MIMO Radar Systems: Deep Learning vs. Traditional Approaches

MIMO Radar Systems: Deep Learning vs. Traditional Approaches

Mostafa Hefnawi, Zakaria Benyahia, Mohamed Aboulfatah, Elhassane Abdelmounim, Taoufiq Gadi
Copyright: © 2023 |Pages: 25
DOI: 10.4018/978-1-6684-5955-3.ch010
OnDemand:
(Individual Chapters)
Available
$37.50
Current Special Offers
No Current Special Offers
TOTAL SAVINGS: $37.50

Abstract

Unlike traditional phased-array radars that need successive scans to cover the entire field of view, MIMO radar transmits orthogonal waveforms from each antenna element simultaneously, allowing the illumination of all targets at once. Also, better detection performance and a high spatial resolution can be obtained using all the components extracted by the matched filters. MIMO radar systems can detect the range, angle, and doppler of the targets, using traditional techniques such as the fast fourier transform (FFT), the multiple signal classifier (MUSIC), and the minimum variance distortionless response (MVDR). On the other hand, deep learning (DL) techniques have been proposed for MIMO radar systems as an alternative to traditional techniques that are computationally expensive and very sensitive to clutters and interferences. This chapter presents the performance of MIMO radar systems in a cluttered environment using both conventional and DL techniques.
Chapter Preview
Top

Background

In MIMO radar, the RF signals received by the elements of the antenna array are arranged in a three-dimensional array called the radar data cube (RDC), as shown in Figure 1 (Gentile, R. & Donovan, M., 2016). To estimate the range and Doppler of the targets, the FFT can be applied along the fast-time and slow-time dimensions of the RDC, respectively. The analysis across the elements of the array is used to estimate the direction-of-arrival (DoA) of the targets. This analysis can be performed with the traditional beamscan, MUSIC, or MVDR techniques. For all three algorithms, the peaks of their spatial spectrum coincide with the DOAs of the targets. The MVDR algorithm can achieve a better angular resolution than the conventional beamscan, and MUSIC provides even better spatial resolution than MVDR. However, the MVDR and MUSIC are more sensitive to sensor position errors and to clutters and interferences. In addition, MUSIC requires the number of sources to be known or accurately estimated.

Figure 1.

Radar Data Cube

978-1-6684-5955-3.ch010.f01
(Gentile, R. & Donovan, M., 2016)

Complete Chapter List

Search this Book:
Reset