---
title: 'Ray Antenna Array (RAA): Cost-Effective Beamforming'
url: https://www.emergentmind.com/topics/ray-antenna-array-raa
type: topic
---

# Ray Antenna Array (RAA): Cost-Effective Beamforming

Ray antenna array (RAA) denotes a cost-effective multi-antenna architecture in which a large number of inexpensive antenna elements are arranged in a ray-like structure, each ray being a simple uniform linear array (sULA) with a deliberately designed orientation, while a ray selection network (RSN) connects appropriate rays to a small number of radio-frequency chains for subsequent baseband processing. Its defining feature is that the antenna elements within each sULA are directly connected, so that each ray forms a beam toward a direction matching the ray orientation without relying on analog or digital beamforming or any phase shifters. In contemporary wireless-communication work, RAA is developed primarily for high-frequency systems such as mmWave and THz, and later extended to integrated sensing and communication (ISAC), low-altitude UAV systems, and full-angle “omnicell” deployments [2505.23394][2505.10306][2509.05677].

## 1. Definition and structural organization

An RAA consists of \(N\) rays and \(M\) antenna elements per ray. Each ray is a sULA with inter-element spacing \(d=\lambda/2\), and the rays are placed in a ray-like structure around a reference axis with orientation angles \(\{\eta_n\}\) chosen to cover a target angular sector. In the communication-oriented formulation, the first element of each sULA is placed at radius \(D\) from an origin to ensure minimum inter-ray element separation, while all elements inside each sULA are directly connected to an RF combiner. The array therefore replaces the analog phase-shifter network of conventional hybrid analog/digital beamforming with passive combining inside each ray and switch-based selection across rays [2505.23394].

A standard steering-vector model for the \(n\)-th ray is
\[
\mathbf{a}(\theta,\eta_n)=\big[1,e^{j\pi \sin(\theta-\eta_n)},\ldots,e^{j\pi(M-1)\sin(\theta-\eta_n)}\big]^T,
\]
with an element-pattern and first-element phase term
\[
b(\theta-\eta_n)=e^{j\frac{2\pi}{\lambda}D\sin(\theta-\eta_n)}\sqrt{G(\theta-\eta_n)}.
\]
The directly connected ray response is then written as
\[
r(\theta,\eta_n)=M\,b(\theta-\eta_n)\,H_M\!\big(\sin(\theta-\eta_n)\big),
\]
where
\[
H_M(x)=e^{j\frac{\pi}{2}(M-1)x}\frac{\sin\!\big(\frac{\pi}{2}Mx\big)}{M\sin\!\big(\frac{\pi}{2}x\big)}
\]
is the Dirichlet kernel. This representation makes explicit that beam formation is realized by array geometry and coherent summation rather than programmable per-element phasing [2505.10306].

The RSN is modeled by a binary selection matrix \(S\in\{0,1\}^{N_{\mathrm{RF}}\times N}\) subject to
\[
\|[S]_{i,:}\|_0=1,\qquad \|[S]_{:,n}\|_0\le 1,
\]
so that each RF chain selects exactly one ray and each ray is connected to at most one RF chain. Because each sULA outputs a single combined port, the hardware count is dominated by antenna elements and RF switches, not phase shifters [2505.23394].

## 2. Beam physics and uniform angular resolution

The central theoretical claim of RAA is that it provides uniform angular resolution across all signal directions. With the ray beam pattern
\[
\mathcal{G}_{\mathrm{RAA}}(\theta,\theta')=M\sqrt{G(\theta-\theta')}\left|H_M\big(\sin(\theta-\theta')\big)\right|,
\]
the first nulls satisfy \(\sin(\theta-\theta')=\pm 2/M\), which yields the resolution formula
\[
\gamma_{\mathrm{RAA}}(\theta')=\gamma_{\mathrm{RAA}}=\arcsin\!\left(\frac{2}{M}\right),\qquad \forall \theta'.
\]
By contrast, for a conventional ULA,
\[
\gamma_{\mathrm{ULA}}(\theta')=\frac{1}{2}\arcsin\!\left(\sin\theta'+\frac{2}{M}\right)-\frac{1}{2}\arcsin\!\left(\sin\theta'-\frac{2}{M}\right),
\]
and
\[
\gamma_{\mathrm{ULA}}(\theta')\ge \gamma_{\mathrm{RAA}}(\theta')=\arcsin\!\left(\frac{2}{M}\right),
\]
with equality only at \(\theta'=0\). The nonuniformity of the ULA arises because DFT-style steering samples \(\sin\theta\), whereas RAA samples the local angular offset \(\theta-\eta_n\) [2505.10306].

