PaperlayerView exact revision on arXiv

Source-linked AI summary

A Survey on Integrated Sensing, Communication, and Computation

Dingzhu Wen, Yong Zhou, Xiaoyang Li, Yuanming Shi, Kaibin Huang, Khaled B. Letaief

arXiv:2408.08074v3cs.ITcs.AIcs.LGeess.SP

TL;DR

The paper addresses the inability of partial sensing, communication, and computation integrations to meet extreme intelligent-task requirements. It surveys ISCC foundations, benefits, signal designs, resource-management strategies, future applications, and unresolved issues. ISCC is presented as a task-oriented approach for coordinating the three modules and improving resource utilization and task performance.

  • Problem

    Existing partial integration technologies face low resource utilization, mismatched module and task goals, and insufficient treatment of tightly coupled sensing, communication, and computation.

  • Method

    The paper provides a comprehensive survey of ICC, ISC, and ISAC, then examines ISCC benefits, challenges, signal designs, resource management, future networks, and open issues.

  • Results

    ISCC enables resource sharing, coordination, and joint optimization among sensing, communication, and computation under multiple objectives or customized task goals.

  • Takeaways & Limitations

    ISCC is positioned as a key technology for improving network resource utilization and achieving customized goals in complicated tasks and future intelligent networks.

  • Takeaways & Limitations

    Current ISCC resource-management strategies have relatively low generalization capability, creating high implementation costs across different settings.

Abstract

from arXiv · show

The forthcoming generation of wireless technology, 6G, aims to usher in an era of ubiquitous intelligent services, where everything is interconnected and intelligent. This vision requires the seamless integration of three fundamental modules: Sensing for information acquisition, communication for information sharing, and computation for information processing and decision-making. These modules are intricately linked, especially in complex tasks such as edge learning and inference. However, the performance of these modules is interdependent, creating a resource competition for time, energy, and bandwidth. Existing techniques like integrated communication and computation (ICC), integrated sensing and computation (ISC), and integrated sensing and communication (ISAC) have made partial strides in addressing this challenge, but they fall short of meeting the extreme performance requirements. To overcome these limitations, it is essential to develop new techniques that comprehensively integrate sensing, communication, and computation. This integrated approach, known as Integrated Sensing, Communication, and Computation (ISCC), offers a systematic perspective for enhancing task performance. This paper begins with a comprehensive survey of historic and related techniques such as ICC, ISC, and ISAC, highlighting their strengths and limitations. It then discusses the benefits, functions, and challenges of ISCC. Subsequently, the state-of-the-art signal designs for ISCC, along with network resource management strategies specifically tailored for ISCC are explored. Furthermore, this paper discusses the exciting research opportunities that lie ahead for implementing ISCC in future advanced networks, and the unresolved issues requiring further investigation. ISCC is expected to unlock the full potential of intelligent connectivity, paving the way for groundbreaking applications and services.

I. INTRODUCTION

Future 6G intelligent services require sensing, communication, and computation to operate together, but existing partial integrations cannot meet extreme task demands. ISCC addresses this gap by coordinating these modules and their competing resources under unified task goals.

  • 6G targets connected intelligence through enhanced and new usage scenarios, including integrated AI and communications, ISAC, and ubiquitous connectivity.
  • Intelligent tasks combine sensing for information acquisition, communication for transmission, and computation for processing and decision-making.
  • Existing ISC, ICC, and ISAC support only parts of intelligent tasks and fail to satisfy extreme requirements such as ultra-high decision accuracy and ultra-low latency.
  • Partial integration suffers from low resource utilization because resources are coordinated between only two modules rather than all three.In sensing- and AirComp-supported cooperative inference, joint allocation can adapt resource budgets to a time-varying environment.
  • Separate module objectives can diverge from the overall task goal, so improving sensing quality or transmitting data is not necessarily beneficial to task performance.For edge AI inference, sensing an unimportant view or transmitting irrelevant features may not improve inference accuracy.
  • Sensing, communication, and computation tightly compete for resources while jointly determining outcomes such as federated-learning convergence.ISCC coordinates and allocates resources among the three modules under a unified task goal.

