In this work, we demonstrate an over-the-air communications system which is solely based on deep neural networks and has, thus far, only been validated by computer simulations for block-based transmissions. Transmitter and receiver can be jointly trained end-to-end for an arbitrary differentiable end-to-end performance metric, e.g., block error rate (BLER). We demonstrate that it is possible to build and train such a system using off-the-shelf software-defined radios (SDRs) and open-source deep learning software libraries. A comparison of the BLER performance of the "learned" system against that of a practical baseline shows competitive performance. We identify several practical challenges of training such a system over-the-air, in particular the missing channel gradient, and propose a learning procedure that circumvents this issue.
%0 Conference Paper
%1 dorner2017learningbased
%A Dörner, Sebastian
%A Cammerer, Sebastian
%A Hoydis, Jakob
%A ten Brink, Stephan
%B 2017 51st Asilomar Conference on Signals, Systems, and Computers
%D 2017
%I IEEE
%K sent ubs_10005 ubs_20007 ubs_30073 ubs_40406 unibibliografie
%P 1791-1795
%R 10.1109/ACSSC.2017.8335670
%T On deep learning-based communication over the air
%X In this work, we demonstrate an over-the-air communications system which is solely based on deep neural networks and has, thus far, only been validated by computer simulations for block-based transmissions. Transmitter and receiver can be jointly trained end-to-end for an arbitrary differentiable end-to-end performance metric, e.g., block error rate (BLER). We demonstrate that it is possible to build and train such a system using off-the-shelf software-defined radios (SDRs) and open-source deep learning software libraries. A comparison of the BLER performance of the "learned" system against that of a practical baseline shows competitive performance. We identify several practical challenges of training such a system over-the-air, in particular the missing channel gradient, and propose a learning procedure that circumvents this issue.
%@ 978-1-5386-1823-3 and 978-1-5386-1824-0 and 978-1-5386-0666-7
@inproceedings{dorner2017learningbased,
abstract = {In this work, we demonstrate an over-the-air communications system which is solely based on deep neural networks and has, thus far, only been validated by computer simulations for block-based transmissions. Transmitter and receiver can be jointly trained end-to-end for an arbitrary differentiable end-to-end performance metric, e.g., block error rate (BLER). We demonstrate that it is possible to build and train such a system using off-the-shelf software-defined radios (SDRs) and open-source deep learning software libraries. A comparison of the BLER performance of the "learned" system against that of a practical baseline shows competitive performance. We identify several practical challenges of training such a system over-the-air, in particular the missing channel gradient, and propose a learning procedure that circumvents this issue.},
added-at = {2020-03-24T12:19:23.000+0100},
author = {Dörner, Sebastian and Cammerer, Sebastian and Hoydis, Jakob and ten Brink, Stephan},
biburl = {https://puma.ub.uni-stuttgart.de/bibtex/2f5baa965b10bd508c0c2e78d7da26066/unibiblio},
booktitle = {2017 51st Asilomar Conference on Signals, Systems, and Computers},
doi = {10.1109/ACSSC.2017.8335670},
eventdate = {2017-10-29/2017-11-01},
eventtitle = {2017 51st Asilomar Conference on Signals, Systems, and Computers},
interhash = {8f1ea7f27624a0a1f35fac7ac2823a79},
intrahash = {f5baa965b10bd508c0c2e78d7da26066},
isbn = {{978-1-5386-1823-3} and {978-1-5386-1824-0} and {978-1-5386-0666-7}},
keywords = {sent ubs_10005 ubs_20007 ubs_30073 ubs_40406 unibibliografie},
language = {eng},
pages = {1791-1795},
publisher = {IEEE},
timestamp = {2020-03-24T11:19:23.000+0100},
title = {On deep learning-based communication over the air},
venue = {Pacific Grove, California},
year = 2017
}