PaperlayerView exact revision on arXiv

Source-linked AI summary

A Tutorial on MIMO-OFDM ISAC: From Far-Field to Near-Field

Qianglong Dai, Yong Zeng, Huizhi Wang, Changsheng You, Chao Zhou, Hongqiang Cheng, Xiaoli Xu, Shi Jin, A. Lee Swindlehurst, Yonina C. Eldar, Robert Schober, Rui Zhang, Xiaohu You

arXiv:2504.19091v2eess.SP

TL;DR

Near-field sensing presents challenges for methods based on far-field assumptions. This tutorial develops a unified MIMO-OFDM ISAC model and surveys estimation frameworks and algorithms across far- and near-field scenarios. It offers a guide to MIMO-OFDM ISAC and identifies open problems for future investigation.

  • Problem

    Directly applying far-field-based algorithms to near-field sensing presents challenges.

    Source: §I · p. 4

  • Method

    The paper presents a unified MIMO-OFDM ISAC model, four parameter-estimation frameworks, and sensing algorithms for far- and near-field scenarios.

    Source: §I · pp. 3–4

  • Results

    The tutorial presents modeling and enabling signal-processing algorithms under both far-field and near-field channel conditions.

    Source: §VII · p. 33

  • Takeaways & Limitations

    The tutorial aims to guide researchers toward integrating MIMO-OFDM ISAC into future 6G networks and discusses open problems for further investigation.

    Source: §VII · p. 33

Abstract

from arXiv · show

Integrated sensing and communication (ISAC) is one of the key usage scenarios for future sixth-generation (6G) mobile communication networks, where communication and sensing (C&S) services are simultaneously provided through shared wireless spectrum, signal processing modules, hardware, and network infrastructure. Such an integration is strengthened by the technology trends in 6G, such as denser network nodes, larger antenna arrays, wider bandwidths, higher frequency bands, and more efficient utilization of spectrum and hardware resources, which incentivize and empower enhanced sensing capabilities. As the dominant waveform used in contemporary communication systems, orthogonal frequency division multiplexing (OFDM) is still expected to be a very competitive technology for 6G, rendering it necessary to thoroughly investigate the potential and challenges of OFDM ISAC. Thus, this paper aims to provide a comprehensive tutorial overview of ISAC systems enabled by large-scale multi-input multi-output (MIMO) and OFDM technologies and to discuss their fundamental principles, advantages, and enabling signal processing methods. To this end, a unified MIMO-OFDM ISAC system model is first introduced, followed by four frameworks for estimating parameters across the spatial, delay, and Doppler domains, including parallel one-domain, sequential one-domain, joint two-domain, and joint three-domain parameter estimation. Next, sensing algorithms and performance analyses are presented in detail for far-field scenarios where uniform plane wave (UPW) propagation is valid, followed by their extensions to near-field scenarios where uniform spherical wave (USW) characteristics need to be considered. Finally, this paper points out open challenges and outlines promising avenues for future research on MIMO-OFDM ISAC.

I. INTRODUCTION

The introduction motivates communication-centric MIMO-OFDM ISAC as a way to share wireless resources and use communication infrastructure for sensing. It frames the tutorial around receiver signal processing, from unified modeling and parameter-estimation frameworks to far-field and near-field algorithms.

I. INTRODUCTION

  • Communication-centric ISAC can reuse deployed wireless networks for pervasive sensing and improve spectrum, energy, and cost efficiency.The introduction contrasts this approach with separate C&S systems and standalone radar deployment.Source: §I · p. 2
  • OFDM supports delay and Doppler estimation through subcarrier- and symbol-domain structure, with high resolution for large bandwidths or long coherent processing intervals.The phase variations induced by delay and Doppler are independent across these domains, allowing decoupled estimation.Source: §I · p. 2
  • Existing ISAC overviews focus on generic ISAC, while further receiver signal processing work for MIMO-OFDM and near-field scenarios motivates this tutorial.The paper focuses on receiver methods extending from far-field to near-field settings.Source: §I · p. 3
  • The tutorial presents a unified MIMO-OFDM ISAC model, four parameter-estimation frameworks, far-field algorithms, and extensions for near-field sensing.The frameworks span parallel one-domain, sequential one-domain, joint two-domain, and joint three-domain estimation.Source: §I · pp. 3–4

