Integrated Sensing & Communication

• ISAC is a wireless paradigm that integrates radar and communication functionalities using shared hardware, spectrum, and signal processing.
• It enables advanced applications such as positioning, localization, tracking, imaging, and environmental monitoring within 6G networks.
• ISAC leverages both integration and coordination gains to overcome spectrum congestion and eliminate duplicated system architectures.

Integrated Sensing and Communication (ISAC) is a wireless paradigm in which the same infrastructure, spectrum, waveforms, hardware architecture, signal processing platform, and information are used to support both data communication and sensing of the physical environment. In contemporary 6G formulations, sensing is not restricted to radar-style target detection; it also encompasses sensing, positioning, localization, tracking, identification, imaging, and environmental monitoring. The paradigm is motivated by spectrum congestion, duplicated hardware in separately engineered radar and communication systems, and the possibility of both integration gain and coordination gain when communication and sensing share resources and environmental information (Zhang et al., 9 Apr 2025, Wymeersch et al., 16 May 2025, Lu et al., 2023).

1. Conceptual scope and integration paradigms

ISAC is also discussed under the labels joint communication and radar (JCR) and dual-functional radar and communication (DFRC). Two taxonomies recur in the literature. One distinguishes radar-communication coexistence (RCC), where radar and communication systems share spectrum but may retain separate hardware and waveforms, from DFRC, where a fully shared hardware platform and a unified transmit waveform simultaneously serve communication and sensing. The other, proposed from a cross-layer 6G viewpoint, describes a four-level integration ladder: spectrum integration, hardware integration, waveform integration, and radio-resource integration (Liu et al., 2022, Wymeersch et al., 16 May 2025, Zhuo et al., 2024).

These taxonomies emphasize that ISAC is not merely “radar plus communications” assembled at the air interface. Several surveys explicitly frame it as a system-level evolution involving antennas, network topology, multi-modal sensing, edge intelligence, and standardization. In that broader view, communication seeks reliable information transfer through coding, modulation, and multi-antenna transmission, whereas sensing includes transmission, reception, and processing of radio signals to infer properties of the environment or targets. Positioning estimates the geometric state of a connected device; localization extends that idea to device-free targets, including detection, tracking, and characterization (Zhang et al., 9 Apr 2025, Wymeersch et al., 16 May 2025).

A recurrent misconception addressed in the literature is that ISAC is synonymous with waveform sharing. Multi-beam ISAC, for example, is explicitly characterized as a separated-signaling ISAC scheme in which communication and sensing occupy separate beams but remain coordinated at the beamspace level. This beamspace separation is especially prominent in mmWave and above, where directional beamforming is required by path loss and can be repurposed as the principal integration mechanism (Zhuo et al., 2024).

2. Mathematical models, metrics, and trade-off formulations

A common analytical starting point is the paired linear-Gaussian model

$$\bm{Y}_c = \bm{H}_c \bm{X} + \bm{Z}_c,\qquad \bm{Y}_s = \bm{H}_s \bm{X} + \bm{Z}_s,$$

where $\bm{X}$ is the transmit signal, $\bm{H}_c$ the communication channel, $\bm{H}_s$ the sensing or target response, and $\bm{Z}_c,\bm{Z}_s$ the noise terms. This abstraction supports information-theoretic, estimation-theoretic, and waveform-design formulations of ISAC (Lu et al., 2023).

One widely used unifying objective is the weighted mutual-information trade-off

$$\max \quad \rho I(X; Y_c | H_c) + (1 - \rho) I(H_s; Y_s | X),$$

which formalizes the balance between communication mutual information and sensing mutual information. For detection, the literature also uses Kullback–Leibler divergence (KLD), while for estimation it employs Cramér–Rao bound (CRB), Bayesian CRB (BCRB), and posterior CRB (PCRB). Communication performance is commonly described by channel capacity, achievable rate, SINR, BER, and spectral efficiency; sensing performance by CRLB, NMSE, probability of detection, probability of false alarm, accuracy, resolution, latency, and coverage (Zhang et al., 9 Apr 2025, Lu et al., 2023, Zhuo et al., 2024, Wymeersch et al., 16 May 2025).

