Source-linked AI summary
Integrated Sensing and Communications Over the Years: An Evolution Perspective
Di Zhang, Yuanhao Cui, Xiaowen Cao, Nanchi Su, Yi Gong, Fan Liu, Weijie Yuan, Xiaojun Jing, J. Andrew Zhang, Jie Xu, Christos Masouros, Dusit Niyato, Marco Di Renzo
TL;DR
ISAC addresses the need for efficient spectrum use and lower hardware costs while supporting communication and precise sensing in B5G and 6G networks. This survey synthesizes its evolution across RF and optical spectrum, network architecture, sensing modalities, edge intelligence, and standardization, highlighting both integration opportunities and deployment constraints. It identifies hybrid architectures, collaborative sensing, multimodal processing, and standardization as central directions while noting optical-ISAC limitations under adverse weather and alignment conditions.
Problem
ISAC must support efficient spectrum use and lower hardware costs while serving B5G and 6G applications requiring communication and precise sensing.
Method
The survey synthesizes ISAC evolution across RF and optical spectrum, network architectures, sensing modalities, edge intelligence, security, and standardization.
Results
48.04 Gbps data rate and 1.02 cm sensing resolution were achieved in a W-band RF system using 16 GHz bandwidth, alongside progress in hybrid architectures and task allocation.
Takeaways & Limitations
The survey presents ISAC as moving toward heterogeneous, scalable, deployment-ready 6G systems through hybrid operation, collaborative processing, field-measured data, and system-level integration.
Takeaways & Limitations
Optical ISAC remains constrained by adverse weather, narrow beam divergence, limited field of view, and stringent alignment requirements.
Abstract
from arXiv · showhide
Integrated Sensing and Communications (ISAC) enables efficient spectrum utilization and reduces hardware costs for beyond 5G (B5G) and 6G networks, facilitating intelligent applications that require both high-performance communication and precise sensing capabilities. This survey provides a comprehensive review of the evolution of ISAC over the years. We examine the expansion of the spectrum across RF and optical ISAC, highlighting the role of advanced technologies, along with key challenges and synergies. We further discuss the advancements in network architecture from single-cell to multi-cell systems, emphasizing the integration of collaborative sensing and interference mitigation strategies. Moreover, we analyze the progress from single-modal to multi-modal sensing, with a focus on the integration of edge intelligence to enable real-time data processing, reduce latency, and enhance decision-making. Finally, we extensively review standardization efforts by 3GPP, IEEE, and ITU, examining the transition of ISAC-related technologies and their implications for the deployment of 6G networks.
A. Background and Motivation
Wireless communication and radar developed separately, creating inefficient resource use and duplicated infrastructure. ISAC emerged as a unified paradigm that co-designs waveforms, spectrum, and hardware for simultaneous sensing and communication.
- Separate communication and radar development created inefficient wireless-resource use and duplicated infrastructure.
- ISAC jointly designs waveforms, spectrum, and hardware to sense the physical environment while transmitting information.Sensing includes radar-type functions such as detection, tracking, recognition, behavior analysis, and environmental awareness.
Modal to Multi-
ISAC research has progressed from focused studies of spectrum, hardware, waveforms, and architectures toward richer AI-enabled and multimodal sensing perspectives. However, existing surveys remain fragmented, leaving a consolidated evolutionary account insufficiently developed.
- ISAC is transitioning toward large-scale deployment through industrial adoption and standardization initiatives, including IEEE 802.11bf and ITU recognition of ISAC as a core 6G application scenario.
- Early ISAC surveys addressed spectrum and hardware reuse, waveforms, MIMO, IoT architectures, signal processing, CSI-based sensing, and joint radar–communication foundations.
- Later work expanded toward physical-layer waveform construction, channel modeling, and interference management for dense or heterogeneous deployments.
- AI-, security-, and multimodal-sensing studies introduced adaptive transmission, secure architectures, data-centric learning, and richer environmental awareness.
- Existing surveys provide depth on individual themes but generally do not synthesize simultaneous evolution across spectrum, networks, sensing modalities, optimization, and standardization.
C. Contributions
The survey organizes ISAC evolution across spectrum, network architecture, sensing modalities, security, and standardization, while reviewing enabling antenna, RIS, and hardware technologies. It connects these dimensions into a system-level account for scalable 6G deployment.
- Sensing from RF to Optical: The survey compares RF and optical ISAC, emphasizing technical challenges, synergies, advanced antenna technologies, and hybrid RF–optical architectures.
- Network Architecture from Single-Cell to Multi-Cell: It traces network evolution from single-cell to multi-cell deployments, covering collaboration, synchronization, interference management, and resource allocation.