C. Related Works and Our Contributions

The survey positions ISCC as a comprehensive framework built on ICC, ISC, and ISAC, then organizes its analysis around ISCC motivations, signal design, resource management, future applications, and unresolved issues.

  • Related partially integrated technologies: The paper surveys ICC, ISC, and ISAC, covering their definitions, goals, metrics, techniques, applications, categories, and relationships as foundations for ISCC.ICC includes MEC and AirComp, while ISC and ISAC are organized around wireless, multi-modal, and mobile crowdsensing.
  • ISCC motivations and challenges: The survey analyzes partial-integration shortages, ISCC benefits and functions, and challenges affecting resource utilization, task performance, and 6G support.
  • Signal design for ISCC: ISCC signal-design research covers beamforming and waveform methods targeting sensing CRB, communication SNR, and computation MSE.The survey distinguishes single-functional, dual-functional, and triple-functional designs, with triple-functional signaling trading higher design complexity for interference cancellation.
  • Network resource management: ISCC resource management has two paradigms: joint allocation for coexisting tasks and task-oriented allocation across tightly coupled processes within complicated tasks.
  • Future directions: The paper examines ISCC implementation in digital-twins, computing-power, and space-air-ground integrated networks, while identifying unresolved issues for future research.The issues include theory, hardware, protocols and standardization, security, privacy, robustness, and backward compatibility.

II. INTEGRATED COMMUNICATION AND COMPUTATION

Integrated communication and computation jointly designs communication and computation processes to improve task performance under shared resource constraints. MEC combines offloading and edge processing, while AirComp computes functions through waveform superposition during transmission.

  • ICC jointly optimizes communication and computation resources under unified objectives such as energy, latency, or computation accuracy.
  • Mobile Edge Computing: MEC moves transmission and processing toward edge servers, using task offloading to exchange communication resources for computation resources.
  • Cloud-based computation becomes burdensome as growing IoT deployments generate large volumes of data requiring centralized collection and processing.
  • Mobile Edge Computing: MEC research addresses joint resource allocation, dependent-task scheduling, and cooperative computing to manage limited resources and unavailable edge servers.
  • Over-the-Air Computation: AirComp treats synchronized co-channel waveform superposition as summation, enabling target functions to be computed during transmission without one-by-one decoding.
  • Over-the-Air Computation: AirComp uses transmit preprocessing, waveform summation, and receiver post-processing to compute nomographic and broader nonlinear functions.
  • Over-the-Air Computation: AirComp transceiver design compensates for channel-weighted summation over non-uniform fading channels to induce the desired signal alignment.

D. Discussion

ICC integrates communication and computation, while ISC integrates sensing transceivers with echo-signal processing. The discussion places these approaches within broader sensing technologies and highlights mobile crowdsensing for large-scale, low-cost data collection.

  • ICC jointly designs communication and computation with resource allocation directed toward unified system objectives.
  • MEC emphasizes joint communication-computation allocation, task scheduling, and cooperative computing, whereas AirComp emphasizes scenario-specific transceiver design.
  • ISC jointly designs sensing transceivers and subsequent sensing echo-signal processing.
  • Mobile crowdsensing uses built-in sensors in ubiquitous mobile devices to collect sensory data across domains for broad-coverage, low-cost sensing.

B. Wireless Sensing Technology

Wireless sensing develops radar waveforms, processing methods, and architectures for extracting target information such as range, velocity, direction, and motion. The surveyed progression moves from continuous-wave radar to pulsed and FMCW designs, then to multi-antenna architectures and intelligent processing.