Resolution formulas remain central because they connect waveform, aperture, and observation duration to sensing capability. Representative expressions used across the literature are

$$\Delta R = \frac{c}{2B}, \qquad \Delta v = \frac{\lambda}{2T_{\text{Tx}}}, \qquad \Delta \phi = \frac{0.89\lambda}{D}.$$

These capture the dependence of range, velocity, and angular resolution on bandwidth, coherent observation time, wavelength, and aperture, respectively (Wymeersch et al., 16 May 2025, Lu et al., 2023).

At the same time, the literature repeatedly notes that no single metric fully captures all communication and sensing objectives. Mutual information is elegant, but several works emphasize that its operational meaning is more direct for data transmission than for estimating delay, angle, or Doppler. This has led to parallel use of rate–distortion, CRB–rate, SINR–CRB, and, in short-packet systems, finite-blocklength rate–reliability–detection trade-offs. One recent formulation couples eMBB rate, URLLC decoding error probability, and target detection probability in a sensing-triggered finite-blocklength ISAC system (Lu et al., 2023, Nikbakht et al., 18 Jun 2025).

3. Architectural evolution: from colocated links to networked sensing fabrics

ISAC architectures span monostatic, bistatic, multistatic, multi-cell, and networked forms. In monostatic sensing, transmitter and receiver are colocated and therefore naturally share clock and transmitted signal knowledge, but this configuration entails a full-duplex-like self-interference problem. In bistatic and multistatic settings, transmitters and sensing receivers are separated, which increases flexibility and geometric diversity at the cost of tighter synchronization and fusion requirements (Wymeersch et al., 16 May 2025, Lu et al., 2023).

A concrete response to the monostatic bottleneck is the coordinated cellular network-supported multistatic radar architecture. In this design, a cluster of synchronized DFRC base stations shares spectrum, data, and CSI; one base station is selected as a multistatic receiver, while the others act as multistatic transmitters. Spatial separation of transmission and echo reception intrinsically circumvents the self-interference that arises when a base station transmits an ISAC signal and receives its own echo on LTE/5G time scales. The associated beamforming problem is non-convex, but alternating optimization with semidefinite relaxation yields a stationary point and, in the reported simulations, substantially outperforms monostatic baselines even when those baselines assume strong self-interference cancellation (Xu et al., 2023).

The literature also describes a shift from single-cell ISAC to multi-cell collaboration and then to explicitly networked ISAC. In a distributed cooperative sensing architecture, multiple users act as sensors that observe a sparse region of interest from different angles; instead of sending raw samples to a leader node, they exchange low-dimensional intermediate estimates using an adapt-then-combine (ATC) diffusion scheme. This removes reliance on a centralized fusion node and ties sensing performance to network topology through a steady-state MSE expression (Li et al., 2024).

At a more abstract level, network ISAC has been formulated on relay graphs in which each link budget is shared between communication flow and sensing rate. For a one-dimensional path network, the sensing–throughput boundary is exactly linear; for general topologies, the Pareto boundary is piecewise linear, with each segment slope determined by the minimum number of sensing-region links that must be converted from sensing to communication to gain additional throughput. This establishes a genuinely network-level trade-off beyond link-level beamforming (Andrews et al., 15 Jan 2026).

Multi-beam ISAC represents another major architectural line. Here the communication beams are generally time-invariant and stable, while the sensing beams are time-varying, scanning or tracking a wide field of view. The attraction of this architecture lies in reusing the same time-frequency resources while separating the two functions spatially in beamspace. The literature presents it as a middle ground between resource division and strict waveform sharing: more efficient than orthogonal partitioning, but more flexible than forcing both functions onto an identical waveform (Zhuo et al., 2024).

4. Waveforms, beamforming, learning, and programmable propagation

OFDM remains the dominant signaling substrate because of its compatibility with cellular systems and its natural mapping of delay and Doppler into frequency and slow-time phase structure. A representative downlink model writes the ISAC waveform as

$s(t)=s_p(t)+s_d(t),$

with pilot and data components embedded in a 5G NR-like frame. Phase-coded OFDM, especially with Golay sequences, has been proposed to improve anti-noise robustness and lower CRLB for delay and Doppler. On top of that waveform layer, iterative low-complexity algorithms have been developed: iterative 2D FFT for short-distance sensing and iterative cyclic cross-correlation for long-distance sensing, both intended to reduce quantization error while retaining FFT/DFT-type complexity (Wei et al., 2023).