- Sensing Modalities from Single-Modal to Multi-Modal: It examines progression from single-modal to multi-modal sensing through edge intelligence, adaptive learning, and high-quality datasets for optimization and benchmarking.
- ISAC Security and Privacy Enhancement: It reviews dual-domain security and privacy vulnerabilities and opportunities for joint sensing and communication functionalities to enhance system-level security.
- Standardization Progress: The survey synthesizes standardization across 3GPP, IEEE, and ITU, including NR and Wi-Fi sensing, spectrum management, and low-altitude network applications.
- Advanced Antenna Technologies: Multi-antenna paradigms include centralized, distributed, and movable or fluid arrays, each offering distinct trade-offs in gain, coverage, diversity, adaptability, complexity, or field of view.
- Reconfigurable Intelligent Surfaces: RISs reconfigure wireless propagation through programmable reflection and transmission control, supporting beamforming, resource allocation, communications, radar, imaging, and environmental monitoring.
- Reconfigurable Intelligent Surfaces: RIS-assisted ISAC remains constrained by reconfiguration latency, wideband channel-estimation and calibration demands, RF-front-end integration, impedance matching, routing, and synchronization.
3) Summary and Discussion:
Optical ISAC expands joint sensing and communication beyond RF by using VLC, FSO, and photonic sensing across diverse deployment scenarios. Its benefits include abundant spectrum and fine spatial resolution, but weather sensitivity, alignment constraints, and missing evaluation frameworks limit scalability.
- RF-to-optical evolution: Optical ISAC leverages abundant unlicensed spectrum and shorter wavelengths to alleviate RF spectrum scarcity and improve spatial resolution.These advantages motivate optical alternatives to RF-based ISAC.
- Optical modalities: Optical ISAC is categorized into VLC, FSO, and photonic sensing for joint sensing and communication across diverse scenarios.VLC supports indoor localization and mapping, FSO supports long-range line-of-sight links, and photonic sensing supports detection, ranging, and environmental mapping.
- Optical limitations: Optical propagation remains vulnerable to fog, rain, smoke, turbulence, narrow beam divergence, limited field of view, and stringent alignment requirements.These impairments constrain precision and scalability, especially in mobile, distributed, outdoor, and long-range deployments.
- Optical limitations: Optical ISAC lacks a unified methodology for jointly evaluating communication throughput and sensing fidelity.This gap impedes joint waveform, resource-allocation, and protocol-layer co-design.
- Hybrid architectures: Hybrid RF–optical ISAC combines complementary architectures, including loosely coupled, tightly integrated, and functionally divided designs.Examples include unified VLC–FSO–RF access networks and LiDAR-aided millimeter-wave beam selection.
- Hybrid architectures: Hybrid resource allocation seeks to balance spectral efficiency, sensing accuracy, illumination constraints, latency, switching overhead, and fail-safe redundancy.Joint power allocation and deep reinforcement learning are proposed for adapting resources between RF and optical links.
3) Practical Implementation Challenges:
Practical hybrid RF–optical ISAC deployment is constrained by synchronization, hardware integration, channel-modeling, and waveform-design mismatches. Progress therefore requires coordinated metrics, adaptive cross-domain control, precise calibration, and hardware architectures that manage these impairments.
- Implementation challenges: Hybrid RF–optical ISAC deployment requires solving inter-modal synchronization, hardware co-integration, and unified channel-modeling challenges.These challenges arise from different timing requirements, hardware impairments, and propagation characteristics across the two domains.
- Synchronization: Optical front-ends demand picosecond-level timing and phase coherence, while RF subsystems typically tolerate sub-nanosecond clock jitter.Dual-loop synchronization is proposed to stabilize the optical pulse train and RF local oscillator.
- Hardware integration: Co-packaging RF and optical components imposes strict SWaP and thermal constraints, while nonlinear amplification can create persistent outage floors.Digital predistortion and photonic integrated circuits are identified as mitigation mechanisms.
- Channel modeling: RF fading models and optical turbulence, misalignment, and weather attenuation hinder unified analytical frameworks for joint sensing and communication optimization.The differing propagation characteristics complicate system-level modeling across modalities.
- System-level trade-offs: Hybrid RF–optical ISAC combines RF coverage and blockage robustness with optical resolution and bandwidth, but cross-domain challenges constrain scalability and efficiency.The paper identifies waveform design, task allocation, synchronization, integration, and channel modeling as continuing challenges.
