---
title: 'DISAC: Distributed Integrated Sensing & Communication'
url: https://www.emergentmind.com/topics/distributed-integrated-sensing-and-communication-disac
type: topic
---

# DISAC: Distributed Integrated Sensing & Communication

Distributed Integrated Sensing and Communication (DISAC) denotes a class of systems in which geographically separated communication nodes are reused as sensing nodes and their observations are fused to support localization, tracking, mapping, and communications over a common network fabric. Recent literature also uses the acronym for “distributed and intelligent integrated sensing and communications” and for “Distributed Intelligent Sensing and Communication,” emphasizing that distribution is frequently coupled with semantic processing, local intelligence, and task-oriented information exchange rather than raw-data aggregation alone [2402.11630] [2402.18271] [2504.12765]. Across these formulations, the common principle is that sensing performance no longer arises from a single co-located transceiver, but from cooperation among base stations, access points, user devices, passive sniffers, or heterogeneous external sensors connected through fronthaul, backhaul, sidelinks, or higher-layer fusion interfaces [2606.18488] [2511.13104].

## 1. Terminology and conceptual scope

DISAC is not a single architecture. In the cell-free literature, it appears as a network in which distributed access points jointly serve users and sense targets through a central processing unit, with user-/target-centric operation and no fixed cell boundaries [2606.18488]. In distributed-MIMO work, it appears as cooperative multi-antenna access points that jointly provide communication, localization, and sensing, often with a central unit and flexible processing splits [2403.19785]. In multi-sensor ISAC, the same concept is described as the sensing counterpart of multi-user MIMO and, in radar terminology, as distributed MIMO radar, with hierarchical sensor-level, intra-site, and inter-site cooperation [2511.13104].

The literature also distinguishes DISAC from narrower ISAC interpretations. One line of work explicitly argues that conventional ISAC is too localized, too KPI-centric, and too focused on 6G radio signals alone, and therefore proposes DISAC as an architecture that incorporates distributed sensing nodes, heterogeneous external sensors, semantic composition of information, and AI-native control [2402.11630]. A closely related 6G-DISAC formulation defines the framework around three pillars—distributed architecture, semantic and goal-oriented communication, and high-resolution sensing—and extends network functions to object tracking, object handover, local processing, and semantic management [2402.18271].

This suggests that DISAC is best understood as an umbrella term. CF-ISAC is one concrete realization of it [2606.18488]; distributed antenna networks with RRHs coordinated by a central processor are another [2210.14880]; vehicular BS–UE multistatic sensing in FR3 is another [2509.25557]. The term therefore covers both infrastructure-centric and user-assisted sensing, and both centralized and decentralized coordination.

## 2. Architectural realizations and sensing geometries

The architectural diversity of DISAC is large, but several recurring templates can be identified.

| Architecture | Distributed elements | Representative papers |
|---|---|---|
| Cell-free ISAC | APs, CPU, fronthaul/backhaul, user-/target-centric operation | [2606.18488] |
| Distributed antenna network | RRHs coordinated by a CP, fronthaul-limited communication and sensing phases | [2210.14880] |
| D-MIMO / cell-free tracking | Cooperative APs with CU/global tracker and dynamic AP management | [2403.19785], [2511.03612] |
| BS–UE multistatic sensing | One BS illuminator, vehicular UEs as distributed receivers | [2509.25557] |
| BS–sniffer multistatic sensing | Distant BS illuminator, passive sniffers near protected site | [2606.29926] |
| Multi-sensor ISAC | General \(N\times N\) transmitter-receiver link matrix with CPCL and hierarchical fusion | [2511.13104] |

A fundamental geometric distinction is between monostatic, bistatic, and multistatic sensing. Cell-free ISAC “naturally support[s] multi-static sensing configurations without requiring full-duplex hardware at each AP,” because transmit and receive functions can be separated across APs [2606.18488]. Multi-sensor ISAC generalizes this idea into an \(N\times N\) link matrix \(H_{m,n}(f)=Y_m(f)/X_n(f)\), whose diagonal entries are monostatic links and off-diagonal entries are bistatic links [2511.13104]. In critical-infrastructure protection, the same multistatic principle is realized by placing three or more passive sniffers around a protected site while reusing a distant BS as illuminator, thereby improving both received power and geometric dilution of precision relative to quasi-monostatic sensing [2606.29926].