This property drives the orientation design. Adjacent rays are chosen so that the first null of one ray aligns with the peak of the previous ray:
\[
\eta_n=n\cdot \arcsin\!\left(\frac{2}{M}\right).
\]
To cover \([-\phi_{\max},+\phi_{\max}]\),
\[
N=2\left\lfloor \frac{\phi_{\max}}{\arcsin(2/M)}\right\rfloor +1.
\]
To maintain minimum inter-ray spacing at the first elements,
\[
D\ge \frac{\lambda}{4\sin\!\big(0.5\cdot \arcsin(2/M)\big)}\approx \frac{M\lambda}{4}
\]
for \(M\gg 1\) [2505.23394].

Uniform resolution is paired with enhanced beamforming gain because each sULA covers only a sub-sector and therefore can employ more directive element patterns than a conventional ULA with wide-coverage elements. Using the 3GPP element pattern model, the peak gain increases as the element \(3\) dB beamwidth decreases, and RAA exploits this by assigning narrow element beams to individual rays rather than forcing each element to illuminate the entire field of view [2505.23394].

## 3. Communication models and optimization algorithms

In uplink multi-user communication, the effective signal collected at the \(N\) ray ports is
\[
\mathbf{x}=\sum_{k=1}^{K}\mathbf{h}_k s_k+\mathbf{z},
\]
with effective user channel
\[
\mathbf{h}_k=\sum_{\ell=1}^{L_k}\alpha_{k,\ell}\,\mathbf{f}(\phi_{k,\ell}),
\qquad
\mathbf{f}(\phi)=\big[f(\phi,\eta_n)\big]_{n\in\mathcal{N}},
\]
and after selection and baseband combining,
\[
y_k=\mathbf{w}_k^H S\mathbf{h}_k s_k+\mathbf{w}_k^H S\sum_{i\ne k}\mathbf{h}_i s_i+\mathbf{w}_k^H S\mathbf{z}.
\]
For fixed \(S\), the sum-rate-maximizing linear combiner is MMSE,
\[
\mathbf{w}_k(S)=C_k^{-1}(S)\,S\mathbf{h}_k,
\qquad
C_k(S)=S\left(\sum_{i\ne k}\mathbf{h}_i\mathbf{h}_i^H+\frac{M}{\bar P_t}I_N\right)S^H,
\]
while the single-user maximum-ratio solution gives
\[
\mathrm{SNR}(S)=\bar P_t\frac{\|S\mathbf{h}\|^2}{M},
\qquad \bar P_t=\frac{P_t}{\sigma^2}.
\]
Because exhaustive search over \(\binom{N}{N_{\mathrm{RF}}}\) ray sets is infeasible, a greedy ray-selection algorithm of complexity \(O(NN_{\mathrm{RF}})\) is used in the uplink, and an alternating optimization with SOCP-based beamforming is used in the downlink max–min-SINR problem [2505.23394].

Representative simulations establish the architecture’s performance in high-frequency sparse channels. With \(M=128\), \(N_{\mathrm{RF}}=8\), and \(K=8\), RAA significantly outperforms ULA-based hybrid analog/digital beamforming in both uplink and downlink when directional elements are used. In the single-user uplink, RAA attains approximately \(5\) dB higher maximum SNR than ULA across SNRs; in multi-user operation, greedy ray selection achieves near-exhaustive sum-rate performance for small systems and continues to outperform ULA for realistic sizes. A hardware-cost example with \(M=128\), \(N_{\mathrm{RF}}=16\), and \(\phi_{\max}=0.499\pi\) gives \(N=201\) rays and yields \(\mathrm{cost}_{\mathrm{RAA}}\approx \$46{,}278\) versus \(\mathrm{cost}_{\mathrm{ULA}}\approx \$268{,}700\), so RAA is approximately \(17.2\%\) of the cost of ULA+HBF [2505.23394].