1) Primary ISAC Tasks and KPIs:

  • Primary ISAC KPIs cover communication efficiency and reliability, sensing detection and estimation, target resolution, and object-imaging quality.For unbiased estimators, the Cramér-Rao bound characterizes a lower bound on mean squared error.Source: 1) Primary ISAC Tasks and KPIs · p. 5

2) ISAC Transmitter:

  • ISAC transmitter design spans waveform, resource, antenna, and beam choices; orthogonal C&S resources can reduce interference but usually sacrifice spectral and energy efficiency.The introduction describes fully unified waveforms as more favorable for maximizing integration gains.Source: 2) ISAC Transmitter · p. 5
  • Larger apertures and higher frequencies offer spatial and delay-resolution gains, while narrow beams increase beam-management challenges.The introduction also identifies hybrid beamforming as an approach studied to address complexity and cost of fully digital MIMO.Source: 2) ISAC Transmitter · p. 5

3) ISAC Channel:

The section distinguishes communication and sensing channel models, receiver-processing demands, and waveform trade-offs, then develops a unified ISAC model across six modes. It also motivates near-field modeling as array apertures and operating frequencies increase.

3) ISAC Channel:

  • Communication channels model transmission quality stochastically, whereas sensing channels model target scattering deterministically and generally require target LoS paths.Communication transmitters and receivers are typically spatially separated; sensing transmitters and receivers may be collocated or separated.Source: 3) ISAC Channel · p. 5
  • As arrays grow and mmWave/THz use expands, the near-field region may reach hundreds of meters, making USW modeling relevant to phase and amplitude variation.USW models allow joint angle-and-distance estimation from spatial-domain processing alone, reducing requirements for high-precision synchronization.Source: 3) ISAC Channel · p. 6
  • Higher operating frequencies increase blockage susceptibility; IRSs, passive reflectors, and channel knowledge maps are among proposed ways to address it.The cited approaches establish virtual wireless links or use local-environment knowledge for environment-aware C&S.Source: 3) ISAC Channel · p. 6

4) ISAC Receiver:

  • Super-resolution methods can provide highly accurate parameter estimates, but sensitivity to array-modeling errors and implementation cost matter for large arrays.The section also describes receiver demands: monostatic sensing faces self-interference, while bistatic sensing requires clock-bias compensation and symbol demodulation.Source: 4) ISAC Receiver · p. 6

B. Potential ISAC Waveforms and OFDM ISAC

  • OFDM offers an inherent sensing capability through its thumbtack-shaped ambiguity function, but high PAPR and Doppler sensitivity can raise hardware cost and degrade high-mobility C&S.Emerging multi-carrier waveforms address some OFDM limitations, but their additional transforms increase architectural and computational complexity.Source: B. Potential ISAC Waveforms and OFDM… · pp. 6–7
  • OFDM remains a likely waveform for future ISAC, while XL-MIMO, wide time-frequency bandwidths, joint three-domain estimation, and near-field estimation raise complexity.The paper presents unified models, three-domain parameter-estimation methods, performance and complexity analyses, and extensions from far-field to near-field scenarios.Source: B. Potential ISAC Waveforms and OFDM… · p. 7

III. FUNDAMENTAL ISAC MODELS

  • The six 3GPP modes divide into BS-side and UE-side sensing, with each side supporting monostatic, transmitter-to-receiver bistatic, and other-node bistatic arrangements.The unified model describes an ISAC transmitter, communication receiver, and sensing receiver, and treats the six modes as special cases.Source: §III · p. 8, A. Unified ISAC Model · p. 8

A. Unified ISAC Model

  • The unified model reduces to three BS-side modes: BS monostatic uses collocated transmitter and sensing receiver, while the two bistatic modes place the transmitter at a UE or another BS.BS monostatic operation uses separate antenna sets and full duplex; UE-to-BS bistatic synchronization is needed for distinct clocks, while BS-to-BS synchronization can use a wired connection.Source: A. Unified ISAC Model · p. 9

B. MIMO-OFDM ISAC Signal Processing and Modeling

The signal model organizes MIMO-OFDM sensing observations across antenna, subcarrier, and symbol dimensions, with operating assumptions and synchronization needs depending on the sensing mode. Four estimation frameworks trade computational complexity against parameter-association requirements and accuracy.