The multistatic geometry is often expressed through bistatic ellipsoidal constraints. In one generic form, each transmitter–receiver pair yields
\[
\left\Vert \bm r_{\rm tx_i} - \bm r_{\rm target} \right\Vert_2 + \left\Vert \bm r_{\rm rx_i} - \bm r_{\rm target} \right\Vert_2 = \tau_i c, \quad 1 \leqslant i \leqslant L,
\]
so localization is obtained from the intersection of multiple delay-defined ellipses or ellipsoids [2511.13104].

A distinct architectural theme is the “virtual aperture.” In the FR3 vehicular DISAC system, one BS transmits a MIMO-OFDM pilot while multiple vehicular UEs receive LOS, reflected, and clutter components. When different UEs observe the same target from different positions, the fusion center effectively acquires a larger spatial sampling extent than any single array could provide. The paper treats this as a distributed multiview aperture over space, not as synthetic aperture over time, and attributes gains in observability, target disambiguation, and localization accuracy to that geometry [2509.25557].

## 3. Signal models and processing pipelines

Despite their architectural differences, DISAC systems usually share a layered processing structure: local channel or echo parameter extraction, clutter suppression or direct-link handling, geometric association, and cross-node fusion.

In FR3 multistatic DISAC, the received signal at UE \(n\) on subcarrier \(k\) and OFDM symbol \(i\) is modeled as
\[
\mathbf{y}_{k,i}^{(n)} = \bigl[{\mathbf W}_{k,i}^{(n)}\bigr]^{\mathrm{H}} \mathbf{H}_{k,i}^{(n)} \mathbf{F}_{k,i}\,\mathbf{d}_{k,i} + \bigl[{\mathbf W}_{k,i}^{(n)}\bigr]^{\mathrm{H}}\mathbf{z}_{k,i}^{(n)},
\]
with a MIMO-OFDM channel including LOS, target-scattered, clutter, dense multipath, realistic UPAs, and inter-node timing offsets [2509.25557]. The UE-side chain then performs beamspace ESPRIT-based path estimation using a tensor/CP-decomposition formulation, field-of-interest angle filtering for clutter removal, pathwise geometric inversion, DBSCAN-based clustering and association for extended targets, and a global weighted least-squares estimator that jointly solves for UE positions, target-related ranges, LOS ranges, and timing offsets [2509.25557].

Cell-free ISAC models often formulate communication and sensing jointly at the AP level. A representative AP-wise transmit model is
\[
{x}_m = \sum_{k = 1}^{K} {w}_{mk} q_k + \sum_{t=1}^{T} {s}_{mt},
\]
while the UL-AP sensing observation includes direct-link interference and target echoes reflected through transmitter–target–receiver paths indexed by \((m,n,t)\) [2606.18488]. This model makes the DISAC coupling explicit: the same distributed transmissions that improve sensing geometry can also generate communication interference unless beamforming, power control, and clustering are jointly optimized.

The multi-sensor ISAC literature emphasizes a different issue: sparse OFDMA/TDMA frames distort naive FFT-based delay–Doppler estimation, so distributed sensing should move toward model-based estimation of \(\{\tau_p,\alpha_p,\gamma_p\}\) from irregularly occupied resource elements [2511.13104]. That line also stresses Cooperative Passive Coherent Location (CPCL), where communication-system synchronization, payload knowledge, and LOS reference recovery support bistatic delay–Doppler estimation without requiring every node to operate as a conventional monostatic radar [2511.13104].

A persistent methodological divide concerns where fusion occurs. Some pipelines rely on fully local measurement extraction followed by measurement-level geometric fusion; the FR3 virtual-aperture work explicitly does not develop fully coherent signal-level fusion across UEs [2509.25557]. Other architectures stress coherent beamforming and coherent echo fusion, and therefore place time, frequency, and phase synchronization at the center of system design [2606.18488]. This suggests that DISAC spans a spectrum from weakly coupled measurement fusion to strongly coherent distributed sensing.

## 4. Resource allocation, synchronization, and distributed constraints

Resource allocation in DISAC is fundamentally multi-domain: spectrum, power, time, fronthaul, synchronization budget, computation, and sensing accuracy are coupled.

A stochastic-geometry formulation makes the communication–sensing competition explicit by imposing
\[
p_s+p_c\le p_t,\qquad B_s+B_c\le B,
\]
and then characterizing the jointly achievable sensing detection probability \(P_D\) and communication coverage probability \(P_C\) under fixed total power and bandwidth budgets [2308.06596]. That analysis shows that the sensing–communication Pareto boundary contracts quickly as the number of distributed ISAC users grows, so densification without stronger coordination degrades both functions [2308.06596].