- Task allocation: Assigning high-precision localization to VLC and control signaling to RF can improve energy efficiency and robustness.Sustaining these gains in dynamic environments requires real-time cross-layer coordination and predictive scheduling.
- Future directions: Future hybrid systems need cross-domain metrics, learning-based task offloading, sub-nanosecond synchronization, and integrated beam-alignment procedures.These directions target joint trade-offs among sensing resolution, communication throughput, and energy efficiency.
III. NETWORK ARCHITECTURE FROM SINGLE-CELL TO MULTI-CELL
ISAC is evolving from single-cell systems managed by one base station toward multi-cell architectures with joint sensing and communication across multiple nodes. This transition targets better robustness, sensing accuracy, and interference mitigation while introducing waveform, topology, and cross-layer optimization challenges.
- Single-cell ISAC: Single-cell ISAC uses one base station to manage sensing and communication in a controlled environment for waveform, interference, and resource-allocation design.Its restricted sensing range and inefficient spectrum optimization limit coverage, spectral efficiency, and scalability.
- Multi-cell ISAC: Multi-cell ISAC enables joint sensing and communication among multiple nodes to improve network robustness, sensing accuracy, and interference mitigation.The transition introduces challenges in waveform design, network topology, and cross-layer optimization.
A. Single-Cell ISAC: A Shift in Waveform Design Methods
Single-cell ISAC must allocate shared resources between communication and sensing, creating a fundamental capacity–distortion trade-off. Capacity–distortion theory and the CRB rate–region framework quantify this coupling in Gaussian channels.
- Resource coupling: Single-cell ISAC couples communication and sensing because allocating more resources to one function constrains the other.High-resolution imaging and target localization require wide bandwidth and fine time–frequency resolution.
- Waveform design: Capacity–distortion theory and the CRB rate–region framework quantify the communication–sensing trade-off in Gaussian channels.These frameworks relate communication capacity to sensing distortion or estimation performance.
1) Waveform with Orthogonal Resource Allocation:
Orthogonal resource allocation separates sensing and communication across time, frequency, space, or code, simplifying implementation and interference suppression. Its evolution toward integrated optimization improves coordination but leaves sensing-performance and interference challenges unresolved.
- Overview: Orthogonal allocation separates sensing and communication across time, frequency, space, or code domains to simplify transmitter design, receiver processing, and interference suppression.These approaches provide integration and coordination gains, but newer work increasingly explores non-orthogonal resource sharing.
- Time Division: Time division assigns distinct slots to sensing and communication waveforms and remains relatively easy to implement in existing systems.A 10-ms frame can allocate subframes for V2V/V2I communication and sensing.
- Time Division: Time-division ISAC has reached commercial pilots, including a 2023 ZTE trial detecting 0.01 m² UAVs within a 1 km range.Deployments in China, Korea, Europe, and the U.S. support its practical readiness in real-world settings.
- Frequency Division: Frequency division assigns separate subcarriers to sensing and communication, enabling vehicle detection and stationary-object identification near traffic lights.High-bandwidth implementations can suffer autocorrelation splitting, spurious peaks, and cross-band parameter inconsistencies.
- Spatial Division: Spatial division uses multi-antenna beamforming to reconcile radar energy projection with communication SINR, although their strongest paths may differ.Sidelobe control can embed communication information while maintaining radar coverage, but dynamic sidelobe fluctuations may reduce sensing accuracy.
- Code Division: Code division places sensing and communication signals on shared time-frequency resources using complementary coded scrambling to combat fading and mutual interference.The approach acts as a soft isolator in multi-user and multi-target scenarios, but codeword design must balance radar determinism with communication randomness.
- Evolution and Challenges: Communication-waveform reuse improves compatibility and cost efficiency, yet communication-centric designs can leave sensing performance suboptimal under imperfect synchronization and dynamic channels.The survey identifies unresolved interference characterization, performance limits, and sensing-optimal waveform design under realistic constraints.
3) Topology of Multi-Cell Network:
Multi-cell ISAC evolves from coordinated base stations toward heterogeneous, centralized, and space–air–ground architectures. These designs extend coverage and sensing diversity but increase synchronization, interference-management, and coordination demands.
- Collaboration of Multiple BSs: Cooperative multi-cell ISAC combines monostatic and bistatic sensing, while coordinated beamforming and interference nulling address dense-deployment interference.Simulations reported up to 78% rate gain over non-cooperative baselines, and decoded communication symbols can refine target estimation without excessive pilot overhead.
- Collaboration of Macro and Micro BSs: Macro–micro collaboration improves localized perception and indoor–outdoor integration where macro base stations face penetration loss and coarse coverage.Strategic microcell deployment supports urban and indoor scenarios through localized coverage and heterogeneous cooperation.