Model-based beamforming is not the only design route. A data-driven alternative formulates ISAC as an end-to-end differentiable auto-encoder (AE) with trainable encoder, beamformer, radar detector, angle estimator, uncertainty estimator, and communication receiver. Its overall loss,

$$\mathcal{J}_{\text{ISAC}} = \omega_r \mathcal{J}_{\text{NLL}} + (1-\omega_r)\mathcal{J}_{\text{CE}},$$

balances radar and communication objectives directly in training. The reported numerical results show performance close to model-based benchmarks under ideal hardware, and markedly better robustness when antenna spacings are perturbed and the analytic model becomes mismatched to reality (Mateos-Ramos et al., 2021).

A large body of work uses reconfigurable surfaces to turn propagation itself into a design variable. In RIS-assisted ISAC, the surface can create virtual LoS links, enhance target illumination and echo reception, suppress interference and clutter, and improve communication QoS and sensing accuracy simultaneously. One tutorial-style review makes a sharper claim: joint sensing and communications designs are most beneficial when the sensing and communication channels are coupled, and RISs offer a means of controlling that beneficial coupling by subspace expansion and subspace rotation (Liu et al., 2022, Chepuri et al., 2022).

Target-mounted reflecting structures provide a more specific mechanism. In an IRS-assisted vehicular ISAC system with an IRS of $M$ sub-surfaces mounted on the target vehicle, ideal coherent combining yields an SNR gain of

$\bm{X}$0

while reducing the CRLB for range and velocity because both bounds scale inversely with $\bm{X}$1. Hybrid STAR-RIS architectures extend this logic to full-space coverage: passive reflective elements support local users and targets on one side of the surface, while low-power active transmissive elements strengthen links to distant users and targets on the other side, at the cost of thermal-noise amplification and higher RIS power consumption (Wei et al., 2022, Yigit et al., 2024).

Practical signal processing must also deal with clutter. In clutter-rich environments, one proposed framework uses multiple communication beams plus one scanning sensing beam, divides the service area into S4S and C4S sectors, and suppresses static clutter via mean phasor cancellation (MPC). Dynamic target detection and angle estimation are then carried out with angle-Doppler spectrum estimation (ADSE) and joint detection over multiple subcarriers (MSJD), while range and velocity are estimated using an extended subspace algorithm (Luo et al., 2023).

5. Application domains and representative system functions

Low-altitude security has become one of the most detailed application domains for ISAC. In this setting, low-altitude airspace is dense, heterogeneous, and hard to monitor with traditional radar or optical systems alone. A cellular ISAC architecture based on macro BSs, micro BSs, and future millimeter-wave BSs upgrades existing infrastructure with software-defined radio and improved antenna arrays, fuses multi-BS sensing data at the edge or cloud, and produces a unified airspace situational map. The application logic is organized into three loops: a sensing and fusion loop that produces a four-dimensional spatiotemporal map with position, velocity, heading, and timestamp; a communication and distribution loop that shares the map via low-latency, high-reliability 5G-A/6G communication and network slicing; and a decision-making and control loop in which aircraft, edge nodes, and central platforms perform collaborative conflict resolution and airspace control (Ren, 20 Jan 2026).

Within the same low-altitude-security framework, ISAC is used not only for tracking but also for cognition and trust. The sensing network captures raw channel state information, radar echoes, micro-Doppler, and high-resolution range profiles, from which lightweight AI models infer mechanical structure, approximate size, rotor number, rotor speed, and surface material. Higher-level intent inference fuses those physical signatures with temporary device identifiers, signal strength history, declared flight plans, electronic fence information, and geographic context, using temporal deep learning models or knowledge graphs to detect abnormal loitering, reconnaissance, formation gathering, straying, or malicious intrusion. A dynamic trusted authentication system then compares claimed identity with continuously observed behavioral fingerprints and outputs an integrated trust credential that binds digital identity, historical behavior, and traceable physical device history (Ren, 20 Jan 2026).

Human and environmental sensing extend ISAC well beyond classical radar tasks. A unified channel-frequency-response view decomposes sensing-relevant signatures into path gain, delay, Doppler, AoA/AoD, and nuisance phase offsets. In human sensing, phase cleanup is followed by Doppler-delay-AoA extraction for localization, vital-sign monitoring, and activity recognition. A reported vital-sign pipeline uses phase differencing and bandpass filtering at approximately 0.1–0.5 Hz for respiration and 0.8–2 Hz for heartbeat. In environmental sensing, longer-timescale amplitude and phase statistics have been used for rainfall, soil moisture, and water level; in one field experiment, ambient LTE signals from seven cells in a non-line-of-sight setting supported water-level estimation with RMSE = 7.36 cm (Wu et al., 18 Jul 2025).