  • Radar sensing transmits radio waves, processes echoes, and supports ranging, detection, and motion recognition.
  • Radar Waveform Design and Signal Processing: Continuous-wave radar cannot determine target range because echo phase is ambiguous without a phase basis.
  • Radar Waveform Design and Signal Processing: Pulsed radar obtains range information by periodically transmitting brief pulses separated by inter-pulse periods.
  • Radar Waveform Design and Signal Processing: FMCW radar improves time efficiency through continuous frequency-modulated transmission while detecting range and velocity with fast-time and slow-time processing.
  • Radar Architectures: Radar architectures include mono-static, bi-static, and multi-static systems, each trading structural simplicity, waveform flexibility, resolution, and observation diversity.
  • Radar Architectures: Phased-array radar electronically steers beams through antenna power and phase control, while MIMO radar uses independent waveforms for adaptive beamforming at higher processing complexity.
  • Intelligent Radar Signal Processing: AI-based radar processing addresses tasks including intelligent waveform and array design, radiation-source recognition, target recognition, and interference suppression.

C. Multi-modal Sensing Technology

Multimodal sensing combines multiple sensory modalities to improve downstream task performance through complementary information. Its machine-learning treatment centers on representation, translation, alignment, fusion, and co-learning under heterogeneous and incomplete data conditions.

  • Multimodal sensing observes entities through multiple modalities, increasing statistical information and improving learning-task accuracy compared with unimodal sensing.
  • Multimodal processing transforms original data into task-suitable latent representations through efficient encoding and fusion of common information.
  • The five key multimodal machine-learning issues are representation, translation, alignment, fusion, and co-learning.
  • Representation: Representation methods must extract useful features from modalities whose data characteristics differ.
  • Translation: Multimodal translation generates one modality from another, using retrieval-based or combination-based approaches.
  • Alignment: Alignment finds relations across modalities but faces missing correspondences, incompatible similarity metrics, and scarce annotated datasets.
  • Fusion: Fusion integrates modalities for prediction through early, late, or hybrid strategies despite synchronization, noise, and complementarity challenges.
  • Co-learning: Co-learning transfers knowledge from resource-rich modalities to resource-poor modalities affected by limited data, pollution, or unreliable labels.

2) Applications:

The section presents applications enabled by multi-modal sensing and mobile crowdsensing, then outlines the data, processing, and task-management techniques needed to deploy crowdsensing efficiently.

  • Multi-modal sensing improves speech, action, event, and emotion applications by combining audio and visual data.
  • Modality translation supports image and video captioning and visual question-answering, while structured representations enable multimedia indexing and retrieval.
  • Mobile crowdsensing uses sensors embedded in ubiquitous devices to collect data for intelligent and people-centric services.
  • Crowdsensing offers abundant data, broad coverage, low cost, scalability, fast response, and support for complex tasks across domains.
  • Efficient crowdsensing deployment requires coordinated data collection, storage, transmission, signal processing, and task-level management.
  • Signal Processing: Crowdsensing signal processing sequentially performs pre-processing, analytics, and post-processing to clean, interpret, and prepare collected data.
  • Task-level Management: Task-level management assigns and schedules sensing tasks, using centralized methods for systematic performance or distributed methods for better privacy preservation.

3) Incentive Mechanisms:

The section discusses privacy and participation challenges in mobile crowdsensing, compares three sensing technologies, and surveys integrated sensing-communication paradigms and designs.

  • 3) Incentive Mechanisms:: Mobile crowdsensing requires collecting phone-sensor data, which can threaten user privacy and increase device power consumption.Noise monitoring illustrates how continuous ambient-sound collection can reduce privacy and battery life.
  • 3) Incentive Mechanisms:: Privacy-preserving approaches include anonymization, encryption, differential privacy, secure sharing, and decentralized processing.
  • 3) Incentive Mechanisms:: Incentive mechanisms balance users’ additional costs against crowdsensing performance gains and seek accurate, reliable participation.
  • E. Discussion: Wireless sensing offers mature processing, long detection distance, and penetration, while multi-modal sensing improves robustness but requires complex AI algorithms.
  • E. Discussion: Mobile crowdsensing provides abundant data, coverage, scalability, low cost, and fast response, but raises data-privacy concerns and participation challenges.
  • A. Overview: ISAC is surveyed through wireless sensing-communication, multi-modal sensing-communication, and mobile crowdsensing-communication paradigms.
  • 1) General Integrated Wireless Sensing and Communication:: Integrated wireless sensing and communication shares hardware and radio resources, while DFRC jointly performs radar and communication using shared hardware and spectrum.
  • 1) General Integrated Wireless Sensing and Communication:: RCC research addresses radar-communication interference through spectrum access, interference management, receiver design, and joint parameter adjustment.

