---
title: 'Wi2SAR: Integrated Wireless and SAR Systems'
url: https://www.emergentmind.com/topics/wi2sar
type: topic
---

# Wi2SAR: Integrated Wireless and SAR Systems

Searching arXiv for the supplied IDs to ground the article in the cited papers.
Wi2SAR denotes a family of research directions at the intersection of wireless systems and search, sensing, or synthetic aperture radar. In the recent literature, the term is used in at least two distinct senses. One sense refers to an autonomous, drone-mounted Wi-Fi system for wilderness search and rescue that discovers and localizes victims through their personal devices by mimicking known Wi-Fi networks [2604.09115]. Another sense treats Wi2SAR as a unified wireless–SAR system in which the same waveform, RF front-end, and baseband processing serve both communication and SAR imaging, as in joint communication and SAR imaging under CP-OFDM and integrated communication and remote sensing in LEO satellite systems under ODDM [2403.17627], [2508.10317]. This terminological breadth places Wi2SAR within a wider continuum spanning Wi-Fi-based passive radar, wireless-assisted remote sensing, and radio-centric search-and-rescue systems.

## 1. Terminological scope and research usage

Wi2SAR is not used uniformly across the literature. In "Take Me Home, Wi-Fi Drone," Wi2SAR is the explicit name of a drone-based wireless system for wilderness search and rescue, and the acronym is tied to locating missing people through commodity Wi-Fi devices without relying on existing infrastructure [2604.09115]. In the integrated communication and sensing literature, by contrast, Wi2SAR is used more broadly to denote a unified wireless–SAR system where the same RF, waveform, and baseband processing serve both high-rate communication and high-resolution SAR imaging [2508.10317].

This split usage matters because the underlying problems are different. The wilderness-search interpretation is centered on victim discovery, direction finding, and autonomous drone navigation. The wireless–SAR interpretation is centered on waveform reuse, channel or reflectivity estimation, and imaging tradeoffs between communication rate and sensing fidelity. A plausible implication is that Wi2SAR functions less as a single canonical technical architecture than as a cross-domain label for systems that repurpose mainstream wireless mechanisms for SAR-adjacent or rescue-oriented objectives.

| Usage of “Wi2SAR” | Core function | Representative paper |
|---|---|---|
| Drone-based wilderness search and rescue | Discover and localize victims via Wi-Fi devices | [2604.09115] |
| Unified wireless–SAR system | Use one waveform/RF chain for communication and SAR imaging | [2508.10317] |
| JCASAR-related Wi2SAR framing | Joint communication and SAR imaging under random signaling | [2403.17627] |

A common misconception is to treat Wi2SAR as synonymous with only one of these meanings. The literature instead shows a broader concept space: infrastructure-free Wi-Fi localization for rescue on one side, and integrated communication–remote-sensing architectures on the other.

## 2. Drone-based Wi2SAR for wilderness search and rescue

In the wilderness-search formulation, Wi2SAR is an autonomous, drone-mounted Wi-Fi system designed for long-range, through-occlusion WiSAR operations, without relying on existing infrastructure [2604.09115]. Its basic insight is to leverage the automatic reconnection behavior of modern Wi-Fi devices to known networks. By mimicking these networks via on-drone Wi-Fi, Wi2SAR facilitates the discovery and localization of victims through their accompanying mobile devices. The system is infrastructure-free: it does not need cell towers, Wi-Fi installations, GNSS from the target, or any app on the victim’s phone.

The end-to-end workflow consists of incident reporting, exploratory search, victim discovery, guided search, and termination. An emergency contact provides the Last Known Position together with the SSID and password of the victim’s home or other known Wi-Fi network. The drone then launches and performs a grid or zigzag sweep around the Last Known Position while the onboard access points broadcast beacons at 100 ms intervals. When the victim device scans, sees the impersonated network, and attempts reconnection, Wi2SAR completes the WPA2-PSK handshake and labels that MAC/BSSID pair as the victim device or at least a high-confidence target. Subsequent packets are used for direction finding and guided convergence [2604.09115].

The rescue-specific novelty lies in the identification mechanism. Probe-based approaches are undermined by MAC randomization, whereas successful WPA2-PSK authentication shows that the device holds the victim’s credentials. This gives a persistent identity tied to the victim’s home network rather than to transient randomized probe behavior. The paper also notes that if several devices share the same SSID and PSK, multiple devices may connect; Wi2SAR therefore tracks each connected MAC separately.

This system is intended to address settings in which visual or thermal drones are blocked by dense canopy, rocks, bad weather, or low contrast; cellular or satellite connectivity is unavailable or weak; and the victim may be unconscious or unable to transmit location. In that sense, Wi2SAR reorients wilderness SAR from human-visible detection to device-centric radio discovery.