UAV safety and anti-collision form another application class. One representative ISAC design fuses communication location information and radar sensing information using an Extended Kalman Filter (EKF), and replaces serial identification-friend-or-foe interrogation with a parallel ISAC-based procedure in which sensing and interrogation occur simultaneously. The reported simulations show up to 50% reduction in IFF/communication delay and 24.2% sensing accuracy improvement when communication and radar have the same sensing accuracy (Jiang et al., 2022).

ISAC has also been linked to time-sensitive traffic generation. In a bi-static MIMO setup, a sensing receiver detects the presence of a target of interest during an ongoing eMBB transmission and, upon detection, triggers a URLLC message in the next block. The transmitter uses dirty-paper coding to mitigate interference among sensing, eMBB, and URLLC layers, producing a finite-blocklength rate–reliability–detection trade-off. The reported numerical analysis shows higher eMBB rate than power-sharing and time-sharing baselines while satisfying both URLLC and sensing constraints (Nikbakht et al., 18 Jun 2025).

6. Open problems, controversies, and standardization

The literature remains explicit that ISAC is promising but incomplete. Survey papers repeatedly identify open issues in the information-theoretic limits of ISAC, in quantifying how much channel information can be inferred from sensory data, in characterizing true Pareto-optimal signaling, in super-resolution sensing under cellular constraints, in future sensing-network architectures, in cross-layer resource management and protocols, and in security, privacy, and multi-object multi-task recognition (Lu et al., 2023).

Synchronization is a particularly persistent bottleneck. Timing offset, carrier frequency offset, and random phase shifts are more damaging to sensing than to communication because small phase errors translate directly into range and Doppler bias. This difficulty is amplified in distributed and passive settings. In cooperative passive sensing with mobile communication systems, one proposed remedy is NLoS and LoS signal cross-correlation (NLCC), which uses the fact that LoS and NLoS components share the same synchronization offsets so that cross-correlation cancels CFO and TO before symbol-level multi-BS fusion (Wei et al., 2024, Lu et al., 2023).

Security and privacy are not peripheral issues. Low-altitude-security studies emphasize jamming, spoofing, hijacking, identity falsification, forged physical fingerprints, tampered GNSS positions, and the absence of mature cross-operator, cross-airspace authentication interoperability. Cross-layer 6G perspectives broaden the threat model to include spoofing, tampering, repudiation, disclosure, denial of service, linkability, identifiability, and non-compliance, and therefore treat sensing as a service that requires authentication, authorization, logging, provenance, confidentiality, integrity protection, and policy management (Ren, 20 Jan 2026, Wymeersch et al., 16 May 2025).

Another controversy concerns how much of ISAC should be optimized at the physical layer alone. Some works argue that average sensing accuracy or throughput are insufficient task objectives. A planning-oriented formulation for connected autonomous vehicles reallocates ISAC power toward planning-bottleneck obstacles, deriving a safety bound that links obstacle inflation to transmit power through CRB scaling and then embedding that bound in bilevel power allocation and motion planning. This suggests that future ISAC design may increasingly be task-oriented, with QoS defined by mission success rather than by isolated PHY metrics (Jin et al., 27 Oct 2025).

Standardization activity indicates that ISAC is moving from concept to service framework. In the cellular domain, 3GPP developments are described from Rel-15 through Rel-19 and beyond, including 3GPP TR 22.837 with 32 ISAC use cases, 3GPP TS 22.137 with 8 KPIs for sensing services, and study items on ISAC channel modeling. In Wi-Fi, IEEE 802.11bf addresses sensing measurement reports, waveform reuse, backward compatibility, and dynamic resource allocation, with the first version frozen in October 2024. At the regulatory and vision level, ITU’s IMT-2030 framework includes ISAC as a key 6G application scenario (Zhang et al., 9 Apr 2025).

Taken together, the arXiv literature portrays ISAC as a progression from shared waveforms and colocated beamforming toward distributed sensing fabrics, programmable propagation environments, AI-assisted inference, and application-level control loops. The field’s defining tension remains unchanged: the same resources must support both communication and sensing, but the most effective operating point depends on geometry, topology, latency, hardware impairments, clutter, adversarial pressure, and the task that the sensed information is meant to enable.