A height estimation method based on a beamspace joint alternating iterative algorithm in MIMO radar

Abstract

This paper discusses the problem of low-elevation target height estimation for multiple-input multiple-output (MIMO) radar in multipath environments. The beamspace compresses the data and is ideal for reducing the computational burden of elevation estimation. To obtain the height parameter of the target accurately, we propose a height estimation method based on a beamspace joint alternating iterative (BJAI) algorithm in MIMO radar. This method mainly converts the reduced-dimensional MIMO radar element space data into beamspace data and whitens them to improve the reliability. Then, a simplified model is used to obtain the initial value of the elevation, and we combine the reflection coefficient and the target elevation angle for alternate estimation. Finally, we calculate the target height using the obtained elevation information. Simulation results verify that the proposed algorithm has high estimation accuracy and strong robustness.

1. Introduction

Meter-wave radar, also known as ultrashort wave or very high frequency (VHF) radar, generally refers to radar operating in the meter-wave band (Gage and Green, 1978; Zheng et al., 2023). Because of its long wavelength and wide beam width, meter-wave radar has a strong ground reflection echo when measuring targets at low elevations. The direct signal of the reflected target and the coherent multipath signal reflected along the ground enter the main beam of the radar together (Zhu et al., 2017; Tan and Nie, 2018). The direct signal and multipath signal are usually difficult to separate in the time, Doppler, and spatial domains. These problems seriously affect the accuracy of the estimated elevation, and it is unrealistic to improve the elevation estimation accuracy simply by increasing the physical aperture of the array. As a new radar system, multiple-input multiple-output (MIMO) radar (Li and Stoica, 2007; Zhang et al., 2012; Shi et al., 2021) has virtual aperture expansion characteristics and obvious advantages over traditional radar in terms of anti-interference, anti-clutter, low interception, and angular resolution (Zhou et al., 2016). Therefore, it is more appropriate to use MIMO radar to handle target height estimation in a multipath environment, but how to find an accurate and robust method is an urgent problem that needs to be solved.

In recent years, there has been rapid development of array signal processing, leading to the emergence of many super-resolution methods (Carlin et al., 2013; Yang et al., 2020; Zheng et al., 2022). These methods can be broadly categorized into subspace algorithms and maximum likelihood (ML) algorithms. Subspace algorithms, such as MUSIC (Yin and Zhang, 2020) and ESPRIT (Feng et al., 2021), are commonly used. However, they require decoherence processing (Shi et al., 2016) to be completed first when estimating the target elevation angle, as they cannot directly process coherent signals. On the other hand, ML (Ziskind and Wax, 1988) algorithms can directly process coherent signals and are widely used in height estimation problems. The angle and reflection surface height joint estimation (ARJE) algorithm (Wang SH et al., 2016) is a method based on eigenvalue decomposition for complex terrains. It offers enhanced robustness by searching for more parameters. Beamspace (Xu and Zhao, 2019) reduces the data transmission volume, storage, and calculation, but sacrifices some degrees of freedom. To reduce the computational burden, the beamspace improved maximum likelihood (BIML) algorithm (Liu J et al., 2010) for MIMO radar was proposed for elevation estimation. The BIML algorithm uses prior knowledge of geometric information to reduce search dimensions and beamspace processing to reduce data dimensions. The three-dimensional beamspace maximum likelihood data fusion (3D-BMLF) algorithm (Chen S et al., 2022) converts element space data into beamspace data and uses eigenvectors to find closed-form solutions. The calculation amount is low, but the accuracy is subject to certain limitations. In addition, the refined maximum likelihood (RML) estimation method (Lo and Litva, 1991) takes advantage of prior information, such as radar height, target range, location of the multipath reflection point, multipath reflection coefficient, and multipath mirror geometry. Although the prior information contributes to the good performance of ML methods, these methods are usually applicable.