B. MIMO-OFDM ISAC Signal Processing and Modeling

  • Set the cyclic-prefix duration to at least max{τcom, τsen}, covering both the communication channel delay spread and the sensing distance needed to avoid ISI.The sensing bound differs because the sensing receiver begins receiving when the transmitter starts transmitting, so nearby targets are not missed.Source: B. MIMO-OFDM ISAC Signal Processing… · p. 10
  • Transmit beamforming depends on user resource allocation, and each resource element carries an information symbol only when allocated to a user.The transmitted beamforming vector and symbol are indexed by subcarrier and OFDM symbol.Source: B. MIMO-OFDM ISAC Signal Processing… · p. 10
  • The frequency-domain observations provide Mr × N × P values organized as a tensor across antenna, subcarrier, and OFDM-symbol domains.The model assumes subcarrier spacing much larger than target Doppler to avoid ICI, and uses an approximation when |υk|/∆f ≪ 1.Source: B. MIMO-OFDM ISAC Signal Processing… · pp. 10–11
  • Monostatic sensing avoids synchronization and symbol-demodulation requirements but needs self-interference cancellation; bistatic sensing requires receiver synchronization and demodulation.Fine synchronization after FFT demodulation can further reduce residual timing and frequency offsets when higher accuracy is needed.Source: B. MIMO-OFDM ISAC Signal Processing… · p. 11
  • The four frameworks range from parallel one-domain estimation, which is low-complexity but risks grouping errors, to joint three-domain estimation, which offers highest accuracy and complexity.Sequential one-domain processing avoids parameter grouping but costs more than parallel processing; joint two-domain processing can improve accuracy at much greater complexity than decoupled methods.Source: B. MIMO-OFDM ISAC Signal Processing… · p. 12
  • The transmit and receive steering vectors depend on target location, which can involve both angle and target distance.The paper gives separate array-response expressions for far-field and near-field scenarios.Source: B. MIMO-OFDM ISAC Signal Processing… · p. 12

C. Array Response Vector: From Far-Field to Near-Field

The array response shifts from a far-field plane-wave model to a near-field spherical-wave model as target distance and array aperture make curvature relevant. The section also describes a monostatic MIMO-OFDM prototype and its hardware.

C. Array Response Vector: From Far-Field to Near-Field

  • With D the array aperture [51],,, r_Rayleigh = 2D^2/λ marks the boundary: for r_k > r_Rayleigh, the UPW far-field model uses first-order Taylor simplification; for r_k < r_Rayleigh, sensing uses the more general expression.Source: C. Array Response Vector: From… · p. 12
  • A single far-field snapshot across antennas identifies target angle, whereas a near-field snapshot can identify both angle and distance.The resulting quadratic phase model is the Fresnel approximation.Source: C. Array Response Vector: From… · pp. 12–13

D. Prototype Design and Hardware Requirements

1) Monostatic MIMO-OFDM ISAC Prototype:

  • The monostatic prototype serves vehicle communications while using OFDM echoes to sense vehicles and obstacles for avoidance services.The base station uses separate phased arrays for transmission and sensing reception; the vehicle uses a phased array for communication reception.Source: 1) Monostatic MIMO-OFDM ISAC Prototype · p. 13

2) Hardware Requirements and Key Parameters:

The section surveys prototype hardware choices and lessons from ISAC development, then introduces the far-field MIMO-OFDM sensing model and decoupled estimation. It explains parallel one-domain estimation and the subsequent need to associate domain-wise estimates with targets.