A different formulation appears in distributed antenna networks coordinated by a central processor. There, the ISAC frame is explicitly split into a communication phase and a sensing phase of durations
\[
(1-\eta)T,\qquad \eta T,\qquad 0<\eta<1,
\]
to avoid severe interference between information transmission and radar echo reception [2210.14880]. The sensing phase itself is pulse-radar structured, and \(\eta\) and the pulse duration \(t_p\) jointly determine \(R_{\min}\), \(R_{\max}\), fronthaul load, and total transmit energy. This time-separated formulation is optimized jointly with communication beamforming, sensing covariance matrices, and fronthaul participation variables under QoS, clutter, and power constraints [2210.14880].

Fronthaul and synchronization are recurrent bottlenecks. The CF-ISAC monograph treats time, frequency, and phase synchronization as prerequisites for coherent beamforming and coherent sensing fusion, and repeatedly identifies synchronization overhead, fronthaul limits, multi-target detection, and interference management as open problems [2606.18488]. The industrial distributed-MIMO measurement paper gives the practical counterpart: common \(1\)PPS and \(10\) MHz references, rubidium clocks, post-calibration and over-the-air calibration were required, and the inability to perform full back-to-back calibration prevented full phase-coherent inter-receiver combining in the sensing estimator [2403.02430].

The semantic DISAC literature adds another layer. It argues that distributed sensing will otherwise overload the network with raw data, so local data processing, semantic extraction, semantic composition, and goal-oriented communication should determine what information is exchanged upward and what remains local [2402.11630]. A related 6G-DISAC framework proposes flexible information models, semantic instructions, and distributed AI interfaces precisely because heterogeneous devices cannot all share the same compression, latency, or compute assumptions [2402.18271].

At a more abstract level, ISAC with distributed rate-limited helpers formalizes DISAC as a capacity–compression–distortion tradeoff. In that model, one helper observes the state sequence \(S^n\), another observes a feedback sequence \(Z^n\), and part of the state information is sent before transmission to aid communication while the remaining information is compressed jointly with feedback to aid sensing reconstruction under distortion constraints [2502.09166]. This suggests that rate-limited side-information exchange is not only an implementation detail; it is an information-theoretic core of distributed sensing-communication design.

## 5. Tracking continuity, association, and handover

As soon as sensing is distributed over multiple fields of view, DISAC must solve object continuity: how to associate measurements across nodes, how to hand over targets across coverage regions, and how to maintain identity under clutter and missed detections.

Extended-target association already appears at the measurement level. In the FR3 virtual-aperture work, each path at each UE is first treated as if it came from a separate point target, localized through per-UE weighted least squares, then clustered in the common spatial domain with DBSCAN before final multi-UE fusion [2509.25557]. This converts asynchronous angle–delay observations into candidate spatial points and performs both clustering and matching before the global estimator. The same paper notes that extended targets, dense multipath, and timing mismatch make path-to-target association an essential part of the DISAC stack rather than a negligible preprocessing step [2509.25557].

Trajectory-level continuity has been studied with random finite set methods. One handover approach uses a local trajectory PMBM filter at each BS and injects selected predicted trajectories into a neighboring BS not as Bernoulli tracks but as trajectory PPP intensity contributions. Handover is triggered by a probabilistic field-of-view entry test, and simulation with two BSs and two targets shows that the receiving BS can maintain trajectory continuity from the target origin rather than only from local first detection [2411.01871]. Another approach formulates multi-target handover as a belief-propagation problem on a factor graph, exchanging only selected target priors and, when useful, selected measurements between neighboring BSs. In the reported urban scenario, handover with measurements attains GOSPA comparable to centralized processing while reducing inter-BS exchange relative to full centralization [2506.23118].

User tracking exhibits analogous structure. A complete 3D cooperative tracking framework uses distributed APs, local delay/AoA extraction, measurement transformation into a common position domain, and a global PHD filter at a central unit. A field-of-view-aware AP management rule activates only the APs whose FoVs contain the predicted UE location, so tracking and AP scheduling are coupled [2511.03612]. This result is important because it shows that DISAC tracking accuracy need not require all distributed nodes to be active at all times.