- Collaboration of RRUs in C-RAN: C-RAN advances multi-cell ISAC by centralizing base-station functions and splitting processing across centralized units, distributed units, and remote radio units.This architecture supports coordinated processing but adds functional and synchronization complexity.
- Space–Air–Ground Integration: Space–air–ground networks extend sensing and connectivity across satellites, aerial platforms, and terrestrial infrastructure for monitoring, disaster response, and remote communications.They support large-scale data acquisition but face propagation delay, Doppler instability, rapidly varying interference, and real-time processing constraints.
- Summary and Discussion: The architectural progression from single-cell to multi-cell and SAGIN improves coverage, spatial diversity, and sensing scale while increasing synchronization, interference, and coordination challenges.Clock asynchronization remains a fundamental limitation because scalable solutions for dynamic, multi-node deployments are still limited.
IV. SENSING METHOD FROM SINGLE-MODAL TO MULTI-MODAL (EDGE PERCEPTION)
ISAC is shifting from single-modal sensing toward multi-modal edge perception that combines heterogeneous data with decentralized processing. This integration supports lower latency, reduced communication burden, privacy preservation, and richer environmental understanding.
- From Single-Modal to Multi-Modal Sensing: Multi-modal ISAC integrates heterogeneous data sources, while edge intelligence enables decentralized real-time processing, lower communication overhead, and improved privacy.The transition is motivated by high-precision decision-making needs in autonomous driving, smart cities, and industrial automation.
- Challenges: Edge intelligence avoids some cloud-centric costs, but large-model transmission, fading, device heterogeneity, and adversarial participants complicate federated training.These constraints affect network load, model convergence, and sensing-input integrity.
- Edge Perception Architecture: Edge perception co-locates sensing, computing, and communication modules to analyze environmental data near end devices before transmitting extracted features.Localized decision-making reduces latency and alleviates backhaul-network burden.
- Applications of Edge Perception: Multi-modal edge fusion improves environmental perception across smart-home, industrial, healthcare, and UAV-swarm applications.AI-based fusion captures subtle patterns and supports predictive maintenance, prompt intervention, and autonomous conflict avoidance.
- Communication Enhancement: Fusing radar, camera, LiDAR, and RF inputs within communication processes can improve interference mitigation, beamforming precision, spectrum use, and resource management.The passage presents vehicular networks as an example of communication gains from localized multi-modal processing.
4) Interplay between ISAC and AI:
The paper presents ISAC and AI as mutually reinforcing: AI optimizes sensing and communication decisions, while ISAC supplies aligned multimodal data for edge inference. Applications show gains in efficiency, accuracy, adaptability, and low-latency operation, alongside synchronization, robustness, energy, and deployment constraints.
- AI-Enabled ISAC: AI-enabled ISAC uses lightweight, adaptive models to optimize beamforming, waveform selection, and sensing strategies under edge-resource constraints.A shallow neural network for hybrid beam selection achieved real-time operation with a 40% computational-cost reduction while maintaining near-optimal performance.
- ISAC-Facilitated Edge AI: ISAC-facilitated edge AI fuses radar, LiDAR, vision, and RF features into a unified, temporally aligned representation for inference.This multimodal representation supports applications including vehicular beamforming, indoor activity recognition, and UAV decision-making.
- AI–ISAC Co-Synergy: AI–ISAC co-synergy improves learning workflows by using sensing quality for client scheduling and power control, while transfer learning lowers communication overhead.The cited framework reports preserved accuracy and privacy with reduced training latency and communication burden.
- Representative Results: Reported results span recognition, sensing, communication, and learning, including 96.43% subject-independent accuracy, 3.7 dB sum-rate gain, and 1.6x faster federated-learning convergence.These figures come from separate application results and should not be interpreted as one unified benchmark.
1) The Shift towards Task-Oriented ISAC:
Task-oriented ISAC couples sensing, communication, and edge computing around efficient and reliable task execution rather than isolated component performance. The section connects this shift to AI-based beamforming, multimodal sensing datasets, and deployment choices governed by latency and model constraints.
- Task-Oriented Design: Task-oriented ISAC couples sensing, transmission, and edge computation so devices optimize task performance under communication and computing constraints.The paper argues that ISAC QoS should be reconsidered from a task-oriented perspective.
- AI for ISAC: AI transforms ISAC into adaptive, context-aware infrastructure, including lightweight neural-network beamforming for reduced complexity and improved spectral efficiency.The cited RIS-aided system illustrates AI support for joint sensing and communication design.