2) Hardware Requirements and Key Parameters:

  • Servers, software, and RF development kits shape prototype capability, while cloud computing adds upload-and-retrieval latency and FPGA processing raises implementation workload.USRP bandwidth limits communication rate and delay resolution; higher-frequency operation requires additional converters, and larger arrays may require synchronized multiple USRPs.Source: 2) Hardware Requirements and Key… · pp. 13–14
  • Digital MIMO arrays support precise digital beamforming but require an RF chain per antenna; phased arrays reduce RF-chain needs, while movable and switched arrays trade time-division operation for low-cost large apertures.Phased arrays use phase shifters and can have degraded C&S performance; movable and switched arrays suit low-mobility settings and have limited Doppler resolution and beamforming gain.Source: 2) Hardware Requirements and Key… · p. 14
  • Higher-frequency antennas and arrays are considerably more expensive, motivating work on lower-cost, lower-power designs.The text identifies holographic radar and reconfigurable antenna arrays as research directions.Source: 2) Hardware Requirements and Key… · p. 14

E. Lessons Learned

  • A unified model represents six ISAC modes, but their channel, coverage, and target-scattering characteristics differ; collaboration across modes can improve detection and parameter estimation.BS-side sensing is more likely to involve near-field targets and spherical-wave processing than UE-side sensing.Source: E. Lessons Learned · p. 14
  • Existing prototype efforts largely address far-field cases; near-field prototype design and experimental validation for massive or XL-MIMO remain open.The surveyed progress includes monostatic, bistatic, and cell-free multistatic prototypes.Source: E. Lessons Learned · p. 15

A. Far-field Sensing Signal Model

  • Far-field sensing estimates AoA, propagation delay, and Doppler; periodogram methods are simple but resolution-limited, while subspace methods offer super-resolution at higher computational cost.Compressed-sensing methods can be more accurate under low SNR and sparse channels, but large dimensions challenge real-time use and dense-target or clutter cases can degrade estimates.Source: A. Far-field Sensing Signal Model · p. 15

B. Decoupled Parameter Estimation

1) Parallel One-domain Parameter Estimation:

2) Sequential One-domain Parameter Estimation:

Sequential one-domain estimation processes target parameters in an order chosen by domain resolution, using beamforming between stages to isolate targets. This directly associates the estimated angle, delay, and Doppler parameters.

2) Sequential One-domain Parameter Estimation:

  • Sequential estimation begins in the highest-resolution domain, then beamforms to extract targets before estimating the remaining parameters.For a many-element array with narrow bandwidth and short CPI, the described order is AoA, delay, then Doppler.Source: 2) Sequential One-domain Parameter… · p. 16
  • After estimating AoAs, beamforming separates spatial signals before delay estimation; delay-domain beamforming can then distinguish targets at the same angle but with different delays.The approximation assumes sufficiently separated AoA clusters so residual signals from other angles can be treated as noise.Source: 2) Sequential One-domain Parameter… · pp. 16–17
  • The resulting target-specific signals support Doppler estimation after angle and delay have been estimated.The procedure associates the three parameters through its sequence rather than a separate parameter-grouping step.Source: 2) Sequential One-domain Parameter… · p. 17

3) One-dimensional Parameter Estimation Algorithms:

The section presents periodogram, MUSIC, ESPRIT, and OMP as one-dimensional estimation options with different resolution, complexity, and target-separation tradeoffs. In the close-target example, MUSIC and OMP resolve targets beyond the periodogram's array-resolution limit, though OMP accuracy worsens with closely spaced steering vectors.

