Coherent, super-resolved radar beamforming using self-supervised learning

Itai Orr, Moshik Cohen, Harel Damari, Meir Halachmi, Mark Raifel, Zeev Zalevsky

Research output: Contribution to journalArticlepeer-review

3 Scopus citations


High-resolution automotive radar sensors are required to meet the high bar of autonomous vehicle needs and regulations. However, current radar systems are limited in their angular resolution, causing a technological gap. An industry and academic trend to improve angular resolution by increasing the number of physical channels also increases system complexity, requires sensitive calibration processes, lowers robustness to hardware malfunctions, and drives higher costs. We offer an alternative approach, named Radar signal Reconstruction using Self Supervision (R2S2), which substantially improves the angular resolution of a given radar array without increasing the number of physical channels. R2S2 is a family of algorithms that use a deep neural network (DNN) with complex range-Doppler radar data as input and trained in a self-supervised method using a loss function that operates in multiple data representation spaces. Improvement of 4× in angular resolution was demonstrated using a real-world dataset collected in urban and highway environments during clear and rainy weather conditions.

Original languageEnglish
Article numbereabk0431
JournalScience Robotics
Issue number61
StatePublished - 15 Dec 2021

Bibliographical note

Publisher Copyright:
Copyright © 2021 The Authors, some rights reserved.


Dive into the research topics of 'Coherent, super-resolved radar beamforming using self-supervised learning'. Together they form a unique fingerprint.

Cite this