2) Wi-Fi Sensing:

Wi-Fi sensing reuses communication signals for sensing tasks by processing channel-state information, but its gains are limited by non-RF sensor use and rapidly varying channels.

  • 2) Wi-Fi Sensing:: Wi-Fi sensing uses Wi-Fi signals for detection, recognition, and estimation at lower cost than traditional radar sensing.
  • 2) Wi-Fi Sensing:: Channel state information describes target effects on signal amplitude and phase across OFDM subcarriers and supplies inputs for sensing models.
  • 2) Wi-Fi Sensing:: Wi-Fi sensing preprocessing extracts amplitude, phase, temporal, statistical, frequency-domain, and time-frequency features from raw CSI.
  • 2) Wi-Fi Sensing:: Additional preprocessing includes detrending environmental fluctuations, interpolating packet loss, and other application-specific operations.
  • 2) Wi-Fi Sensing:: Integrated wireless sensing and communication cannot fully exploit non-RF sensors, and rapidly changing electromagnetic conditions can reduce its performance-enhancement potential.
  • 2) Wi-Fi Sensing:: Multi-modal sensing combines radio, visible-light, and infrared reflections to acquire physical-environment parameters more precisely and efficiently.

1) Challenges and Techniques:

Integrated sensing, communication, and computation addresses the limitations of partial integration by coordinating the three modules, their resources, and algorithms for complicated tasks. The section surveys multimodal sensing-communication techniques, haptic communications, mobile crowdsensing, and ISCC’s symbiotic design.

  • Integrated Multimodal Sensing and Communication: Integrated multimodal sensing and communication faces heterogeneous data formats and physical properties, plus a lack of aligned multimodal datasets.Existing techniques address dataset construction, CSI estimation, and multimodal sensing enhancement.
  • Haptic Communications: Haptic communications require ultra-high reliability and low latency because sensors may sample above 1 kHz and immersive applications are sensitive to loss and delay.Haptic sensing includes kinesthetic and tactile modalities, creating substantial communication overhead.
  • Integrated Mobile Crowdsensing and Communication: Joint mobile crowdsensing and communication can adapt sensing, collection, and processing to wireless conditions and jointly allocate resources to maximize data utility.Separated designs degrade resource utilization and leave research opportunities.
  • ISCC: ISCC jointly integrates sensing, communication, and computation through shared resources, coordination, and joint algorithm development.It is presented as a comprehensive alternative to partial integration technologies.
  • ISCC: Sensing provides prior knowledge for communication and data for computation, communication enables cooperative sharing, and computation improves sensing and communication algorithms.These roles form the sensing-communication-computation symbiosis.
  • ISCC: Partial integration can mismatch module metrics with overall task goals and ignore coupling while modules compete for time and energy.In online federated learning, convergence depends on sensory samples, sensing and communication SNRs, computation speed, and communication capacity.

B. ISCC’s Functions in 6G

ISCC is described as a 6G technology for improving resource utilization and task performance across network KPIs and usage scenarios. Its benefits include shared waveforms, adaptive coordination, and pipeline design, while challenges remain in optimization, interference, complexity, resource management, unified criteria, and generalization.

