Multistatic ISAC: Distributed Sensing & Communication

• Multistatic ISAC is a distributed sensing and communications approach that uses spatially separated nodes to jointly estimate target delay, Doppler, and angle.
• It leverages cooperative localization and synchronization to fuse bistatic measurements, enhancing geometric observability and robustness under interference.
• Advanced designs integrate waveform recovery, CRB-driven optimization, and state-space tracking to manage challenges like self-interference and synchronization.

Multistatic Integrated Sensing and Communications (ISAC) denotes ISAC deployments in which transmission and sensing reception are spatially separated across multiple nodes rather than co-located in a single monostatic transceiver. In the multistatic case, one or more transmitters illuminate a scene with communication waveforms, while multiple geographically separated receivers observe direct and target-reflected paths and cooperate to infer delay, Doppler, angle, position, velocity, or target tracks. Across recent work, multistatic ISAC is treated as a distributed MIMO radar problem embedded in a communication network, with particular emphasis on cellular downlink reuse, passive or semi-passive receivers, cooperative localization, synchronization by direct-path references, and joint sensing-communications resource optimization (Thomä et al., 17 Nov 2025).

1. Architectural concept and relation to monostatic and bistatic sensing

Multistatic ISAC is commonly distinguished from two simpler geometries. In mono-static ISAC, the same equipment transmits and receives the sensing echo. In bi-static ISAC, one receiver is used for sensing, separated from the transmitter. In multi-static ISAC, multiple receivers at different locations receive the reflected target echo and cooperate (Wang et al., 2024). In coordinated cellular formulations, one base station may act as the receiver while the other coordinated base stations act as transmitters, so sensing and communications coexist in the same time-frequency resources but the sensing transmit and receive functions are physically separated across base stations (Xu et al., 2023).

A central motivation for the multistatic architecture is structural avoidance of self-interference. Monostatic ISAC can suffer from severe self-interference because the transmitted waveform leaks into the sensing receiver chain, and the desired echo is often much weaker after round-trip loss. One coordinated-cellular example notes that for a target at $300$ m, the round-trip time is only about $2\,\mu s$, while LTE/5G symbol durations are on the order of tens of microseconds, so the transmit and receive phases overlap naturally; the same example reports an echo path loss at $2.4$ GHz of roughly $180$ dB (Xu et al., 2023). Multistatic deployments address this by physically separating transmit and sensing roles, thereby intrinsically circumventing the dominant self-interference bottleneck rather than relying only on cancellation.

Recent system papers also generalize the notion further. Multi-Sensor ISAC (MS-ISAC) is described as corresponding to multi-user MIMO communication and, in radar terminology, to distributed MIMO radar. In that view, with $N$ nodes, the full distributed MIMO matrix contains $N$ monostatic and $N^2-N$ bistatic measurements (Thomä et al., 17 Nov 2025). This formulation moves multistatic ISAC away from a single-link interpretation and toward a network service in which illumination, synchronization, estimation, and fusion are distributed across infrastructure nodes, radio units, sniffers, or user devices.

The architecture varies by application. Cellular proposals include one central BS transmitting OFDM signals and several neighboring BSs acting as cooperative receivers (Han et al., 2023); one gNB transmitting periodic 5G PRS while multiple distributed receivers estimate bistatic delay and radial velocity (Sagduyu et al., 3 Jul 2025); and one serving gNB actively illuminating a UAV while multiple software-defined sensor nodes act as passive receivers (Dickerson et al., 19 Jun 2026). Critical-infrastructure proposals assume at least three passive sniffing sensors near a protected site, registered as ordinary UEs and connected by standard downlink/uplink signaling to the illuminating base station (Thomä et al., 29 Jun 2026).

A common misconception is that multistatic ISAC is merely “more receivers.” The literature consistently treats it instead as a geometry-, synchronization-, and fusion-dependent regime. This suggests that the primary gain is not node count in isolation, but the combination of spatial diversity, favorable bistatic geometry, and cooperative estimation.

2. Geometric and signal-theoretic foundations

The core multistatic observable is the bistatic path length. In a canonical single-transmitter, multi-receiver geometry with transmitter at $\mathbf t$, receivers at $\mathbf r_m$, and target at $\mathbf p$, the bistatic delay/length is

$2\,\mu s$0

and the corresponding bistatic radial velocity is

$2\,\mu s$1

In $2\,\mu s$2D, each bistatic range measurement defines an ellipse, and three independent ellipses intersect at a unique point; in $2\,\mu s$3D, four measurements are needed (Sagduyu et al., 3 Jul 2025). A more general distributed-MIMO formulation expresses the same localization constraint as

$2\,\mu s$4

with each measured excess time of arrival defining an ellipse or ellipsoid (Thomä et al., 17 Nov 2025).