3) One-dimensional Parameter Estimation Algorithms:

  • Delay and Doppler resolution are inversely proportional to system bandwidth and CPI, respectively; angular resolution depends on target angle and improves with larger effective aperture.The section relates the angular-resolution behavior to beamwidth and the array's Rayleigh limit.Source: 3) One-dimensional Parameter… · pp. 17–18
  • The periodogram produces higher spectral peaks at ground-truth AoAs, but strong-target sidelobes may obscure weaker echoes.Its peak narrows as target angle approaches 0°, making the target easier to distinguish in the described example.Source: 3) One-dimensional Parameter… · p. 18
  • MUSIC estimates parameters using noise- and signal-subspace orthogonality, whereas ESPRIT uses rotational invariance between overlapping subarrays to recover AoAs.MUSIC's spectrum peaks correspond to target AoAs; ESPRIT obtains angles from the estimated rotation operator.Source: 3) One-dimensional Parameter… · pp. 18–19
  • MUSIC gives sharper peaks and essentially no sidelobes compared with the periodogram, improving weak-target detection, but requires the number of AoAs to be known.In the described simulation, both methods accurately estimate all three targets.Source: 3) One-dimensional Parameter… · p. 19
  • For targets spaced below the 7.1° resolution limit of a 16-element ULA, periodogram peaks blend, while MUSIC and OMP resolve all three; OMP accuracy deteriorates with steering-vector correlation.OMP estimation also degrades for dense target distributions, although its separate target estimates prevent weak targets from being masked by strong ones.Source: Figure 12 · p. 19, 3) One-dimensional Parameter… · p. 20
  • OMP approximates sparse recovery by repeatedly selecting dictionary atoms most correlated with the residual, updating the residual by least squares, and estimating target angles from the selected grid.The optimization formulation uses an ℓ2,0 sparsity term and is NP-hard, motivating greedy approximation.Source: 3) One-dimensional Parameter… · p. 19

1) Joint Two-domain Parameter Estimation Model:

Joint two-domain estimation first separates targets by angle, then estimates delay and Doppler together for targets at each angle. Compared with decoupled estimation, this approach can improve accuracy at the cost of greater computational complexity.

1) Joint Two-domain Parameter Estimation Model:

  • Joint two-domain estimation can achieve higher accuracy than decoupled estimation, but requires greater computational complexity.Source: 1) Joint Two-domain Parameter… · p. 20
  • Joint delay-Doppler estimation first estimates target AoAs and beamforms to separate signals by angle, then estimates delay and Doppler simultaneously for each angle.The method covers joint delay-Doppler, AoA-delay, and AoA-Doppler cases; the latter two are handled similarly.Source: 1) Joint Two-domain Parameter… · pp. 20–21

2) 2D Joint Parameter Estimation Algorithms:

The 2D methods jointly estimate delay and Doppler after spatial beamforming. A periodogram uses a separable delay-Doppler spectrum, while 2D-MUSIC applies spatial smoothing to enable super-resolution.

2) 2D Joint Parameter Estimation Algorithms:

  • The 2D-periodogram spectrum factors into independent delay and Doppler powers, so targets can be distinguished when they are resolvable in either domain.An FFT in the subcarrier and symbol domains accelerates the delay-Doppler search.Source: 2) 2D Joint Parameter Estimation… · p. 21

N FFT

  • Negative Doppler estimates require shifting the Doppler spectrum to center its zero-frequency component before locating peaks.Source: N FFT · p. 21
  • 2D-MUSIC vectorizes the data matrix and applies modified spatial-smoothing preprocessing across subcarrier and symbol domains to address the single-snapshot covariance limitation; spectrum peaks identify target delays and Doppler shifts.Source: N FFT · p. 21
  • 2D-MUSIC produces smoother spectra with sharper peaks than the 2D-periodogram, improving delay-Doppler resolution and reducing weak-target masking by strong-target sidelobes.Correctly choosing the signal-subspace dimension remains difficult, particularly when weak targets must be included.Source: N FFT · p. 21

D. Joint Three-domain Parameter Estimation

Joint three-domain estimation uses the received signal to estimate target angle, delay, and Doppler together. The section also compares algorithm complexity and notes the computational burden of joint methods.

