Pith. sign in

REVIEW 4 major objections 5 minor 11 references

Interference Mitigation for OFDM-based Integrated Sensing and Communications with Arbitrary Modulation Formats

T0 review · 4 major / 5 minor · reviewed 2026-08-04 · deepseek-v4-flash

Pith's one-line read The paper argues that two-pass successive target cancellation lets OFDM-ISAC use arbitrary, non-constant-modulus alphabets like 64-QAM with sensing accuracy matching constant-modulus QPSK.

desk verdict Clean derivation and a useful cancellation trick; the 'always' claim needs a detection-theoretic caveat. read the letter →

arxiv 2509.07754 v1 pith:Y5MLGRVI submitted 2025-09-09 eess.SP

classification eess.SP
keywords OFDM-ISACintegratedsensingandcommunicationsinterferencemitigationsuccessivetargetcancellationconstantmodulusquadratureamplitudemodulationkurtosisraytracing
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

The paper tackles the sensing half of integrated sensing and communications (ISAC): when an OFDM communication waveform carries data, any constellation that varies in amplitude (like 64-QAM) creates a modulation-dependent interference in the radar range-Doppler map, with power scaling with the transmitted symbol kurtosis and the reflected target power. The authors derive this interference term explicitly, then show that coherent successive target cancellation (CSTC) removes much of it. Their enhancement, ECSTC, adds a second pass that subtracts the synthesized interference of every other target before re-estimating each target, bringing estimation error close to the Cramér-Rao bound in specular multi-target scenes and, in scattering scenes, to the level of constant-modulus QPSK. The practical point is that a 6G base station need not sacrifice high-order, high-data-rate constellations to keep its radar function accurate.

What carries the argument

The central mechanism is the matched-filter range-Doppler map combined with per-target interference reconstruction. The paper shows that for a non-unit-modulus alphabet, |X|^2 = 1 + D, so after matched filtering the channel estimate carries an extra term D times the superposition of all target reflections; this term is the modulation-dependent interference. Each detected target is used to synthesize an interference matrix A_l by re-applying the model, and ECSTC subtracts all matrices except the target of interest before the final parameter estimation. The kurtosis of the alphabet appears as the single scalar controlling interference power.

What would settle it

Run the same ray-tracing scene but order the targets by a CA-CFAR detector's detections from the noisy RDM instead of the true nearest-to-farthest sequence; if the ECSTC MSE at high SNR rises above the QPSK matched-filter baseline, the central claim fails.

Watch

Extended reading notes

Core claim

The paper claims that the sensing loss caused by arbitrary (non-constant-modulus) modulation alphabets in OFDM-ISAC can be almost entirely undone by an enhanced successive-cancellation procedure. For a matched-filtered OFDM radar, the transmitted symbol power fluctuation |X|^2-1 multiplies the channel and acts as additive interference with power |a_l|^2 (kappa-1), where kappa is the fourth moment of the alphabet. CSTC detects and subtracts the strongest targets one by one; ECSTC re-estimates each target after removing all other targets' reconstructed interference matrices. In ray-tracing simulations with 16 extended targets, ECSTC pushes 64-QAM range and velocity MSE down to the QPSK level a

Load-bearing premise

The simulations process the known 16 targets from nearest to farthest, so the true target order and count are not inferred from the noisy range-Doppler map; a deployed system that must detect and order targets first would feed less accurate first-pass estimates into the cancellation loop.

Editorial extensions

If this is right

  • 64-QAM (or any high-order QAM) can be used for ISAC sensing without the usual accuracy penalty, as long as the receiver performs ECSTC.
  • The extra computation is modest: one sum of already-computed interference matrices per target, with no new matched-filter passes.
  • In low-SNR conditions the enhancement is unnecessary; in high-SNR conditions it closes the dominant interference gap.
  • The derived kurtosis scaling gives a design target: constellations with lower fourth moment generate less sensing interference, so shaping schemes can be co-optimized.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • If a real receiver must discover the target set and ordering from the noisy RDM (e.g., via CFAR) rather than using the known nearest-to-farthest order, the first-pass estimates feeding ECSTC would contain detection and ordering errors, so the reported MSE gains could shrink. This is our inference, not stated in the paper.
  • The same cancellation principle might transfer to MIMO-ISAC, where the interference matrices would be per transmit-receive path but the kurtosis-derived interference term remains.
  • The scattering simulation uses a single configurable scattering probability; testing ECSTC on measured or varied scattering surfaces would show how robust the cancellation is to model mismatch.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