The received-signal models used across the literature differ by waveform, but they share the same physical content: delay, Doppler, path gain, steering vectors, and clutter. In a multistatic MIMO-OFDM cellular system, the received baseband signal at receiver $2\,\mu s$5 is modeled as

$2\,\mu s$6

with $2\,\mu s$7 and $2\,\mu s$8 (Han et al., 2023). In 5G-PRS-based sensing, after direct-path removal the residual at receiver $2\,\mu s$9 is modeled as

$2.4$0

where $2.4$1 and $2.4$2 (Sagduyu et al., 3 Jul 2025).

A separate but foundational issue is target scattering. Distributed ISAC papers emphasize that in multistatic sensing the decisive object is not only the communication channel, but the bistatic reflectivity of the target as a function of bistatic angle, frequency, distance, polarization, orientation, motion, and micro-motion. Because the target is moving, the reflectivity is treated as a time-varying function along the trajectory, not as a single scalar (Thomä et al., 2022). This is especially important for performance prediction, because each transmitter–receiver pair sees a different scattering aspect.

Recent synthetic-network analyses also formalize the geometry dependence through the vector

$2.4$3

which appears in both delay and Doppler gradients,

$2.4$4

making explicit that the same multistatic geometry governs both localization and velocity observability (Pu et al., 19 Nov 2025). This suggests that receiver placement and waveform scheduling are coupled design variables rather than separable stages.

3. Synchronization, asynchrony, and reference recovery

Synchronization is one of the defining technical difficulties of multistatic ISAC. In realistic deployments, the observed phase evolution mixes target motion with transmitter motion and hardware offsets. For an asynchronous multistatic configuration with one moving transmitter, one moving target, and $2.4$5 static receivers, the continuous-time CIR at receiver $2.4$6 is modeled as

$2.4$7

with

$2.4$8

so the measured phase contains TX-motion Doppler, target Doppler, receiver CFO, receiver phase offset, path-length phase, and noise (Bhalli et al., 16 May 2026).

A key recent result is a constructive solution to this unsynchronized regime. The method first subtracts the LoS phase from the target phase at each receiver,

$2.4$9

which cancels the receiver-specific CFO/PO term because it is common to both paths. A second temporal difference,

$180$0

removes the static path-phase term. Using measured AoAs, known receiver locations, and reconstructed transmitter and target positions, the originally underdetermined system reduces to four unknowns, and the paper concludes that the system is solvable if

$180$1

Thus, at least four receivers are needed to resolve target Doppler in the considered unsynchronized multistatic mobile-node setting (Bhalli et al., 16 May 2026).

A closely related experimental line demonstrates LoS-based compensation in hardware. An asynchronous $180$2 GHz IEEE 802.11ay prototype with one transmitter and two receivers compensates timing offset by aligning the first detected CIR peak, assumed to be LoS, and compensates carrier frequency offset using the LoS-path phase

$180$3

No dedicated synchronization link is used; both TO and CFO are estimated solely from the LoS path (Canil et al., 2023). The same direct-path principle reappears in Cooperative Passive Coherent Location (CPCL), where excess time of flight and excess Doppler are estimated relative to the direct BS-to-sniffer link, so explicit Tx–Rx synchronization is relaxed by differential coherent measurement (Thomä et al., 29 Jun 2026).

The reference-recovery problem becomes harder in sparse OFDMA/TDMA operation. MS-ISAC work therefore emphasizes model-based estimation rather than naive FFT processing, because sparse time-frequency occupancy distorts the ambiguity and scattering functions (Thomä et al., 17 Nov 2025). This suggests that synchronization and parameter estimation in multistatic ISAC are inseparable: direct-path recovery, waveform reconstruction, and geometric differencing are part of the sensing algorithm itself rather than a purely preliminary calibration stage.

4. Estimation, tracking, and optimization methodologies

Multistatic ISAC estimation pipelines span direct signal processing, geometric inversion, CRB-driven optimization, and state-space tracking. In 5G-PRS sensing, a coherent cross-ambiguity function is evaluated at each receiver to form a range–Doppler map,

$180$4

from which the strongest peak yields bistatic delay and radial velocity. The bistatic delays are then fused through nonlinear least-squares trilateration solved by Levenberg–Marquardt, while the radial-speed equations are inverted by ridge-regularized least squares,

$180$5

with the resulting state estimates smoothed by standard Kalman filters or extended Kalman filters (Sagduyu et al., 3 Jul 2025).