D. Joint Three-domain Parameter Estimation

  • The 3D-periodogram uses FFTs across antenna, subcarrier, and symbol domains to reduce computation when searching angle, delay, and Doppler.Source: D. Joint Three-domain Parameter… · p. 22
  • 3D-MUSIC estimates AoA, delay, and Doppler from peaks in a three-domain spectrum, provided the signal-subspace dimension is chosen correctly.Its steering vector combines Doppler, delay, and spatial responses; modified spatial smoothing forms snapshots from the received signal.Source: D. Joint Three-domain Parameter… · p. 22

1) Complexity Analysis:

  • Among 1D methods, periodogram, FFT-MUSIC, and PM-ESPRIT offer the lowest complexity overall, though rankings vary with antenna count and available snapshots.Using partial or a single snapshot can reduce periodogram and OMP complexity at the cost of increased error.Source: 1) Complexity Analysis · p. 23
  • Joint 2D/3D methods require high computation and large-dimensional matrix operations; coarse-fine grid search and signal decimation can reduce complexity and hardware demands.Source: 1) Complexity Analysis · pp. 23–24, Table VIII · p. 24

2) Estimation Accuracy and Resolution Analysis:

Far-field estimation algorithms show accuracy and resolution trade-offs that depend on SNR and sensing domain. Near-field operation instead requires spherical-wave spatial processing, because applying far-field assumptions can spread energy and distort angle estimates.

2) Estimation Accuracy and Resolution Analysis:

  • At about -20 dB, 1D algorithm performance improves sharply; the periodogram performs best overall, while ESPRIT has the lowest estimation accuracy in the tested setup.PM and ESPRIT perform better spatially than in delay and Doppler, where ICI affects estimation; PM-MUSIC can fail for delay at high SNR due to signal coherence.Source: 2) Estimation Accuracy and Resolution… · p. 24
  • For angular resolution, the periodogram resolves targets at about 7° at both tested SNRs; MUSIC and ESPRIT resolve about 4° at lower SNR, while at higher SNR ESPRIT reaches about 0.6° and MUSIC about 1°.The super-resolution methods improve resolution over the periodogram, especially at high SNR.Source: 2) Estimation Accuracy and Resolution… · p. 24, Figure 17 · p. 25

F. Lessons Learned

  • Far-field processing offers trade-offs among complexity, resolution, and accuracy, while different signal SNRs and processing requirements leave C&S with different coverage ranges.Super-resolution methods are not directly applicable to spatial sensing with analog or hybrid beamforming arrays; target detection and FPGA implementation remain open challenges.Source: F. Lessons Learned · p. 25

V. NEAR-FIELD MIMO-OFDM ISAC

  • Near-field processing retains the main signal-processing strategies but must extend spatial estimation from AoA alone to joint angle-range estimation using spherical-wave characteristics.Spherical-wave phase varies nonlinearly across the array; in wideband settings, beam splitting can misalign beams across frequencies, though angular coverage can be controlled for rapid CSI acquisition.Source: §V · p. 25

A. Near-field Sensing Signal Model

  • The near-field array response depends on both target angle and distance, enabling spatial-domain estimation of both independently of delay-based distance estimation.The paper states that this can improve distance accuracy and alleviate synchronization errors associated with traditional delay-based distance estimation.Source: A. Near-field Sensing Signal Model · p. 26

B. Directly Applying Far-field Algorithms to Near-field

  • Applying a far-field periodogram in the near field gives an accurate AoA for the 150 m target but multiple peaks around the 30 m target's true AoA.The near-field/far-field model mismatch produces energy spread, motivating dedicated near-field localization algorithms.Source: B. Directly Applying Far-field… · p. 26, Figure 20 · p. 26

C. Extending Far-field Algorithms to Near-field

1) Near-field Beam Focusing:

2) Near-field MUSIC:

The tutorial extends near-field sensing methods and explains why some far-field algorithms fail under spherical-wave propagation. It presents approaches that separate angle and distance estimation, with simulations highlighting their differing accuracy and complexity.