4 major / 5 minor

Summary. The paper analyzes the sensing performance of OFDM-based ISAC when the communication symbols are drawn from an arbitrary, possibly non-constant-modulus constellation. The key theoretical contribution is a derivation showing that non-constant-modulus alphabets introduce an additive interference term whose per-target power is |a_l|^2(kappa-1), where kappa is the fourth-order moment of the constellation. The paper then evaluates an existing coherent successive target cancellation (CSTC) scheme and proposes an enhanced version (ECSTC) that performs a second parameter-estimation pass with all first-pass target interference removed. Simulations use a ray-tracing scene with 16 cubes, with and without diffuse scattering, and compare QPSK and 64-QAM operation. The central claim is that ECSTC makes a communication-centric OFDM system with 64-QAM achieve sensing accuracy comparable to that of constant-modulus QPSK.

Significance. If the central claim holds, the result is practically important: it would decouple the sensing accuracy from the constellation alphabet in OFDM-ISAC, allowing higher-order modulation without sensing loss. The theoretical derivation in Sec. III is clean and self-contained, and the ECSTC algorithm is simple and computationally cheap (a second addition of already-computed matrices). The use of Sionna ray tracing with diffuse scattering is a strength relative to idealized point-target papers. However, the load-bearing empirical claims are supported only under a favorable oracle setup: known target processing order and perfect detection. The paper would be significantly strengthened by a detection-theoretic evaluation or at least a robustness study. As it stands, the general claim 'always leads to a performance comparable with the case of constant-modulus alphabets' goes beyond the simulation evidence.

major comments (4)
  1. [Sec. VI-A, Sec. VI-B, Eq. (22)] The central 'always comparable' claim in Sec. VI-B is supported only by simulations in which 'the targets are processed from nearest to farthest' (Sec. VI-A) and in which every target is detected before cancellation. Eq. (22) subtracts A_l' for all l' != l, which presupposes that the first pass found all targets and produced usable estimates. No CA-CFAR or automatic detection/association/ordering is simulated, and no missed-detection, false-alarm, or processing-order sensitivity analysis is provided. If a real system must detect and order targets from the noisy RDM, a missed target's modulation-dependent term (kappa-1)|a_l|^2 from Eq. (19) remains uncancelled. The paper should either replace the 'always' wording with a claim restricted to the simulated oracle scenario or add experiments with automatic detection and randomized/estimated target ordering.
  2. [Sec. V-B, Figs. 3 and 4] The MSE results are averaged over K=1000 independent scene realizations, but no error bars, confidence intervals, or variance of the MSE estimate are reported. Many conclusions rely on the absence of a visible gap between the QPSK and 64-QAM ECSTC curves, and between the ECSTC curves and the CRB. Without interval estimates it is not possible to assess whether these differences are statistically significant. Please report standard errors or bootstrap intervals on the MSE values, at least for the key comparisons in Fig. 4.
  3. [Sec. VI-B, Eqs. (26)-(27)] The CRB used in Figs. 3 and 4 is the AWGN-only, single-point-target bound of Eqs. (26)-(27). The comparison is made, however, under a 16-target channel and, in the scattering cases, under diffuse multi-ray reflections. For scattering, each cube generates a set of sub-rays, so the effective parameter space is different from that of a single specular point target; the displayed CRB may not be a valid lower bound for those cases. This does not invalidate the proposed algorithm, but it weakens the 'close to CRB' argument. Please clarify whether the CRB line is intended only for the specular cases, or provide a bound that accounts for the actual scattering model and multi-target interference.
  4. [Sec. V-C, Sec. VI-B] The scattering parameters P_scatt = 10^-3 and scattering coefficient 0.9 are fixed without sensitivity analysis, and the conclusions in Sec. VI-B are stated as unconditional ('always'). Diffuse scattering strength is a free parameter that can strongly affect the residual interference after cancellation, since ECSTC reconstructs a single reflection per target. A sensitivity study over scattering coefficient/probability, or a clear boundary on the validity regime, is needed before the 'always' claim can be accepted.
minor comments (5)
  1. [Sec. II-B] Typo: 'We note hat the inter-carrier interference' should read 'We note that the inter-carrier interference'.
  2. [Sec. IV-A, Eq. (20)] The normalization and dimension of the IFFT/FFT operations in Eq. (20) are not defined. Please specify whether these are the same N- and M-point transforms used in Eq. (10), and what normalization convention is applied.
  3. [Fig. 4] The legend entries for the scattering curves are difficult to parse (e.g., 'QPSK scattering 64-QAM MF with scattering 64-QAM ECSTC with scattering'). Please use a clear table or distinct line styles with a formal legend.
  4. [Sec. VI-B] Minor typo: 'The remainder of the back scattered still causes noticeable interference' should read 'The remainder of the backscattered signal still causes noticeable interference'.
  5. [Sec. III, after Eq. (19)] The statement that the D-term 'behaves as additional Gaussian interference in the RDM' is attributed to reference [5], which is a self-citation. Since this property is used to interpret the residual after cancellation, a short numerical verification or a direct derivation would make the paper more self-contained.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: ECSTC is validated by independent ray-tracing simulation, the kurtosis derivation is self-contained, and the self-cited [5] is interpretive rather than load-bearing.

