Sensing-Guided Communication Dual-Function

• SGCDF is a dual-function design that integrates radar-like sensing with communication by sharing hardware, spectrum, and signal processing.
• It employs a decoupled beam-pattern design where a sensing-optimal transmit covariance is computed first and then minimally altered for communication feasibility.
• Empirical results demonstrate that SGCDF yields lower DoA RMSE and robust performance under high communication demand, maintaining effective beam-patterns.

Searching arXiv for papers on Sensing-Guided Communication Dual-Function and closely related ISAC/DFRC work. arxiv_search({"query":"\"Sensing-Guided Communication Dual-Function\" OR \"Integrated Sensing and Communication\" beam pattern design decoupled approach", "max_results": 10, "sort_by": "submittedDate", "sort_order": "descending"}) arXiv search results include recent ISAC/DFRC papers relevant to SGCDF, notably "Dual-Function Beam Pattern Design for Multi-Target ISAC Systems: A Decoupled Approach" (Junior et al., 27 Sep 2025) and the review "Integrated Sensing and Communication: Towards Multifunctional Perceptive Network" (Cui et al., 16 Oct 2025), along with related DFRC/OTFS/RIS works. arxiv_search({"query":"\"sensing-assisted communication\" dual-scale channel estimation ISAC", "max_results": 10, "sort_by": "submittedDate", "sort_order": "descending"}) Sensing-Guided Communication Dual-Function (SGCDF) denotes a dual-function integrated sensing and communication design in which sensing and communication share hardware, spectrum, and signal processing, while sensing outputs are used explicitly to guide communication decisions such as beamforming, link configuration, scheduling, and channel estimation. In the narrow sense used by the paper that introduces the term, SGCDF is a beam-pattern design methodology for mono-static multi-user multi-target ISAC systems: a sensing-optimal transmit covariance is first designed and then projected onto a communication-feasible set with minimal beam-pattern distortion. In the broader ISAC literature, the same design philosophy appears under closely related labels such as sensing-assisted communication, joint communication and sensing, dual-function radar-communication, and perceptive networks (Junior et al., 27 Sep 2025, Cui et al., 16 Oct 2025).

1. Terminology, scope, and conceptual status

The term SGCDF is not a universally standardized umbrella label across the ISAC literature. In "Dual-Function Beam Pattern Design for Multi-Target ISAC Systems: A Decoupled Approach" (Junior et al., 27 Sep 2025), it is defined as a beam pattern design methodology in which the sensing-desired covariance $\boldsymbol{R}_X^\star$ is designed first and then a dual-function covariance is found so that communication QoS is satisfied while the sensing pattern is minimally disturbed. In the review "Integrated Sensing and Communication: Towards Multifunctional Perceptive Network" (Cui et al., 16 Oct 2025), the same underlying idea appears at system level: communication waveforms and infrastructures act as radar-like sensors, and sensing outputs are exploited to improve beamforming, link adaptation, and resource scheduling.

This dual usage explains a common source of ambiguity. In a strict sense, SGCDF refers to the decoupled covariance design of (Junior et al., 27 Sep 2025). In a broader encyclopedic sense, it denotes a sensing-first dual-function operating mode within ISAC, especially where the same base station, access point, vehicle, roadside unit, or UAV performs both communication and sensing and feeds sensing results back into communication control. The broader literature aligns this with sensing-assisted communication technology, perceptive networks, and network-level ISAC (Cui et al., 16 Oct 2025, Liu et al., 2021).

The broader concept is grounded in two shifts described in the ISAC reviews. First, networks evolve from capacity-centric data pipes to multifunctional perceptive platforms. Second, sensing ceases to be a bolt-on feature and becomes natively integrated into service, architecture, and operation: device-free sensing, monostatic-to-networked sensing, and sensing-data transmission and fusion are all part of the same network fabric. This suggests that SGCDF is best understood as a coordination-oriented specialization of ISAC rather than a separate research area (Cui et al., 16 Oct 2025, Liu et al., 2021).