- LLM-Enabled ISAC: Large language models can support multimodal sensing reasoning through perception, grounding, and alignment modules, including motion and environment inference from transformed sensor signals.Their use in ISAC remains constrained by latency, model complexity, and hardware limitations.
- Deployment Choices: Edge or hybrid deployment suits real-time beam prediction and mobility management, whereas cloud deployment better fits latency-tolerant semantic reasoning and long-term optimization.Relevant KPIs include inference delay, power consumption, model size, and Top-κ accuracy.
- Open Challenges: LLM-enabled ISAC offers a pathway toward scalable intelligent S&C infrastructure, but interpretability, energy efficiency, and real-time constraints remain unresolved.These constraints limit how broadly LLMs can be deployed across ISAC use cases.
- Sensing Datasets: Comprehensive multimodal sensing datasets require defined objectives, heterogeneous data collection, and systematic cleaning and labeling procedures.The dataset-building process is presented as foundational for edge-intelligence research and AI applications in ISAC.
1) High-Quality Datasets:
High-quality datasets support ISAC and edge-intelligence research by providing controllable simulations and higher-fidelity field measurements, while distributed learning and integrated S&C introduce unresolved performance, security, and interoperability challenges.
- Dataset resources: SDP provides 400 hours of field-measured RF data spanning seven sensing scenarios and tasks including detection, tracking, recognition, vital signs, and imaging.It was launched by IEEE ISAC-ETI in February 2024.
- Dataset resources: Field-measured datasets offer higher fidelity, diversity, and representativeness than simulations for training models that generalize to practical ISAC scenarios.Simulation datasets remain useful because they provide controllable environments for performance analysis and algorithm optimization.
- Open challenges: Heterogeneous sensing data require distributed learning methods that account for device diversity and coupled application requirements, which current approaches often overlook.Ignoring device heterogeneity can produce inefficiency and degraded performance.
- Security and privacy: Integrated S&C expands privacy and security exposure through sensitive sensing information, shared infrastructure, intercepted signals, and hybrid attack surfaces.The cited challenges include target-location leakage, eavesdropping, spoofing, impersonation, and man-in-the-middle attacks.
- Security and privacy: Sensing-assisted physical-layer security can estimate eavesdropper locations and adapt beamforming or artificial-noise strategies without requiring directly obtained eavesdropping CSI.The reviewed approaches seek to degrade signals at eavesdropper locations while supporting legitimate users.
- Open challenges: Current security research remains limited by scarce hybrid attacker models, idealized assumptions about CSI and environments, and insufficient experimental validation.The paper calls for prototypes, datasets, hardware testbeds, and standards-oriented evaluation.
- Standardization: ISAC standardization is progressing across 3GPP, IEEE, and ITU, but sensing capability negotiation, triggering, result reporting, and protocol-stack design remain open interoperability issues.The standards discussion covers NR, Wi-Fi sensing, spectrum management, and emerging application domains.
VII. CONCLUSION AND FUTURE DIRECTIONS
The conclusion reviews ISAC’s demonstrated benefits and identifies future directions in channel modeling, multi-objective resource optimization, privacy, and security. It presents ISAC as increasingly capable of supporting adaptive services while emphasizing unresolved complexity, security, and hardware-integration challenges.
- Conclusion: ISAC has demonstrated benefits across intelligent transportation, smart cities, industrial IoT, security, sensing, and standardization domains.The paper frames these developments within 5G and beyond applications.
- Applications: Real-time road-condition data can support autonomous-driving safety and efficiency, while industrial IoT applications enable high-precision monitoring and predictive maintenance.The paper also describes adaptive infrastructure in smart cities and new services enabled by these innovations.
- Challenges: ISAC’s evolution introduces complex resource management, stronger security requirements, and challenges in scalable hardware integration.These challenges accompany the embedding of sensing functions into communication infrastructure.
- Future directions: Future channel models should represent multipath reflections, clutter, scattering, multiple bounces, and object-specific information for high-fidelity mmWave and sub-THz sensing.The paper identifies ongoing 3GPP RAN1 work as an important step but calls for richer models.
- Future directions: ISAC-aware scheduling should jointly balance spatial, temporal, and spectral resources while incorporating sensing KPIs such as angle resolution and Doppler sensitivity.RSMA and OTFS waveform design are identified as promising technologies for this direction.
- Future directions: Future ISAC systems require native security mechanisms, including federated access control, encrypted beamforming, and zero-trust architectures, aligned with data-protection regulations.The paper links these requirements to exposure of environmental and user-related data.