Direction of Arrival Estimation with Virtual Antenna Array Using FMCW Radar Simulated Data

TL;DR

This paper compares FFT, MUSIC, and compressed sensing algorithms for direction of arrival estimation in FMCW automotive radar, highlighting the trade-offs between speed and resolution using simulated data.

Contribution

It evaluates and contrasts the performance of different angle estimation algorithms for FMCW radar, emphasizing the advantages of super resolution methods over FFT.

Findings

FFT is the fastest but has poor angular resolution.

MUSIC and compressed sensing offer super resolution capabilities.

Super resolution algorithms outperform FFT in resolving closely spaced targets.

Abstract

The FMCW radars are widely used for automotive radar systems. The basic idea for FMCW radars is to generate a linear frequency ramp as transmit signal. The difference frequency, (i.e., beat frequency) between the transmitted and received signal is determined after down conversion. The FFT operation on beat frequency signal can recognize targets at different range and velocity. Increasing demand on safety functionality leads to the Direction of Arrival (DOA) estimation to resolve two closely located targets. Consequently, the problem of angle estimation for 77GHz FMCW automotive radar simulated data has been investigated in this term project. In particular, we examined the performances of FFT, MUSIC and compressed sensing in angle estimation task, and it was found that although FFT is the fastest algorithm, it has very poor angular resolution when compared with others which are both super resolution algorithms. The code for this project report is available at https://github.com/ekurtgl/FMCW-MIMO-Radar-Simulation.