Segunda/Monday - Sexta/Friday (8:00/8:00AM – 18:00/6:00PM)
+55 91 3201-7181

Raymobtime is a methodology for collecting realistic datasets for simulating wireless communications. It uses ray-tracing and 3D scenarios with mobility and time evolution, for obtaining consistency over time, frequency and space. We incorporate simulations of LIDAR (via Blensor), cameras (via Blender) and positions to enable investigations using machine learning and other techniques. We have been using Remcom’s Wireless Insite for ray-tracing and the open source Simulator of Urban Mobility (SUMO) for mobility simulation (of vehicles, pedestrians, drones, etc). We also use Cadmapper and Open Street Map to simplify importing realistic outdoor scenarios. For more details, please check our publications.


UFPA, UNIFESSPA and North Carolina State University (NCSU), invite you to participate in the ITU Artificial Intelligence/Machine Learning in 5G Challenge, a competition which is scheduled to run from now until the end of the year. Participation in the Challenge is free of charge and open to all interested parties from countries that are a member of ITU. If you are interested in one of the following topics below, please signal your interest by filling out the form on the website.
Detailed information about the Challenge can be found in the document ITU AI/ML 5G Challenge – Applying AI/ML
in 5G networks. A Primer , available on the Challenge website.



Dataset nameWireless Insite Version3D scenarioFrequencyNumber of receivers and typeTime between scenesTime between episodesNumber of episodesNumber of scenes per episodeNumber of valid channels
s0003.2 Rosslyn60GHz10 Mobile100 ms30 s1165041K
s0013.2 Rosslyn2.8; 5 GHz10 Fixed5 ms37 s2001020K
s0023.2 Rosslyn2.8; 60 GHz 10 Fixed1 s3 s1800118K
s0033.2 Rosslyn2.8; 5 GHz10 Fixed1 ms35 s2001020K
s0043.2Rosslyn60 GHz10 Mobile1 s30 s5000135K
s0053.2Rosslyn2.8; 5 GHz10 Fixed10 ms35 s12580100K
s0063.2Rosslyn28; 60 GHz10 Fixed1 ms35 s2001020K
s0103.3Rosslyn60 GHz10 Mobile0.5 s5 s1005030K
s0113.3Rosslyn60 GHz10 Mobile0.5 s6 s762013K
s0123.3Rosslyn60 GHz10 Fixed0.5 s6 s1052021K


Dataset nameWireless Insite Version3D scenarioFrequencyNumber of receivers and typeTime between scenesTime between episodesNumber of episodesNumber of scenes per episodeNumber of valid channels
s0073.3Beijing2.8; 60 GHz10 Mobile1 s5 s504015K
s0083.2Rosslyn60GHz10 Mobile0.1 s30 s2086111K
s0093.3Rosslyn60GHz10 Mobile0.1 s30 s2000110K


Dataset nameWireless Insite Version3D scenarioFrequencyNumber of TransmittersNumber of ReceiversTime between scenesTime between episodesNumber of episodesNumber of scenes per episodeNumber of valid channels
v0013.3Rosslyn60 GHz25100 ms30 s20508.5k
v0023.3Rosslyn60GHz150.1 s0.1 s2500112.5K

Links of interest:

  • Wiki with technical information about the datasets (file format, definitions of episode and scene, etc.)
  • General Code in Python and Matlab / Octave for reading and interpreting Raymobtime files (hdf5 files).
  • Code in Python for baseline systems associated to the 2020 ITU Artificial Intelligence/Machine Learning in 5G Challenge
Example of ray-tracing simulation in a 3D scenario with the received powers of each ray indicated in colors.


Please feel free to create an issue at our Github.


WSIL/UT Austin Team


When using Raymobtime datasets/codes or any (modified) part of them, please cite this paper:

[1] A. Klautau, P. Batista, N. González-Prelcic, Y. Wang and R. W. Heath Jr., “5G MIMO Data for Machine Learning: Application to Beam-Selection using Deep Learning” in 2018 Information Theory and Applications Workshop (ITA) – PDF available HERE.

Raymobtime related work and historical perspective:

[2] A. Klautau, N. González-Prelcic and R. W. Heath Jr., “LIDAR Data for Deep Learning-Based mmWave Beam-Selection” in IEEE Wireless Communications Letters, vol. 8, no. 3, pp. 909-912, June 2019, doi: 10.1109/LWC.2019.2899571 (paper showing that deep neural networks with LIDAR data as input can be efficiently used to reduce the overhead associated to beam-selection in communication networks).

[3] Oliveira, A., Dias, M. Trindade, I. Klautau, A. “Ray-Tracing 5G Channels from Scenarios with Mobility Control of Vehicles and Pedestrians.” XXXVII SBRT, 2019.

[4] Marcos Yuichi Takeda, Aldebaro Klautau, Amine Mezghani, Robert W. Heath Jr., “MIMO Channel Estimation with Non-Ideal ADCS: Deep Learning Versus GAMP“, IEEE 29th International Workshop on Machine Learning for Signal Processing (MLSP).

[5] Nascimento, A., Frazão, W., Oliveira, A., Gomes, D., & Klautau, A. “MultiModal Dataset for Machine Learning Applied to Telecommunications.” XXXVIII SBRT, 2020.

Python code for repeatedly invoking ray-tracing to obtain a consistent time-evolution of communication channels paired with other simulators has been developed and was made public in 2018 by Pedro Batista and Aldebaro Klautau. Pedro was a Ph.D. student at UFPA, Brazil, while Aldebaro was a visiting scholar at The University of Texas, Austin, hosted by Prof. Robert Heath. The MIMO software was based on scripts available at WSIL, Prof. Heath’s group. Since then, the code and datasets have evolved with the collaboration of several people, including Profs. Nuria González-Prelcic (NCSU) and Diego Gomes (UNIFESSPA).  Based on the idea of interfacing Blender and Remcom’s Wireless InSite, a dataset limited to communication channels and images was used in a competition organized at IEEE ICC 2020. In 2020 Raymobtime datasets have been used in open machine learning competitions organized by ITU regarding beam-selection ( and channel estimation (