## 3. Architecture, hardware, and packet-level discovery pipeline

The drone implementation is built around a DJI Matrice 350 RTK platform and a Raspberry Pi Compute Module 4 connected via DJI E-port to obtain GPS and IMU data [2604.09115]. The RF front-end consists of a 3D-printed Luneburg Lens of diameter 15 cm, mounted under the drone with RF absorbing foam between the lens and the airframe, and 10 FPC Wi-Fi antennas attached to the lens surface: 1 at zenith, 3 on a 60° elevation ring, and 6 on a 30° elevation ring. The radio subsystem uses 5 Intel AX200 Wi-Fi NICs on PCIe risers, each running two virtual interfaces, one in AP mode and one in monitor mode.

The software stack is partitioned into a Victim Discovery Module, a Direction Finding Module, and a Drone Navigation Module. The Victim Discovery Module configures the access points with the same SSID and PSK as the victim’s network, broadcasts beacons, handles WPA2-PSK authentication, sniffs uplink packets on all antennas, builds per-packet RSS vectors using MAC address and sequence number, and filters and identifies the target device. The Direction Finding Module uses a pre-measured beam template of the lens to estimate incident direction from RSS only. The Drone Navigation Module uses DJI Payload SDK telemetry, rotates AoA from the body frame to the world frame, and implements a two-phase search scheme with a stop criterion based on elevation approaching \(90^\circ\) [2604.09115].

A notable architectural point is the multi-NIC, dual-mode design. Antenna-switching approaches do not observe the same MPDU on all ports and are sensitive to drone motion between samples. Wi2SAR instead uses each NIC to drive two antennas in MIMO, with AP mode handling beacons, ACKs, and data while monitor mode sniffs all frames and reports source MAC, sequence number, and RSS. RSS snapshot aggregation is keyed by source MAC and sequence number, producing a vector \(y \in \mathbb{R}^{N}\) with \(N=10\) antennas. Missing entries are filled by a very low RSS placeholder before the finalized snapshot is passed to the direction-finding module [2604.09115].

The implementation details are explicitly lightweight relative to phased-array AoA systems. Driver adaptations enable AP and monitor concurrency and promiscuous reporting in AP mode, which is necessary because pure monitor mode cannot send hardware-timed ACKs within SIFS. Reported processing performance is approximately 48 ms per direction-finding snapshot, with aggregate update rate around 7.8 Hz for up to 7 concurrent targets, CPU load between 48% and 52% on the CM4, and memory usage around 635 MB [2604.09115].

## 4. RSS-only direction finding with a Luneburg Lens and autonomous guidance

The direction-finding subsystem uses a spherical gradient-index Luneburg Lens with relative permittivity profile
\[
\varepsilon(r) = 2 - \left(\frac{r}{R}\right)^2,\quad 0 \le r \le R
\]
and refractive index
\[
n(r) = \sqrt{2 - \left(\frac{r}{R}\right)^2}.
\]
A plane wave entering from direction \((\theta,\phi)\) is focused to a point on the opposite surface of the sphere. In Wi2SAR, the lens is tuned mainly for 5 GHz Wi-Fi, is 3D-printed in PLA with gyroid infill to approximate the GRIN profile, and is characterized once in an anechoic chamber to obtain a static template \(B_0(\cdot)\) [2604.09115].

The estimator is explicitly RSS-only. For upper-hemisphere directions
\[
\Omega = \{ (\theta,\phi)\mid \theta\in[0,\pi/2],\ \phi\in(-\pi,\pi] \},
\]
the incident direction unit vector is
\[
\mathbf{u}(\theta,\phi) =
\begin{bmatrix}
\cos\theta\cos\phi \\
\cos\theta\sin\phi \\
\sin\theta
\end{bmatrix}.
\]
Measured RSS at antenna \(i\) is modeled in dBm as
\[
y_i = B_0^{\mathrm{dB}}\big(R_{\mathbf{u}}^{-1} L_i\big) + \beta + n_i,
\]
where \(\beta\) absorbs direction-independent terms such as transmit power, path loss, and RF-chain constants. After mean removal, the system solves
\[
\hat{\mathbf{u}}
= \arg\min_{\mathbf{u} \in \Omega}
\left\|\tilde{\mathbf{y}} - \tilde{\mathbf{s}}(\mathbf{u})\right\|_2^2,
\]
or equivalently
\[
\hat{\mathbf{u}} =
\arg\max_{\mathbf{u}\in\Omega}
\frac{\tilde{\mathbf{s}}(\mathbf{u})^\top \tilde{\mathbf{y}}}
{\|\tilde{\mathbf{s}}(\mathbf{u})\|_2 \,\|\tilde{\mathbf{y}}\|_2}.
\]
This formulation removes dependence on unknown absolute transmit power and average path loss, relying instead on the shape of the RSS pattern across antennas [2604.09115].