2. System models and mathematical foundations

A canonical SGCDF instantiation is the mono-static dual-function MIMO base station studied in (Junior et al., 27 Sep 2025). The base station has $M_T$ transmit antennas and $M_R$ receive antennas, serves $K$ single-antenna users in downlink, and senses $T$ point targets via direction-of-arrival estimation. It operates in full-duplex mono-static ISAC with perfect self-interference cancellation over a coherence interval of $L$ time samples. The transmit signal at snapshot $\ell$ is

$$\boldsymbol{x}_\ell = \boldsymbol{W}_C \boldsymbol{c}_\ell + \boldsymbol{W}_S \boldsymbol{s}_\ell = \boldsymbol{W}\tilde{\boldsymbol{x}}_\ell,$$

with communication precoder $\boldsymbol{W}_C$, sensing precoder $\boldsymbol{W}_S$, and augmented precoder $M_T$0. Under the large-$M_T$1 assumptions in the paper, the transmit covariance, or beam pattern matrix, is

$M_T$2

On the communication side, user $M_T$3 receives

$M_T$4

with a Rician channel model and user rate

$M_T$5

On the sensing side, the echo model is

$M_T$6

where $M_T$7 and the unknown parameter vector is the target-angle vector $M_T$8. Sensing performance is measured through DoA estimation MSE and controlled by the CRLB via the Fisher information matrix

$M_T$9

with the trace-opt criterion $M_R$0 used as the sensing objective (Junior et al., 27 Sep 2025).

At a more abstract ISAC level, the same coupling is often written as

$M_R$1

with communication rate

$M_R$2

and a sensing loss function $M_R$3 that may represent detection error, MSE, or CRB. This yields the generic sensing-guided design template

$M_R$4

which makes the dual-function precoder itself the locus of sensing–communication coupling (Cui et al., 16 Oct 2025).

3. Decoupled beam-pattern design methodology

The defining technical contribution of SGCDF in (Junior et al., 27 Sep 2025) is the replacement of a conventional joint CRLB-plus-rate formulation by a decoupled two-subproblem design. The original joint formulation minimizes

$M_R$5

subject to per-antenna power constraints and per-user SINR or rate constraints. After lifting to covariance variables, it becomes a high-dimensional SDP. The paper identifies two problems with this baseline. First, its complexity scales as

$M_R$6

Second, minimizing CRLB under communication constraints does not explicitly preserve the physical beam-pattern gain toward targets; under high communication load, the optimizer may strongly reshape $M_R$7 to satisfy user SINR, which can degrade practical DoA MSE (Junior et al., 27 Sep 2025).

SGCDF addresses this by separating sensing-optimal pattern design from communication feasibility. The first subproblem,

$M_R$8

subject to per-antenna power, $M_R$9, and $K$0, ignores communication altogether and yields the sensing-optimal covariance $K$1. The second subproblem,

$K$2

projects that sensing-optimal covariance onto the set of communication-feasible covariance matrices subject to per-antenna power, user SINR constraints, and positive-semidefinite splitting between communication and sensing covariances (Junior et al., 27 Sep 2025).

The Frobenius-norm projection is the central sensing-guided mechanism: communication feasibility is imposed, but deviation from the sensing-desired covariance is penalized directly. This preserves target-directed lobes and beam-pattern structure more faithfully than a pure CRLB-driven joint SDP. The paper reports a key empirical result: the conventional joint "Comm\Sensing" formulation can attain a lower CRLB than SGCDF and still exhibit worse empirical DoA RMSE, whereas SGCDF’s pattern preservation yields lower practical RMSE despite slightly larger CRLB values (Junior et al., 27 Sep 2025).

After solving the two subproblems, beamformers are recovered by rank-1 projection for the communication covariances, while the sensing covariance is factorized by SVD,

$K$3

This makes SGCDF both a covariance-domain optimization principle and a concrete beamformer synthesis method (Junior et al., 27 Sep 2025).