  • ISCC Enhances 6G KPIs: ISCC can improve spectrum efficiency by sharing frequency bands among sensing, communication, and computation waveforms or using one waveform for all three functions.AI can also process real-time sensory data for spectrum detection and prediction.
  • ISCC Enhances 6G KPIs: ISCC can estimate or predict device motion to compensate for Doppler effects and enable soft handoff, especially for fast-moving devices.The mechanism uses carefully designed multimodal sensing strategies.
  • ISCC Enhances 6G KPIs: Adaptive time allocation and pipeline overlap among sensing, communication, and computation can reduce end-to-end task-completion latency.An example overlaps new-sample sensing with local-feature extraction and transmission during continuous AI inference.
  • ISCC Enhances 6G KPIs: ISCC can enhance connection density through sensory-data-based estimates of device locations and channel gains, enabling improved beamforming, timing, and subcarrier allocation.The stated setting is a dense network.
  • ISCC Enhances 6G KPIs: ISCC can improve energy efficiency by reusing transmitter signals across functions and adaptively managing device and server energy among the three processes.The benefit follows from coordinating sensing, communication, and computation energy use.
  • ISCC Usage Scenarios: ISCC supports usage scenarios including edge intelligence, metaverse, IoT, smart cities, vehicular networks, and Wi-Fi sensing.For edge intelligence, it targets simultaneous sensing, delivery, and computation under latency and on-device resource constraints.
  • ISCC Challenges: Key ISCC challenges include multi-objective optimization, interference management, high signal-design complexity, and complicated joint resource management.These difficulties arise from heterogeneous metrics, coexistence of waveforms, multifunctional signals, and tight module coupling.
  • ISCC Challenges: ISCC also lacks unified criteria and broad generalization across tasks, networks, and scenarios.Existing schemes are often designed for specific applications or operating environments.

VI. SIGNAL DESIGN FOR ISCC

ISCC signal design coordinates beamforming and waveforms to balance sensing, communication, and computation performance. Designs range from independent signals to a unified triple-functional signal, trading interference reduction against design complexity.

  • Overview: ISCC waveforms balance sensing performance measured by CRB, communication performance measured by SNR, and computation performance measured by AirComp MSE.Beamforming also equalizes mobile-device channels and suppresses noise.
  • Signal-design types: Single-functional designs use independent signals for sensing, communication, and computation but cause severe interference among functionalities.The design is straightforward in functional separation but suffers from cross-functional interference.
  • Signal-design types: Dual-functional designs combine sensing with AirComp using separate signals, reducing interference relative to fully independent functional signals.Antenna arrays are divided between sensing and AirComp, with sensing information extracted from received echoes and data aggregated through analog signal addition.
  • Signal-design types: Triple-functional designs use one signal for sensing, communication, and computation, trading increased design complexity for interference cancellation.The unified signal acts as both a radar probing pulse and a data carrier.

2) Dual-Functional Signal Design:

Dual-functional and triple-functional ISCC designs share sensing, communication, and computation more tightly than separate designs. They use beamforming and resource-management strategies to balance sensing accuracy, AirComp performance, and network-resource competition.

  • 2) Dual-Functional Signal Design:: Dual-functional designs divide each mobile device’s antenna array between sensing and AirComp, using echoes for target extraction and analog wave addition for computation.The sensing target response matrix evaluates sensing performance.
  • 2) Dual-Functional Signal Design:: Triple-functional Air-ISCC uses one signal for sensing, communication, and computation, with mobile-device transmission beamformers and server-side aggregation beamformers.Maximum likelihood estimation determines radar targets, while server beamforming equalizes received signals for AirComp.
  • 2) Dual-Functional Signal Design:: Omnidirectional and directional radar beampatterns are designed jointly with AirComp, but directional beampatterns provide lower AirComp accuracy than omnidirectional designs.The work also examines the balance between optimal radar sensing beampatterns and AirComp accuracy under power constraints.
  • 2) Dual-Functional Signal Design:: Air-ISCC supports target localization, vehicular networks, and federated learning by combining sensing with communication and AirComp.In federated learning, signals sense targets for local training while transmitting local training results for global model updates.
  • 2) Dual-Functional Signal Design:: Joint resource allocation allows sensing, communication, and computation tasks to coexist while competing for shared network resources.One example maximizes a Cobb-Douglas utility combining sensing precision, communication capacity, and computation capability.