Empirically, the lens provides about 10 dB RSS gain in ground tests, enables reception at 384 m where a bare antenna cannot decode, and extends aerial victim discovery range by up to 104% in LoS and 91% in NLoS relative to bare antennas, depending on band and scenario [2604.09115]. The paper reports a global direction-finding statistic of \(\mathrm{MedPR} \approx 0.95\), corresponding to a median angular error of approximately \(18.4^\circ\), with less than 4% of samples having \(\mathrm{PR} \le 0\). In comparison with a CSI-based ArrayTrack-style baseline, a 2D version of Wi2SAR using the Luneburg-Lens RSS method attains median error around \(6.0^\circ\) over full \(360^\circ\) FoV, while ArrayTrack without continuous recalibration degrades to median error around \(46.0^\circ\) [2604.09115].

Navigation is divided into two phases. In Phase 1, the drone executes a pre-planned zigzag or grid pattern centered around the Last Known Position, with grid spacing chosen as twice the reliable operational range of the lens-based Wi-Fi link. In Phase 2, each new RSS snapshot yields an AoA estimate in the drone body frame, which is rotated to the world frame using the IMU, and the drone moves along that direction. The stop criterion is triggered when the elevation angle approaches \(90^\circ\) and remains stable, at which point the drone’s GPS coordinate is reported as victim location [2604.09115].

Field results are concrete. In a \(400\times 400\) m search area, the system discovered all 5 phones in about 13.5 minutes with 100% discovery rate. In a single-blind \(40{,}000\ \mathrm{m}^2\) forest trial, the first auto-reconnect occurred at 115 s, the stop criterion triggered at 224 s, and the final horizontal localization error was approximately 5 m [2604.09115].

## 5. Wi2SAR as integrated communication and SAR imaging

A different research line uses Wi2SAR to denote systems in which the same waveform and hardware perform both communications and SAR imaging. In "Waveform Design for Joint Communication and SAR Imaging Under Random Signaling," the system is a monostatic SAR carried by a UAV, flying parallel to the \(y\)-axis at altitude \(H_p\), and simultaneously supporting SAR imaging and downlink communication to a user through a CP-OFDM waveform known at the UAV but unknown to the communication user [2403.17627]. The same OFDM signal illuminates the scene for SAR imaging and carries information to the user.

The communication model uses \(N\) subcarriers, spacing \(\Delta f\), bandwidth \(B=N\Delta f\), symbol duration \(T=1/\Delta f\), and cyclic prefix \(T_{\text{CP}}\). For Gaussian signaling,
\[
S_k \sim \mathcal{CN}(0,P_k),
\]
and the achievable per-symbol rate is
\[
C = B \sum_{i=0}^{N-1} \log\!\left( 1 + \frac{P_i |h_i|^2}{\sigma_n^2} \right).
\]
On the SAR side, range resolution is
\[
\rho_r = \frac{c}{2B},
\]
the swath is discretized into \(M\) range cells, and the CP must satisfy
\[
T_{\text{CP}} \ge M - 1 \quad \Rightarrow \quad T_{\text{CP}} = \frac{M-1}{B}
\]
for IRCI-free CP-OFDM SAR under the cited construction. Under the Swath Width Matched Pulse configuration, \(M=N\), and after discarding edge samples, range processing becomes
\[
\boldsymbol{y} = \boldsymbol{S}\boldsymbol{d} + \boldsymbol{w},
\]
with \(\boldsymbol{S}\) circulant and \(\boldsymbol{d}\) the unknown reflectivity vector [2403.17627].

Instead of matched filtering, the paper proposes least-squares range profiling:
\[
\hat{\boldsymbol{d}} = \left( \boldsymbol{S}^H \boldsymbol{S} \right)^{-1} \boldsymbol{S}^H \boldsymbol{y},
\]
with
\[
\mathrm{MSE} = \sigma^2 \sum_{k=0}^{N-1} \frac{1}{|S_k|^2}.
\]
Because \(\boldsymbol{S}\) is circulant, diagonalization by DFT makes the dependence of estimation error on per-subcarrier magnitude explicit. For deterministic constant-modulus OFDM under fixed total power, uniform power across subcarriers minimizes the MSE. For Gaussian signaling, the paper derives an expected MSE of
\[
\mathrm{EMSE} = A\sigma^2 \sum_{k=0}^{N-1} \frac{1}{P_k},
\]
where the constant \(A\) is obtained in practice by truncating a divergent integral over the Rayleigh amplitude distribution [2403.17627].

