Extremely Large Antenna Spacing Method for Enhanced Wideband Near-Field Sensing

TL;DR

This work tackles the challenge of high‑resolution wideband sensing at mmWave by enabling near‑field sensing with a practically scalable ELAA. It proposes ELAS, a spacing scheme that yields a large effective Rx aperture with few elements, allowing the Rx to operate in the near field while the Tx remains in the far field with simple beam steering. The authors derive the NF range–angle response, define a super‑resolution region where NF effects dominate the range resolution, and show that wideband bandwidth and NF focusing jointly improve localization accuracy for an extended target modeled as scatterers. Numerical results using RMSE and GOSPA demonstrate significant gains of the NF ELAS architecture over a FF baseline, with the NF gains most pronounced inside the super‑resolution region and bandwidth mainly mitigating sidelobes outside it.

Abstract

This paper proposes a monostatic wideband system for integrated sensing and communication (ISAC) at millimeter-wave frequencies, based on multiple-input multiple-output (MIMO) orthogonal frequency-division multiplexing (OFDM). The system operates in a hybrid near-/far-field regime. The transmitter (Tx) operates in the far field (FF) and uses low-complexity beam steering. The receiver (Rx), on the other hand, operates in a pervasive near field (NF), enabled by a very large effective array aperture. To enable a fully digital implementation, we introduce an extremely large antenna spacing (ELAS) design. This design attains the required aperture with only a few widely spaced antenna elements while avoiding grating lobes in the composite Tx-Rx response. We analytically characterize the NF range-angle response of this architecture and study the interplay between NF effects and waveform bandwidth. This leads to the definition of a super-resolution region, where NF propagation at the Rx dominates the achievable range resolution and surpasses the classical, bandwidth-limited resolution. As a case study, we consider an extended target modeled as a collection of scatterers and assess localization performance via maximum-likelihood estimation. Numerical results evaluated in terms of root mean square error (RMSE) and generalized optimal sub-pattern assignment (GOSPA) show that operating in NF conditions with the ELAS-based design yields significant gains compared to a conventional FF baseline at both the Tx and Rx.

Figures (8)

• Figure 1: Monostatic MIMO ISAC setup with hybrid near-/far-field operation. The system is composed of a Tx ULA with half-wavelength spacing and a sparsely spaced Rx ELAA implementing the ELAS design to observe an ET located in the area of interest.
• Figure 2: Normalized Tx FF array factor (a), Rx NF array factor (b), and composite array factor (c) evaluated in the angular domain under the angular sampling method assumption, by considering the proposed monostatic radar configuration with the following system parameters: $f_\mathrm{c}=60\,$GHz, $d_\mathrm{t}=\lambda/2$, $d_\mathrm{r}=N_\mathrm{t}\lambda/2$ and $N_\mathrm{t}=N_\mathrm{r}=32$.
• Figure 3: Rx NF array factor in the range domain under the distance sampling method assumption CuiDai22, obtained focusing at multiple distances $\overline{r}$ from the transceiver at $\overline{\theta}=0^\circ$ with $f_\mathrm{c}=60\,$GHz, $d_\mathrm{t}=\lambda/2$, $d_\mathrm{r}=N_\mathrm{t}\lambda/2$ and $N_\mathrm{t}=N_\mathrm{r}=32$.
• Figure 4: Range profile obtained focusing locations at different distances ($\overline{r}=1,7,15\,$m) from the transceiver at $\overline{\theta}=0^\circ$. Note that, with $f_\mathrm{c}=60\,$GHz, $B=200\,$MHz, $K=1024\,$, $d_\mathrm{t}=\lambda/2$, $d_\mathrm{r}=N_\mathrm{t}\lambda/2$ and $N_\mathrm{t}=N_\mathrm{r}=32$, the effective NF region and super-resolution area are limited by $r_\mathrm{DF,r}=376.1\,$m and $r_\mathrm{sr}=11.9\,$m, respectively.
• Figure 5: Range-resolution map $\Delta r(x,y)$ over the monitored area for (a) $B=50\,$MHz, (b) $B=200\,$MHz, and (c) $B=750\,$MHz by using the ELAS design with $f_\mathrm{c}=60\,$GHz and $N_\mathrm{t}=N_\mathrm{r}=32$. These maps illustrate how the super-resolution region (low $\Delta r$ area) shrinks as the bandwidth increases. White stars labelled 1–3 indicate the ET centroid positions for the three simulation setups: (1) $\mathbf{p}_0=[5,3]^\mathsf{T}$, (2) $\mathbf{p}_0=[3.5,0]^\mathsf{T}$, and (3) $\mathbf{p}_0=[17,-8]^\mathsf{T}$.