2) Near-field MUSIC:

  • Near-field MUSIC forms a two-dimensional spectrum from the noise subspace obtained by eigendecomposing the covariance matrix.Source: 2) Near-field MUSIC · p. 26

3) Near-field ESPRIT and Root-MUSIC:

4) Simulations and discussions:

D. Low-complexity Near-Field Estimation Algorithms

1) Second-order cumulant (SoC):

  • SoC selects antenna-pair offsets to eliminate the distance-dependent phase and decouple angle and distance, but SoC-AoA has reduced spatial resolution.Source: 1) Second-order cumulant (SoC) · p. 27

2) Reduced dimension (RD) and reduced rank (RR) algorithms:

3) FFT-enhanced low-complexity and super-resolution algorithm:

The tutorial describes near-field methods that lower the cost of angle-distance estimation through search-region pruning or array symmetry. These methods retain a two-dimensional MUSIC search in selected regions or use sequential one-dimensional estimates.

3) FFT-enhanced low-complexity and super-resolution algorithm:

  • The FFT-enhanced method prunes angle and distance regions without targets, then applies 2D MUSIC within the remaining clusters to reduce search complexity while retaining super-resolution.Source: 3) FFT-enhanced low-complexity and… · p. 28

4) Symmetry-based near-field estimation algorithms:

  • Modified MUSIC reconstructs angle-dependent observations from covariance anti-diagonal samples, applies spatial smoothing, and estimates distance with a subsequent one-dimensional search.Source: 4) Symmetry-based near-field… · pp. 28–29
  • Generalized ESPRIT partitions a symmetric array into two subarrays to estimate target angles, then uses a one-dimensional search for distance.Source: 4) Symmetry-based near-field… · p. 29

E. Performance Analysis and Comparison

The analysis compares near-field sensing algorithms in complexity and RMSE, then identifies trade-offs and implementation constraints. Accuracy, complexity, array spacing, target-count knowledge, and parameter coupling all shape algorithm use.

E. Performance Analysis and Comparison

F. Lessons Learned

  • Near-field algorithms trade complexity against accuracy: 2D and RD/RR methods have high RMSEs and higher complexity, while modified MUSIC and SoC are lower-complexity but less accurate.Source: F. Lessons Learned · p. 30
  • Most near-field algorithms require spacing below a quarter wavelength, which can increase mutual coupling and reduce physical aperture and spatial resolution; SoC further halves effective aperture.Source: F. Lessons Learned · p. 30
  • Nearly all discussed algorithms require prior target-count knowledge, and decoupled distance estimates depend strongly on angle accuracy; SoC and RD cannot accurately estimate distance for targets sharing a direction.Source: F. Lessons Learned · pp. 30–31
  • Separate angle and distance estimation may require parameter grouping, although SoC-AoA and RR avoid that step by searching distance for each estimated angle.Source: F. Lessons Learned · p. 31

VI. EXTENDED DISCUSSION FOR MIMO-OFDM ISAC

The discussion identifies system and signal-processing challenges for MIMO-OFDM ISAC and outlines directions for further work. The conclusion presents the tutorial’s modeling and algorithm coverage across far-field and near-field conditions.

A. Wideband XL-MIMO OFDM ISAC

  • Wideband XL-MIMO creates beam and Doppler squint as array-wide delay and Doppler variation become significant, requiring conventional sensing methods to be reformulated.Source: A. Wideband XL-MIMO OFDM ISAC · p. 31

B. Signal Pre-processing for MIMO-OFDM ISAC

C. Resource Allocation and Beam Management

  • Sparse resource allocation can improve sensing resolution but trades off unambiguous range or sidelobes, while NOMA faces error propagation in successive interference cancellation.Source: C. Resource Allocation and Beam… · p. 32

D. Sparse MIMO ISAC

  • Sparse arrays offer large virtual apertures with fewer antennas, but grating lobes create ambiguity and near-field spherical-wave phase variation prevents far-field virtual-array formation.Source: D. Sparse MIMO ISAC · p. 32

E. NLoS Sensing and Localization in ISAC

F. Mutualism between ISAC and CKM

G. ISAC for 6G System-Level Requirements

H. Security and Privacy in MIMO-OFDM ISAC

Loading 2504.19091v2…