This leads to a convex optimization problem that trades SAR imaging quality against communication rate:
\[
\begin{aligned}
\min_{\{P_i\}} \quad & A\sigma^2 \sum_{i=0}^{N-1}\frac{1}{P_i} \\
\text{s.t.}\quad & \sum_{i=0}^{N-1} P_i = P,\quad P_i \ge 0, \\
& \log\det(\boldsymbol{I} + \boldsymbol{\Lambda}_P \boldsymbol{H}) \ge R_0 .
\end{aligned}
\]
The imaging-optimal limit yields uniform power allocation, whereas the communication-optimal limit yields the water-filling solution
\[
P_k = \left( \frac{1}{\nu} - \frac{\sigma^2}{|h_k|^2} \right)^+.
\]
The imaging chain is explicitly
\[
\text{raw data} \rightarrow \text{LS range profiling} \rightarrow \text{RCMC} \rightarrow \text{azimuth matched filtering} \rightarrow \text{SAR image}.
\]
Simulation results show that constant-modulus OFDM produces sharp range profiles and clear car-shaped imagery, Gaussian OFDM with uniform power is blurrier, and Gaussian OFDM with communication-optimal power allocation can make the car shape hardly recognizable, illustrating the communication–imaging tradeoff [2403.17627].

In this strand of work, Wi2SAR is thus not about locating a person’s handset but about reusing communication waveforms for SAR imaging, with LS-based reflectivity estimation and subcarrier power allocation as the central technical levers.

## 6. Delay–Doppler Wi2SAR, passive Wi-Fi radar, and coexistence context

An even more explicit integrated wireless–SAR architecture appears in "Integrated Communication and Remote Sensing in LEO Satellite Systems: Protocol, Architecture and Prototype," which describes a LEO satellite system using a unified ODDM waveform and a single RF front-end for both broadband service and SAR-based remote sensing [2508.10317]. The system uses a delay–Doppler domain representation with DD-orthogonal pulses, and after matched filtering the discrete DD-domain input–output relation becomes an exact 2D circular convolution. A pilot ODDM symbol serves simultaneously as a 5G-style downlink pilot and as a SAR pulse; the same channel-sensing block produces \(\widehat{\mathbf{H}^{\mathrm c}}\) for communications or \(\widehat{\mathbf{H}^{\mathrm r}}\) for SAR. The paper states that SAR range reconstruction is IRCI-free because it is obtained directly from estimated DD taps rather than from ambiguity-function convolution, and reports a sub-6 GHz simulation campaign together with a 28 GHz SDR prototype for real-time SAR imaging and information transmission [2508.10317].

This delay–Doppler formulation broadens the Wi2SAR design space beyond CP-OFDM. A plausible implication is that the unifying principle is not a specific waveform family but the reuse of communication-native signaling structures for radar-like inversion, channel sensing, or imaging under shared RF hardware.

Two adjacent literatures further frame the boundaries of Wi2SAR. One is passive Wi-Fi radar. "Hybrid Fusion for 802.11ax Wi-Fi-based Passive Radars Exploiting Beamforming Feedbacks" studies a passive Wi-Fi radar that listens to 802.11ax NDP sounding and unencrypted beamforming feedbacks, combining NDP-derived joint AoD/AoA with BFF-derived LoS AoDs in a maximum-likelihood fusion framework [2404.14269]. This is not an active Wi2SAR architecture in the same sense as the UAV or satellite systems, but it demonstrates how standard-compliant Wi-Fi control traffic can be repurposed for localization. The other is coexistence-oriented Wi-Fi–radar research. "Spectrum Sharing Between A Surveillance Radar and Secondary Wi-Fi Networks" formulates protection regions, aggregate-interference models, and Wi-Fi throughput under 3 GHz radar coexistence constraints, showing that radar awareness and realistic interference margins strongly affect viability [1602.00802]. That work addresses spectrum sharing rather than unified sensing, yet it defines the regulatory and interference environment in which some Wi2SAR deployments would have to operate.

Taken together, these lines of work show that Wi2SAR spans at least three technical regimes: rescue-oriented Wi-Fi discovery and localization; active integrated communication–SAR imaging with shared waveform and RF chain; and Wi-Fi-enabled sensing or coexistence mechanisms that reuse wireless signaling for radar functions or operate near incumbent radar systems. The term’s breadth is therefore not accidental but reflects an evolving convergence between commodity wireless protocols and traditionally specialized sensing pipelines.

Source: https://www.emergentmind.com/topics/wi2sar