full rationale

The paper's central claim—that ECSTC restores constant-modulus sensing performance for 64-QAM—is established by Monte Carlo ray-tracing simulations (Sec. VI) in which the MSE between true and estimated target parameters is measured. No model parameter is fitted to the reported MSE curves, and the cancellation algorithm in Eqs. (20)-(22) follows from the system model rather than from a self-citation. The Sec. III derivation of modulation-dependent interference power is definitional: with unit-power, zero-mean alphabets, Eq. (19) gives P_I,ell = |a_ell|^2 (kappa - 1), where kappa = E|X|^4. This is a variance computation, not a circular prediction. The self-citation to [5] for the statement that the D-term interference 'behaves as additional Gaussian interference in the RDM' (Sec. III) is not load-bearing: ECSTC's reconstruction of A_l via |X_{n,m}|^2 in Eq. (20) does not require Gaussianity, and the simulation serves as an external benchmark. The paper's 'always comparable' claim in Sec. VI-B is broader than what the simulation actually tests, since Sec. VI-A states 'The targets are processed from nearest to farthest' and no automatic detection/ordering or missed-detection analysis is provided; however, that is an evidentiary gap or overclaim, not a circular reduction. No equation is equivalent to its inputs by construction, no fitted parameter is renamed as a prediction, and no author-imported uniqueness theorem forces the result.

Assumptions & free parameters 3 free parameters · 6 assumptions · 0 invented entities

The central claim rests primarily on the Sec. III derivation (self-contained) and on the Sionna simulation. The hand-picked scattering parameters and the assumed Gaussian interference character are the main non-derived inputs. No new physical entities are introduced.

free parameters (3)
  • Scattering probability P_scatt = 1e-3
    Hand-chosen for Sionna RT; controls diffuse reflection intensity and directly shapes the scattering simulation results used to support the claim.
  • Scattering coefficient = 0.9
    Hand-chosen; sets 90% of incident energy reflected diffusely, a key scenario assumption for the scattering curves.
  • Target scene geometry = 16 cubes, 8 m edge, d in [45,145] m, v in [-50,50] m/s
    Hand-chosen scenario; the claim 'always comparable' is only demonstrated for this geometry and no sensitivity analysis is shown.
assumptions (6)
  • domain assumption No ISI: all reflection delays are shorter than the cyclic prefix
    Stated in Sec. II-B; without it, the receive matrix model (6) and the cancellation equations (20)-(22) do not hold.
  • domain assumption Negligible ICI: subcarrier spacing is at least 10x the Doppler spread
    Invoked in Sec. II-B after Eq. (6) citing [6]; ensures Doppler appears as a per-symbol phase only.
  • standard math Matched filter is unbiased and noise-preserving for unit-power alphabets
    Used in Sec. II-C and III to justify Eq. (9); a standard statistical property of the linear matched filter.
  • domain assumption The D-term interference behaves as additional Gaussian noise in the RDM
    Assumed in Sec. III after Eq. (19), citing the self-authored [5]; not re-derived here. It motivates the interference-power interpretation but is not required for the cancellation logic.
  • standard math CRB formulas (26)-(27) are valid for the simulated per-target estimation
    Taken from [8] as the reference bound; the paper applies an 'average SNR per target' which may not correspond to the true per-target CRB when target amplitudes vary.
  • ad hoc to paper Sionna RT scattering parameters approximate real diffuse reflections
    The scene in Sec. V-C uses P_scatt=1e-3 and coefficient 0.9 to model rough surfaces; no comparison to measured channels is given.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Interference Mitigation for OFDM-based Integrated Sensing and Communications with Arbitrary Modulation Formats." pith.science (2026). https://pith.science/paper/Y5MLGRVI

@misc{pith2026250907754,
  author       = {Pith},
  title        = {Pith review of: Interference Mitigation for OFDM-based Integrated Sensing and Communications with Arbitrary Modulation Formats},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/Y5MLGRVI}},
  note         = {Machine review of arXiv:2509.07754}
}
read the original abstract

Integrated sensing and communication will be a key feature of future mobile networks, enabling highly efficient systems and numerous new applications by leveraging communication signals for sensing. In this paper, we analyze the impact of arbitrary modulation alphabets on the sensing performance of communication-centric OFDM systems as expected in the next-generation 6G networks. We evaluate existing interference mitigation techniques, such as coherent successive target cancellation, and propose an enhanced version of this algorithm. A systematic performance evaluation in multi-target scenarios, including the effects of scattering, demonstrates that our proposed interference mitigation methods achieve performance comparable to sensing-optimal constant modulus signals while utilizing higher order constellations for more efficient communications.

