Wireless MIMO Channel Dataset

This is a dataset of wireless MIMO channels. Using 3D city models from real world locations, we conducted ray-tracing simulation to model the channel from base stations (BS) to a dense grid of users. The raw path data has been postprocessed to emulate a MIMO setting with multiple antennas on both sides. We use the dataset to train CNN on the task of predicting an optimal beam index distribution over an area, given geospatial information of the environment. The experiments and results are described in our paper "Beam Index Map Prediction in Unseen Environments From Geospatial Data" and the code can be found on our Github page https://github.com/fabja19/beam_index_prediction .

If you find this dataset useful, please consider citing our work:

F. Jaensch, G. Caire and B. Demir, "Beam Index Map Prediction in Unseen Environments From Geospatial Data," in IEEE Wireless Communications Letters

https://ieeexplore.ieee.org/document/11363011

bibtex
@ARTICLE{11363011,
  author={Jaensch, Fabian and Caire, Giuseppe and Demir, Begüm},
  journal={IEEE Wireless Communications Letters}, 
  title={Beam Index Map Prediction in Unseen Environments From Geospatial Data}, 
  year={2026},
  volume={},
  number={},
  pages={1-1},
  keywords={Tensors;Indexes;Training;Urban areas;Azimuth;Base stations;Vectors;Buildings;Vegetation mapping;Spatial resolution;Convolutional Neural Networks;Machine Learning;MIMO},
  doi={10.1109/LWC.2026.3657504}}

effective_channel_tensors.zip
Contains the effective channel tensors after applying predefined beam codebooks. These are used to determine the optimal beam indices in each location.

gis.zip
Contains nDSMs of buildings and vegetation and aerial images for each city map.

model_checkpoint.zip
Weights and config of the best model in our experiments.

path_data{1...4}.zip
Contains the raw output data of the ray-tracing simulations, including power, time of arrival and directions of arrival and departure per path. It is not required for the experiments described in the paper.

script_create_effective_channel_tensors.py
Used to process raw path data to channel tensors.