2) Coexistence of Sensing and MEC Tasks:

ISCC resource management coordinates sensing, communication, and computation for tasks in which these modules share resources and jointly determine performance. Applications include vehicular perception, inference, crowdsensing, and cooperative sensing systems.

  • 2) Coexistence of Sensing and MEC Tasks:: ISCC allocates network resources among sensing, communication, and computation modules of complex tasks to achieve customized goals.The modules compete for limited radio resources, while task objectives can differ from conventional throughput maximization.
  • 2) Coexistence of Sensing and MEC Tasks:: Online federated learning couples sensing, computation, and communication because devices acquire data, calculate local gradients, upload them, and support global-model updates.Learning performance depends on outcomes across these tightly connected stages.
  • 2) Coexistence of Sensing and MEC Tasks:: Real-time inference accuracy depends on distortion from sensing, quantization, and channel noise while the three processes compete for network resources.An example aggregates local feature vectors through AirComp before downstream inference.
  • 2) Coexistence of Sensing and MEC Tasks:: Task-oriented resource allocation schemes target low latency in autonomous driving and road-traffic applications.Cooperative perception is optimized using spatial-temporal value and latency, while collaborative sensory fusion targets task completion rate.
  • 2) Coexistence of Sensing and MEC Tasks:: Mobile crowdsensing combines environmental sensing, preprocessing, offloading, and server-side processing to maximize data utility under constrained resources.Data utility depends on acquired data quantity and quality, while hostile channels and limited resources can prevent transmission or processing.
  • 2) Coexistence of Sensing and MEC Tasks:: Semantic modeling and hybrid perceptive frames support control of ISCC-based reconfigurable mobile sensor networks.The framework characterizes relations among sensors and multimodal sensory information.

6) ISCC for Robotics:

ISCC extends beyond signal design to robotics, digital-twins networks, and future implementation settings. These systems coordinate sensing, communication, and computation under resource, latency, accuracy, and privacy constraints, while general resource management remains difficult to generalize.

  • 6) ISCC for Robotics:: ISCC-enabled UAVs sense real-time data, compute control commands, and transmit them to robots for task execution.A task-oriented closed-loop optimization approach minimizes total linear quadratic regulator cost under bandwidth, energy, and sensing-distortion constraints.
  • D. Discussion: Resource sharing, coordination, and joint optimization can enhance network-resource utilization and task performance under multiple objectives or customized goals.These benefits arise from jointly handling sensing, communication, and computation rather than treating them as isolated modules.
  • D. Discussion: ISCC resource-management problems are more complex than separated designs because tightly coupled variables must be jointly optimized.Different scenarios and tasks require customized schemes, which limits generalization and raises implementation costs.
  • D. Discussion: ISCC resource management is most suitable for critical tasks with extreme performance requirements or limited network resources; otherwise, separated designs are sufficient.The paper identifies further ISCC resource-management needs for holographic communication, autonomous driving, intelligent agriculture, and smart healthcare.
  • A. ISCC for Digital-Twins Enabled Wireless Networks: Digital-twins systems construct virtual representations of physical objects using distributed sensing, 3D mapping, real-time tracing, and machine learning.They generate simulation data, analyze it, and make decisions based on the generated data.
  • A. ISCC for Digital-Twins Enabled Wireless Networks: An ISCC-based digital-twins network uses heterogeneous sensors to perceive a real wireless network and an edge server to execute computation-intensive algorithms for virtual-representation construction.The proposed platform includes hybrid AI training for digital-twins models.
  • A. ISCC for Digital-Twins Enabled Wireless Networks: Hybrid AI training integrates federated learning, federated split learning, and centralized learning for sensors with different resources, capabilities, and privacy requirements.The training objective minimizes long-term device resource consumption subject to learning accuracy, convergence latency, and privacy constraints.