## 4. ISAC and low-altitude UAV applications

RAA has been extended from communication-only designs to ISAC, especially for low-altitude UAV swarm scenarios in which off-boresight operation is routine and conventional arrays suffer severe sensing and communication degradation. In the OFDM-based UAV-swarm formulation, the wideband channel is modeled as
\[
\mathbf{h}(t,\tau)=\sum_{l=0}^{L}\alpha_l\,\mathbf{r}(\theta_l)\,\delta(\tau-\tau_l)e^{j2\pi f_{D,l}t},
\]
and after RSN selection,
\[
\mathbf{y}(t)=S\sum_{l=0}^{L}\alpha_l\mathbf{r}(\theta_l)e^{j2\pi f_{D,l}t}x(t-\tau_l)+S\mathbf{n}(t).
\]
AoA estimation is performed by constructing an angle-domain covariance matrix and applying MUSIC with spectrum
\[
P_{\mathrm{MUSIC}}(\theta)=\frac{1}{\mathbf{h}_{\mathrm{s}}^H(\theta)\,E_n E_n^H\,\mathbf{h}_{\mathrm{s}}(\theta)},
\]
followed by zero-forcing spatial filtering and a two-dimensional periodogram for delay–Doppler recovery [2505.10306].

The ISAC significance of the uniform-resolution theorem is direct: the same value
\[
\gamma_{\mathrm{RAA}}=\arcsin\!\left(\frac{2}{M}\right)
\]
holds for all directions, so moving aerial targets do not experience the angle-dependent resolution loss characteristic of conventional ULAs. In the low-altitude UAV-swarm setting, simulations with \(f_c=39\) GHz, \(M=128\), \(\eta_{\max}=\pi/2\), \(N=201\), \(N_{\mathrm{RF}}=8\), \(N_{\mathrm{sc}}=512\), and \(M_{\mathrm{sym}}=2048\) show significant performance improvement over conventional antenna arrays in sensing angular resolution and communication spectral efficiency [2505.10306].

A related low-altitude ISAC formulation optimizes joint ray selection and beamforming to maximize the minimum user SINR subject to sensing-SNR constraints. In that setting, the sensing metric is
\[
\gamma_s(U,W)=\sum_{k=1}^{K}\frac{\big|\beta\,\mathbf{r}(\theta)^H U^H \mathbf{w}_k\big|^2}{\sigma_s^2},
\]
and the communication metric is
\[
\gamma_k(U,W)=\frac{\big|\bar{\mathbf{h}}_k^H U^H \mathbf{w}_k\big|^2}{\sum_{i\ne k}\big|\bar{\mathbf{h}}_k^H U^H \mathbf{w}_i\big|^2+\sigma^2}.
\]
Because RAA provides uniform angular resolution and eliminates coverage holes, it is particularly suitable for low-altitude monitoring above a base station, where conventional arrays degrade in both gain and resolution [2606.19146].

## 5. Full-angle RAA and the omnicell paradigm

The original RAA formulations cover a prescribed angular sector. Full-angle RAA extends the orientation domain to the full azimuth, allowing the same cost–resolution tradeoff to hold over \(2\pi\). In this construction the base station is placed at the center of the cell, and the ray orientations expand to the full-angle domain so that the architecture supports the “omnicell” wireless communication system rather than conventional cell sectoring [2509.05677].

For large \(M\), the full-angle design uses approximately
\[
N\approx \lfloor M\pi \rfloor,
\qquad
D\approx \frac{M\lambda}{4},
\]
while preserving the same ray spacing
\[
\eta_n=n\arcsin\!\left(\frac{2}{M}\right).
\]
The intended result is that uniform angular resolution, high beamforming gain, and reduced hardware cost are now available in every direction, not merely inside a predefined sector. Under 3GPP UMa NLoS simulation settings, a full-angle RAA omnicell with \(M=64\), \(N=201\), and \(N_{\mathrm{RF}}=10\) achieves higher multi-user sum rate than both ULA- and UCA-based cell-sectoring systems, because the narrow non-overlapping ray beams reduce inter-user interference and avoid sector-edge degradation [2509.05677].