These handover and association formulations also clarify a common misconception. DISAC does not require a single fusion philosophy. Some works use centralized global trackers or CPU-based fusion [2511.03612]; others preserve local tracking and exchange compact trajectory-level information only when targets approach neighboring FoVs [2411.01871]. A plausible implication is that scalable DISAC will rely on hybrid strategies rather than uniform full-network fusion.

## 6. Empirical evidence, applications, and open problems

The empirical record already shows that distributed sensing gains are measurable under realistic channels, but it also shows that gains depend strongly on geometry, synchronization, and processing choices.

In an industrial sub-6 GHz distributed-MIMO measurement campaign with \(13\) transceiver units, \(f_c=3.75\) GHz, and \(40\) MHz bandwidth, distributed transmission/reception produced average array gains of \(13.8\) dB and \(14.4\) dB in the two reported scenarios and eliminated deep fading dips. The same platform enabled cooperative uplink positioning with a trajectory-tracking mean-square error of \(18.4\) cm, despite only about \(35\) MHz effective sensing bandwidth [2403.02430]. The authors argue from these measurements that new channel models are needed that are spatially consistent and deal with nonstationary channel properties among distributed antennas [2403.02430].

At upper mid-band FR3, a practically grounded BS-plus-vehicular-UE DISAC system at \(15\) GHz with \(100\) MHz bandwidth, a \(32\times 32\) BS UPA, and \(8\times 8\) UE UPAs reports that in \(80\%\) of Monte Carlo snapshots the target localization error is below \(32\) cm and the UE positioning error is below roughly \(43\)–\(44\) cm. In a representative two-UE, two-target snapshot, weighted least-squares DISAC localized one target that single-UE ISAC failed to detect, and achieved \(0.075784\) m and \(0.049561\) m errors for the two targets [2509.25557].

In real measured indoor distributed-MIMO channels at \(5.6\) GHz and \(400\) MHz bandwidth, cooperative 3D user tracking with a global PHD filter and FoV-aware AP management achieved an average RMS trajectory error of \(0.09\) m while using, on average, \(4.78\) active panels out of \(8\). The paper explicitly reports that keeping all APs active all the time is unnecessary for maintaining nearly the same tracking performance [2511.03612].

System-level D-MIMO studies reinforce the communication side of the DISAC claim. In one scalable phase-coherent D-MIMO case study with \(200\) APs and \(5\) UEs, the simulated uplink spectral efficiency per UE at \(10\) dB SNR was \(5.93\) bits/s/Hz with joint ISAC support, \(3.00\) with localization only, \(4.55\) with sensing only, and \(0.357\) without ISAC support. The same study reports that phase-coherent D-MIMO positioning can reduce PEB from about \(15.22\) m to \(2.94\) m in a conventional scheme and from about \(9.41\times 10^{-4}\) m to \(1.82\times 10^{-4}\) m in a phase-coherent scheme as the AP count increases from \(4\) to \(12\) [2403.19785].

The application portfolio is correspondingly broad. Architectural DISAC work identifies smart-factory shop floors and vulnerable-road-user protection at smart intersections as representative use cases, emphasizing precision, safety, operational efficiency, real-time responsiveness, and heterogeneous device integration [2504.12765]. Critical-infrastructure protection work specializes this to lower-airspace monitoring, where one BS illuminator and three or more passive sniffers around a protected site provide better coverage and HDoP than a quasi-monostatic baseline [2606.29926].

Open problems recur with striking consistency. The CF-ISAC monograph lists synchronization and calibration across distributed APs, multi-target sensing and association, scalable cooperation and clustering, fronthaul capacity and latency, realistic clutter and target models, and deployment under mobility and hardware impairments [2606.18488]. DISAC vision papers add semantic representation, semantic-native metrics, flexible local/central processing splits, and standardization of new network functions for distributed sensing and semantic orchestration [2402.11630] [2402.18271]. Multi-sensor ISAC adds sparse-frame estimation, multistatic reflectivity, functional split, and distributed data fusion [2511.13104].

A recurring conclusion across these lines of work is that DISAC is neither a purely PHY-layer waveform problem nor merely a networking overlay. It is a networked sensing architecture in which geometry, synchronization, compression, semantic selection, association, and communication design are inseparable. That is the unifying technical lesson of the current DISAC literature.

Source: https://www.emergentmind.com/topics/distributed-integrated-sensing-and-communication-disac