2) ISCC Based Digital-Twins Network Application:

ISCC-based digital-twin applications combine real-time, task-relevant sensory data with communication and computation across devices, edge servers, and wider networks. These designs use preprocessing, feature sharing, and distributed resources to support fast decisions while reducing communication and protecting privacy.

  • Digital-twin requirements: Digital twins require real-time sensory data that are sufficient, task-relevant, and high quality throughout sensing, transmission, and processing.Relevant design factors include sensing capability, data importance, computation and communication resources, channel gains, battery life, transmit power, and task latency.
  • Resource competition: Multiple decision-making tasks compete for network ISCC resources, motivating opportunistic sensor scheduling that selects sensors producing timely, high-quality data important to the tasks.The scheduling objective is to save network resources while supporting simultaneous tasks.
  • Micro-service architectures: Micro-service architectures divide complex ISCC tasks into fine-grained computing blocks that can be assigned across devices according to their computing powers.Universal blocks can support applications such as criminal tracking, traffic monitoring, and urban management while sharing processing components.
  • Edge-device cooperation: Edge-device cooperation performs sensing and feature extraction on devices while assigning complex DNN computation to edge servers.This avoids raw sensory-data transmission, lowers communication through low-dimensional features, and enables computation offloading.
  • Multi-node cooperation: Multiple ISCC nodes can cooperate through a directed-graph model to improve task performance while saving sensing, computation, and communication resources.The model represents sensing nodes and other participating devices in a computing-power network.
  • SAGIN applications: SAGIN-supported ISCC extends intelligent sensing and computation to remote areas through satellite, aerial, and terrestrial communication systems.In hierarchical designs, UAVs collect and preprocess local sensory data, forward compressed features, and satellites perform global fusion for real-time tasks.

2) Remote Sensing based ISCC Framework in SAGINs:

Remote-sensing ISCC in SAGINs combines multimodal satellite sensing, onboard computation, and satellite-ground communications for real-time applications. Its implementation remains constrained by challenging channels, unresolved theory and hardware requirements, protocol compatibility, and security, privacy, and robustness trade-offs.

  • Remote-sensing framework: Remote-sensing SAGINs combine panchromatic, multispectral, thermal-infrared, and radar data with onboard computation and satellite-ground communication.Processed features are transmitted to ground stations for tasks including precision agriculture and climate-change detection.
  • Fusion design: Late fusion shares only final satellite decisions with the ground station, improving communication efficiency while retaining less information.The passage contrasts this with feature-level processing that saves communication load but sacrifices part of the information.
  • Implementation constraints: Long distances, weather, atmospheric disturbances, electromagnetic interference, obstacles, satellite mobility, and large path loss can degrade SAGIN communication and remote-sensing performance.These conditions produce weak and varying UAV-to-satellite channels.
  • Unresolved issues: ISCC lacks a unified theoretical framework that characterizes the integration of sensing, communication, and computation, limiting guidance for common strategies across tasks and networks.The unresolved gap is especially identified from the information-theoretic perspective.
  • Unresolved issues: New hardware is needed to coordinate the three modules, including pipeline designs that can improve time efficiency but require dedicated sensing-computation hardware.The example performs on-device stochastic gradient descent immediately after each sensory sample.
  • Unresolved issues: ISCC protocols and standardization remain uncharted, with compatibility with existing wireless communication systems identified as an important requirement.Future systems must also support older protocols and devices equipped with legacy hardware.
  • Unresolved issues: Tighter sensing-communication-computation coupling increases information-leakage risks, while privacy and robustness protections can conflict with task performance.The paper identifies leakage through communication loads and computational results, and trade-offs involving privacy, robustness, and performance.
  • Conclusion: The survey presents ISCC as a framework for improving resource utilization and supporting customized goals of complicated tasks, while identifying signal design, resource management, and future implementation issues.It surveys historical partial-integration technologies, ISCC benefits and challenges, signal designs, network resource management, advanced-network implementations, and unresolved research questions.
Loading 2408.08074v3…