The cost argument persists at the system level. In a representative mmWave example, the omnicell RAA hardware cost is reported as \(\$28{,}892\), while the ULA sectoring system costs \(\$76{,}801\), so the full-angle RAA implementation is \(37.62\%\) of the ULA-sectoring cost. The paper positions this as a direct consequence of replacing large phase-shifter networks by RF switches and directly connected rays [2509.05677].

## 6. Limitations and three-dimensional successors

Despite its advantages, planar RAA has two recurrent limitations in the literature. First, because the rays are placed in the same plane, the architecture is prone to signal blockage. Second, in three-dimensional scenarios it has no elevation angle resolution capability. These weaknesses are especially visible in low-altitude UAV swarms and in XL-MIMO deployments where the angular scene is intrinsically three-dimensional [2603.17620][2512.07330].

One response is the spherical directly-connected antenna array (DCAA), which places multiple simple uniform planar arrays over a spherical surface. This design preserves the directly connected, phase-shifter-free philosophy of RAA while providing three-dimensional coverage and mitigating blockage. Its angular-resolution formulas are
\[
\gamma_v^{\mathrm{DCAA}}(\phi',\theta')=\arcsin\!\left(\frac{2}{M}\right),
\qquad
\gamma_h^{\mathrm{DCAA}}(\phi',\theta')=\arcsin\!\left(\frac{2}{M\cos\theta'}\right),
\]
so elevation resolution is uniform and azimuth resolution is elevation dependent. In UAV-swarm ISAC simulations, spherical DCAA yields fewer missed targets, lower RMSE in angle estimation, and higher spectral efficiency than conventional UPA with KPC-based HBF [2603.17620].

A second response is the cylinder DCAA, motivated explicitly as a remedy for the signal-blockage issue of planar ray arrays. It stacks multiple simple uniform circular arrays in a layered three-dimensional structure, partitions each sUCA into two directly connected semi-UCAs, and uses passive variable-length delay lines rather than phase shifters. In the dense-connectivity scenario with \(M=128\), \(N_{\mathrm{RF}}=30\), and \(N\approx 104\), the cylinder DCAA hardware cost is approximately \(\$89{,}560.64\), while ULA+HBF costs approximately \(\$1{,}511{,}427.84\), i.e. about \(5.93\%\) of the HBF cost. The same study reports higher uplink and downlink sum rates than ULA+HBF and convergence in no more than six iterations for the corresponding optimization algorithms [2512.07330].

## 7. Terminological ambiguity and adjacent meanings of “RAA”

A persistent source of confusion is that the acronym “RAA” is not unique. In the wireless-communication literature discussed above, it denotes Ray Antenna Array. In other research streams, however, the same acronym names different objects. “Rotatable Antenna Array” describes arrays whose elements or the whole array can rotate in three-dimensional space and whose design variables are boresight orientations rather than ray selection [2501.02595]. “Reconfigurable Antenna Arrays” denotes arrays whose individual radiators change radiation pattern, polarization, or operating frequency through PIN diodes, varactors, RF MEMS, or related actuation mechanisms [2510.17113]. In sparse RF-lens work, “RAA” can mean “reconfiguration of antenna array,” namely a placement strategy on the focal surface rather than a ray-arranged directly connected array [2306.16739]. In cosmic-ray instrumentation, “RAA” is also used informally for radio antenna arrays, including autonomous large-area arrays for extensive-air-shower detection [1301.2555][1905.04986].

The Ray Antenna Array literature is therefore best identified not by the acronym alone but by a specific cluster of defining properties: rays realized as sULAs, directly connected elements, RSN-based selection, absence of phase shifters, uniform angular resolution, and cost-effective beamforming in mmWave/THz and ISAC settings. Within that definition, the field has progressed from sector-limited planar designs to low-altitude ISAC, full-angle omnicell systems, and blockage-aware three-dimensional descendants, while preserving the original architectural premise that geometry and selection can replace much of the analog beamforming hardware [2505.18163][2606.19146].

Source: https://www.emergentmind.com/topics/ray-antenna-array-raa