A distinct geometric approach augments bistatic range and bistatic range rate with DOA. In a 2-D multistatic ISAC system with one static receiver and $180$6 static transmitters, the target-path range is

$180$7

the bistatic range is

$180$8

and the BR/DOA relation yields

$180$9

Because DOA supplies angular information directly, the proposed method can conduct location estimation with a single TX-RX pair and velocity estimation with two TX-RX pairs, whereas prior 2SWLS methods require at least three TX-RX pairs (Zhuge et al., 2024).

Beamforming and receiver-selection formulations are often expressed through Fisher information and CRB objectives. One multi-static system with one transmitter and multiple receivers derives a closed-form CRB for joint estimation of transmission delay and Doppler shift,

$N$0

and minimizes it subject to communication rate, power, and cooperation-cost constraints. The mixed-integer non-convex problem is decomposed into an RE selection subproblem and a transmit beamforming subproblem, solved respectively by a minimax linkage-based clustering method and successive convex approximation (Wang et al., 2024).

More recent formulations optimize fairness across targets rather than only aggregate sensing quality. In an $N$1-fair multistatic MIMO-OFDM downlink, communication users also act as passive bistatic receivers, and the sensing objective is

$N$2

subject to per-user minimum rate and total transmit power constraints. The non-convex problem is reformulated with a smooth penalty and solved on a complex sphere manifold by Riemannian conjugate gradient (Noh et al., 31 Mar 2026).

At the network level, full joint processing can outperform staged fusion. In a space–time–frequency synthetic ISAC network, a concentrated MLE performs a network-wide normalized matched-filter search over position and velocity, whereas a two-stage information fusion method first estimates per-path delay and radial speed and then solves a weighted nonlinear least-squares problem via Gauss–Newton. The reported conclusion is that fully synthesized network processing is essential, because estimations by individual base stations followed by fusion are consistently inferior and unstable at low SNR (Pu et al., 19 Nov 2025).

Tracking closes the estimation loop. In OTFS-aided vehicular multistatic ISAC, triangulation from an anchor and multiple receivers is combined with Kalman filtering under a correlated random walk model, with the state transition

$N$3

to smooth position and velocity trajectories (Rani et al., 10 Feb 2026). Taken together, these methods show that multistatic ISAC estimation is not a single algorithmic family but a stack: direct-path recovery, per-link delay/Doppler/angle inference, geometric or CRB-based fusion, and dynamic-state tracking.

5. Deployment modes and application domains

The cellular domain has provided the dominant deployment templates. Coordinated cellular-network-supported multistatic radar architectures separate transmission and reception across coordinated BSs, while multistatic MIMO-OFDM cellular sensing uses one central BS as transmitter and several neighboring BSs as cooperative receivers to enable wide-area sensing at low cost (Xu et al., 2023). In 5G-native sensing, periodic downlink PRS or uplink DM-RS are reused as sensing waveforms, and multistatic processing is attached to standard scheduling, beamforming, and baseband functions rather than to dedicated radar hardware (Sagduyu et al., 3 Jul 2025).

UAV surveillance has become a prominent application. One adaptive 5G NR framework models sensing and communications as a shared resource-allocation problem in which a serving gNB sends active sensing waveforms while multiple software-defined sensor nodes act as passive receivers. Sensing feasibility is controlled by the sensing overhead

$N$4

and active sensing is feasible only if

$N$5

When active sensing is insufficient under congestion, external signals of opportunity provide supplemental passive sensing opportunities through distributed SDS nodes (Dickerson et al., 19 Jun 2026).

Critical-infrastructure protection motivates another deployment mode: passive sniffers near the protected site. In that model, at least three sniffers are deployed near the site, registered as ordinary UEs, and use the full downlink communication resource for passive bistatic sensing according to CPCL principles. Their local excess time-of-flight and excess Doppler estimates are returned via uplink for network-level fusion (Thomä et al., 29 Jun 2026). The same deployment logic emphasizes that multistatic sensing is especially attractive when the protected area is geographically bounded and can host multiple receivers.

Vehicular scenarios favor OTFS. In multistatic OTFS-ISAC, one anchor transmits and several spatially separated receivers capture bistatic reflections, while Kalman-based tracking exploits delay-Doppler robustness in doubly selective channels (Rani et al., 10 Feb 2026). Synthetic-network formulations extend the idea by combining multiple transmitters and receivers across time intervals and frequency bands to synthesize aperture, observation time, and bandwidth simultaneously (Pu et al., 19 Nov 2025).

Interference-limited coexistence creates yet another deployment regime. In tightly integrated low-earth-orbit satellite and terrestrial multistatic ISAC, each target echo is collected by one of $N$6 distributed radar receivers, and target–receiver association must explicitly account for angular proximity to satellite interference. The resulting architecture co-designs terrestrial beamforming, satellite power allocation, and multistatic-aware target association (Jee et al., 20 Nov 2025).