4. Waveforms, resource coupling, and sensing-guided communication mechanisms

SGCDF is not restricted to covariance design. In the broader ISAC literature, it includes any design in which sensing products shape communication-layer choices. The review (Cui et al., 16 Oct 2025) identifies several such mechanisms. The first is sensing-assisted beamforming: when scatterers constituting the communication channel align with sensing targets, sensing measurements enhance downlink beamforming accuracy. The second is predictive beamforming in high-mobility scenarios, where sensing observations $K$4 are used to predict the future channel

$K$5

and then choose

$K$6

A third is the use of electromagnetic maps, which provide quasi-static channel knowledge so that base stations can pre-configure narrow beams and reduce candidate beam search from 64 directions to 8 (Cui et al., 16 Oct 2025).

Waveform design supplies the observation substrate for these mechanisms. The literature distinguishes non-overlapped resource allocation, unified communication-centric waveforms such as OFDM, sensing-centric deterministic waveforms, and jointly optimized designs. Communication-centric OFDM and OTFS are especially prominent because they let sensing and communication share the same frame structure and often the same pilots or delay–Doppler processing chain. In the OTFS-based virtual-array design of (Wang et al., 2024), private TF bins assigned exclusively to specific transmit antennas synthesize a virtual array for high-resolution angle–range–velocity estimation, while the communication rate reduction per antenna is $K$7 symbols and, in the example with $K$8, $K$9, $T$0, and four private bins, the relative rate loss is about $T$1. This is a concrete example of sensing-guided resource allocation: the number of private bins is adjusted according to angular resolution requirements (Wang et al., 2024).

A different form of resource coupling appears in the dual-scale sensing-assisted communication design of (Zhiyue et al., 2024). There, the large-scale spatial correlation used by communication is estimated from sensing, with CRB-based angular error mapped into channel-space error covariance:

$T$2

Large-scale sensing duration $T$3 and the number of small-scale updates $T$4 are jointly optimized to maximize achievable rate subject to sensing accuracy constraints. For fixed $T$5 and $T$6, the optimal block partition is as uniform as possible,

$T$7

showing that CSI freshness and sensing accuracy are jointly schedulable design variables rather than separate subsystems (Zhiyue et al., 2024).

At the communication-centric extreme, dual-domain waveform superposition overlays a delay–Doppler-designed sensing signal onto legacy OFDM in the frequency–time domain. With a sensing signal power abatement of at least $T$8 dB relative to communication, the design can exceed legacy regulatory communication bandwidth for sensing, achieve up to $T$9 dB improvement in the CRB on delay estimation, and impose a negligible penalty on achievable rate. This establishes that SGCDF need not require communication redesign from scratch; it can also arise as a low-power sensing overlay whose outputs later steer communication control (Tagliaferri et al., 2022).

5. Algorithms, prototypes, and empirical behavior

The covariance-domain SGCDF formulation in (Junior et al., 27 Sep 2025) is paired with a low-complexity implementation, RMO-SGCDF, based on Riemannian optimization over the complex oblique manifold

$L$0

where per-antenna full power is enforced through

$L$1

The method uses Riemannian conjugate gradient iterations with tangent-space projection and retraction. For the sensing subproblem, the Euclidean gradient is

$L$2

while the communication-feasibility subproblem is written as a sum of squared distances to second-order-cone projections. The paper reports that RMO-SGCDF reduces average run-time by about $L$3 for the sensing subproblem and by more than $L$4 when both subproblems are included, while maintaining DoA RMSE close to full SGCDF and outperforming conventional joint methods (Junior et al., 27 Sep 2025).

Simulation behavior in the same paper is consistent with the sensing-first interpretation. With $L$5, $L$6, and $L$7 targets at fixed angles $L$8, $L$9, and $\ell$0, SGCDF and RMO-SGCDF achieve significantly higher beam-pattern gain at target angles than dual-function baselines and consistently lower RMSE than "Comm\Sensing", "ZF\Sensing", and "C\S Robust". As the overload factor $\ell$1 increases, all schemes degrade, but SGCDF and RMO-SGCDF show slower RMSE growth, indicating stronger robustness under high communication demand (Junior et al., 27 Sep 2025).

