 aoa_estimator.py --cluster_name "LOS 3" --location_idx 18  &&
python3 aoa_estimator.py --cluster_name "LOS 4" --location_idx 17 &&
python3 aoa_estimator.py --cluster_name "NLOS 2" --location_idx 16 &&
python3 aoa_estimator.py --cluster_name "NLOS 2" --location_idx 23 &&
python3 aoa_estimator.py --cluster_name "NLOS 3" --location_idx 18

which produce figures in plots/fig5s/

Get started in accessing the measurement for your purpose:

demo_access_user_channels.py contains two examples for using the channels.

Example 1: get the channel of each location in each cluster

get_all_cluster_measured_channel will returns Dict[str, Dict[int, PerLocationChannel]]. For example, get_all_cluster_measured_channel()["NLOS 1"][11] will return the measured wideband channel of user at Location 11 in NLOS 1 cluster. And the measured wideband channel at each location PerLocationChannel is of shape (NUM_BASE_STATION_ANTENNAS, FRMAE_NUM, SUB_CARRIRE_NUM). For example, Location 11 in NLOS 1 cluster has (64, 226, 52), which are 226 channels from users to 64 base-station antennas across all 52 data sub-carriers.

Example 2: get the channel of a user randomly moving in each cluster

Since we measured channels in each cluster continuously, you can also use the measurement data to get a user moving randomly inside each cluster. get_single_cluster_channel_mobile returns a dictionary that maps the cluster name to the channels of a user randonly moving inside the cluster. By running example below, you will see

• User at Cluster LOS 1 has channel (64, 7939, 52)
• User at Cluster LOS 2 has channel (64, 7958, 52) ...
• User at Cluster NLOS 4 has channel (64, 3856, 52)
• User at Cluster NLOS 5 has channel (64, 3862, 52)

And User at Cluster LOS 1 has channel (64, 7939, 52) means that user that measured channel of user moving in LOS cluster will have 64 base-station antenna, 7939 frame across 52 sub-carriers.

To run the demo, try the following command

python3 demo_access_user_channels.py

Reference

If you use the code or dataset from the project, please cite both of the following paper:

  @ARTICLE{9629284,
  author={Du, Xu and Sabharwal, Ashutosh},
  journal={IEEE Transactions on Vehicular Technology},
  title={Massive MIMO Channels with Inter-User Angle Correlation: Open-Access Dataset, Analysis and Measurement-Based Validation},
  year={2021},
  volume={},
  number={},
  pages={1-1},
  doi={10.1109/TVT.2021.3131606}}


  @ARTICLE{8368089,
  author={Zhang, Xing and Zhong, Lin and Sabharwal, Ashutosh},
  journal={IEEE Transactions on Wireless Communications},
  title={Directional Training for FDD Massive MIMO},
  year={2018},
  volume={17},
  number={8},
  pages={5183-5197},
  doi={10.1109/TWC.2018.2838600}}