Fast maximum likelihood DOA estimation in the two-target case with applications to automotive radar

Designing the Waveform Bandwidth and Time Duration of Automotive Radars for Better Collision Warning Performance

“…The procedure of a radar-based CWS can be divided into three steps: detecting object, estimating parameters and making decisions [34], as shown in Fig. 2.…”

“…A deduction from (34) is that σ 2 Z,opt < σ 2 Z,con as long as τ 0 = ∆ d /∆ v . That is, the optimized design reduces the error index and further improves the CWS performance.…”

Coprime beamforming: fast estimation of more sources than sensors

“…The increasing application of radar in emerging technologies such as automotive sensing and health devices has raised the importance of the problem of multiple‐source direction‐of‐arrival (DOA) estimation [1–4]. Conventional subspace‐based DOA estimation methods, such as multiple signal classification (MUSIC) and estimation of signal parameters via rotational invariance technique (ESPRIT), can detect sources using a uniform linear array (ULA) of N sensors [5–7].…”

Efficient implementation of the ML estimator for high‐resolution angle estimation in an unmanned ground vehicle

“…A few studies have reported reducing the computational burden of the ML estimator for automotive radars. In [12], the objective function is simplified using the centro‐Hermitian property of projection matrices and the necessary operators are pre‐calculated off‐line and stored for fast ML estimation. An efficient one‐dimensional (1D) search method is proposed in [13] based on Taylor series expansion of the objective function, which is valid for cases of small angular difference, i.e.…”

“…One or two targets are assumed to exist in the selected data, which is practical given that in most situations of automotive radars, multiple targets can be distinguished by their range and/or relative velocity. Therefore, especially if the data are extracted after range–Doppler (R–D) processing, the situations with <2 targets in selected processing cells can be understood as practical and relevant [12]. The proposed method enables global search procedures with appropriate memory requirements and computational costs.…”