Hardware feasibility is supported by the OFDM-based SDR prototype in (Xu et al., 2022). That system uses a $\ell$2 MIMO downlink at $\ell$3 GHz with a weighted least-squares trade-off parameter $\ell$4:

$\ell$5

The testbed shows that, for omnidirectional sensing with $\ell$6, the dual-functional waveform remains within about $\ell$7–$\ell$8 dB of pure communication BER performance while maintaining an omnidirectional beampattern within about $\ell$9 dB. For directional sensing with the same $$\boldsymbol{x}_\ell = \boldsymbol{W}_C \boldsymbol{c}_\ell + \boldsymbol{W}_S \boldsymbol{s}_\ell = \boldsymbol{W}\tilde{\boldsymbol{x}}_\ell,$$0, the communication penalty is about $$\boldsymbol{x}_\ell = \boldsymbol{W}_C \boldsymbol{c}_\ell + \boldsymbol{W}_S \boldsymbol{s}_\ell = \boldsymbol{W}\tilde{\boldsymbol{x}}_\ell,$$1 dB, illustrating the stronger cost of stringent sensing constraints. The experimental result that dual-functional waveforms can maintain fine radar beampatterns and comparable BER to communication-centric solutions supports the practical relevance of sensing-guided trade-off design (Xu et al., 2022).

6. Applications, limitations, and open directions

SGCDF appears most naturally in scenarios where geometry changes faster than pilot-driven communication control can track. The review (Cui et al., 16 Oct 2025) highlights low-altitude airspace, where predictive beamforming and electromagnetic maps address UAV and eVTOL mobility; vehicular systems, where RF-SLAM and clutter suppression support predictive scheduling and blockage-aware beamforming; and indoor sensing, where delay–Doppler imaging from Wi-Fi or NR waveforms can guide AP selection, handover, and interference coordination. The same review also reports that ISAC-based schemes can yield more than $$\boldsymbol{x}_\ell = \boldsymbol{W}_C \boldsymbol{c}_\ell + \boldsymbol{W}_S \boldsymbol{s}_\ell = \boldsymbol{W}\tilde{\boldsymbol{x}}_\ell,$$2 throughput gain in some multi-antenna scenarios compared to conventional communication when localization error is small, indicating that sensing guidance can produce direct communication gains (Cui et al., 16 Oct 2025).

Several misconceptions are clarified by the current literature. SGCDF is not equivalent to simple waveform coexistence, nor is it synonymous with any dual-function radar-communication architecture. Its distinctive feature is explicit feedback from sensing to communication design. Conversely, a purely communication-constrained sensing design is related but not identical. The term itself is also not universal: (Junior et al., 27 Sep 2025) uses it for a specific decoupled beam-pattern method, while (Cui et al., 16 Oct 2025) uses closely aligned but broader terminology such as sensing-assisted communication and perceptive networks.

The limitations of current SGCDF formulations are also explicit. The beam-pattern design of (Junior et al., 27 Sep 2025) assumes narrowband flat-fading channels, perfect instantaneous CSI at the base station, perfect self-interference cancellation, point targets, known target count, no clutter or multipath in the radar channel, and a single-cell single-base-station setting. The dual-scale sensing-assisted communication model of (Zhiyue et al., 2024) assumes single-cell downlink, single-antenna users, MRT beamforming, and angular CRB as the sensing metric. These assumptions delimit current analytical tractability but also identify the main open directions: robust SGCDF under CSI and DoA uncertainty, extended-target and clutter-aware models, wideband and near-field operation, multi-cell and multi-node cooperation, and network-level coordination (Junior et al., 27 Sep 2025, Zhiyue et al., 2024).

Future work in the reviews is concentrated on AI/ML-enabled predictive functions, adaptive waveform and constellation selection, cooperative multi-source sensing, and protocol support for distributed inference and semantic-level data sharing. A plausible implication is that SGCDF will increasingly become a network-control abstraction rather than only a beamformer design method: sensing products such as electromagnetic maps, delay–Doppler images, and fused multi-node tracks will act as persistent side information for beam management, handover, interference control, and scheduling across perceptive networks (Cui et al., 16 Oct 2025, Liu et al., 2021).