Experimental work confirms that the architecture is not only conceptual. A $N$7 GHz IEEE 802.11ay prototype demonstrates concurrent target tracking and micro-Doppler estimation from multiple viewpoints in an asynchronous multistatic setup (Canil et al., 2023). Conversely, Wi-Fi ISAC work argues that bistatic or multistatic Wi-Fi still poses multiple challenges for radar-like sensing and therefore redesigns Wi-Fi toward monostatic operation, which is a reminder that multistatic sensing is not uniformly preferable in every hardware regime (Chen et al., 2024).

6. Performance characteristics, misconceptions, and open problems

Reported performance trends are consistent but nuanced. In unsynchronized multistatic Doppler estimation with a mobile transmitter and multiple static receivers, error increases with AoA noise, while for $N$8 the reported performance is $N$9 Hz; larger inter-receiver separation improves Doppler estimation, and the method is reported to be largely invariant to target speed in the tested scenarios (Bhalli et al., 16 May 2026). In multistatic receiver-selection and beamforming, representative CRB values are about $N$0 for mono-static, about $N$1 for bi-static, about $N$2 for multi-static with $N$3, and about $N$4 for multi-static with $N$5, with performance improving as the number of cooperative receivers increases (Wang et al., 2024).

Localization results show that the multistatic advantage depends on the metric. One cellular MIMO-OFDM study reports that multistatic position RMSE is slightly higher than monostatic, because monostatic benefits from round-trip distance structure, but multistatic yields much better coverage and much lower velocity RMSE because it provides enough independent measurements to recover the full $N$6D velocity vector (Han et al., 2023). This directly contradicts the simplistic claim that multistatic is uniformly superior in every respect. A more accurate interpretation is that multistatic primarily improves observability, coverage, and robustness, while some single-metric advantages of monostatic geometry can persist.

System-level analyses in dense mmWave networks further sharpen this point. Dual-mode networked sensing papers identify a transition from monostatic-dominant to multistatic-dominant sensing as a function of self-interference cancellation efficiency. In the multistatic-dominant regime, using six multistatic BSs instead of a single bistatic receiver improved sensing coverage probability by over $N$7; the same analysis reports that dual-mode networked sensing with four cooperative BSs can double throughput, while multistatic sensing alone improves throughput by over $N$8 (Nabil et al., 19 Feb 2025). This suggests that the value of multistatic cooperation is greatest precisely when monostatic full-duplex assumptions become fragile.

UAV surveillance studies report strong geometry effects. In one 5G NR setup, average multistatic PEB drops from $N$9 m to $N^2-N$0 m as $N^2-N$1 increases from $N^2-N$2 to $N^2-N$3, while monostatic PEB stays around $N^2-N$4 m to $N^2-N$5 m (Dickerson et al., 19 Jun 2026). Experimental $N^2-N$6 GHz work likewise reports viewpoint-dependent micro-Doppler, with stronger Doppler at one receiver position and reduced Doppler as the bistatic angle increases toward $N^2-N$7, consistent with the factor

$N^2-N$8

(Canil et al., 2023). Multistatic processing therefore changes not only estimation accuracy but also the physical content of the measured signature.

Interference among coexisting ISAC nodes is another recurring limitation. In a shared-spectrum mono-/bistatic coexistence scenario, adaptive successive interference cancellation is shown to improve BER, EVM, and radar-image SINR. Reported gains include worst BER dropping from about $N^2-N$9 to $\mathbf t$0, average EVM improvement of $\mathbf t$1 dB at $\mathbf t$2 dBm, and average monostatic image-SINR gain of $\mathbf t$3 dB after bistatic LoS cancellation (Jeong et al., 28 Jul 2025). This indicates that multistatic gain can be nullified unless mutual interference is explicitly managed.

Open problems are repeatedly stated. They include receiver synchronization and clock bias estimation, target association across receivers, nonlinear trilateration local minima, sparse OFDMA/TDMA resource grids, clutter and sidelobes, waveform recovery from communication payloads, geometry-dependent ill-conditioning, cooperation overhead, and scalability of exhaustive multi-target association (Sagduyu et al., 3 Jul 2025). Propagation-modeling work adds the need for geometrically consistent, dynamic target reflectivity models and dedicated measurement ranges for extended, time-varying bistatic targets (Thomä et al., 2022). A plausible implication is that future multistatic ISAC research will be constrained at least as much by calibration, data association, and scene modeling as by nominal waveform design.

In aggregate, the literature presents multistatic ISAC as a distributed sensing architecture that reuses communication infrastructure for illumination, reception, synchronization, and fusion. Its defining benefits are structural mitigation of self-interference, spatial diversity, and improved geometric observability; its defining costs are synchronization complexity, cooperation overhead, and the need for robust fusion under clutter, interference, and partial observability.