Figures

Figures reproduced from arXiv: 2509.07754 by the authors.

Figure 1
Figure 1. Block diagram of the CSTC algorithm and our proposed enhancement (blue) [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. Simulation scene using Sionna RT [10]. The distance and velocity of [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. MSE per target at SNRY = 0 dB. B. Figures of Merit In this paper, we evaluate the mean squared error (MSE) be￾tween the true and estimated target distance ˆdℓ and velocity vˆℓ MSEd = 1 K X K i=1 [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗
Figures from the paper (1 more)
Figure 4
Figure 4. Figure 4: MSE between the true target distance dℓ and the estimate dˆ ℓ [PITH_FULL_IMAGE:figures/full_fig_p005_4.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

11 extracted references · 9 canonical work pages

  1. [5]

    Geiger, F

    B. Geiger, F. Liu, S. Lu, A. Rode, and L. Schmalen, ``Joint optimization of geometric and probabilistic constellation shaping for OFDM-ISAC systems,'' in Proc. IEEE International Symposium on Joint Communications and Sensing (JC&S), Oulu, Finland, Jan. 2025

  2. [1]

    Shatov et al., ``Joint radar and communications: Architectures, use cases, aspects of radio access, signal processing, and hardware,'' IEEE Access, vol

    V. Shatov et al., ``Joint radar and communications: Architectures, use cases, aspects of radio access, signal processing, and hardware,'' IEEE Access, vol. 12, pp. 47\,888--47\,914, 2024

  3. [2]

    T. Wild, A. Grudnitsky, S. Mandelli, M. Henninger, J. Guan, and F. Schaich, `` 6G integrated sensing and communication: From vision to realization,'' in Proc. European Radar Conference ( EuRAD ) , Sep. 2023, pp. 355--358

  4. [3]

    Mandelli, M

    S. Mandelli, M. Henninger, M. Bauhofer, and T. Wild, ``Survey on integrated sensing and communication performance modeling and use cases feasibility,'' in Proc. International Conference on 6G Networking (6GNet), May 2023

  5. [4]

    Xiong, F

    Y. Xiong, F. Liu, Y. Cui, W. Yuan, T. X. Han, and G. Caire, ``On the fundamental tradeoff of integrated sensing and communications under Gaussian channels,'' IEEE Trans. Inf. Theory , vol. 69, no. 9, pp. 5723--5751, 2023

  6. [6]

    B. Nuss, J. Mayer, and T. Zwick, ``Limitations of MIMO and multi-user access for OFDM radar in automotive applications,'' in Proc. IEEE MTT-S International Conference on Microwaves for Intelligent Mobility (ICMIM), 2018, pp. 1--4

  7. [7]

    Geiger, S

    B. Geiger, S. Mandelli, M. Henninger, D. G. Gaviria, C. Muth, and L. Schmalen, ``Integrated long-range sensing and communications in multi target scenarios using CP-OFDM ,'' in Proc. Intl. ITG Conf. on Systems, Communications, and Coding (SCC), Karlsruhe, Germany, Mar. 2025

  8. [8]

    Braun, C

    M. Braun, C. Sturm, and F. K. Jondral, ``Maximum likelihood speed and distance estimation for OFDM radar,'' in Proc. IEEE Radar Conference, 2010, pp. 256--261

Show all 11 references
  1. [9]

    M. A. Richards et al., Fundamentals of Radar Signal Processing . 1em plus 0.5em minus 0.4em Mcgraw-Hill New York , 2005, vol. 1

  2. [10]

    Hoydis et al

    J. Hoydis et al. (2023) Sionna: An open-source library for next-generation physical layer research. [Online]. Available: https://arxiv.org/abs/2203.11854

  3. [11]

    11em plus .33em minus .07em 4000 4000 100 4000 4000 500 `\.=1000 = #1 \@IEEEnotcompsoconly \@IEEEcompsoconly #1 * [1] 0pt [0pt][0pt] #1 * [1] 0pt [0pt][0pt] #1 * \| ** #1 \@IEEEauthorblockNstyle \@IEEEcompsocnotconfonly \@IEEEauthorblockAstyle \@IEEEcompsocnotconfonly \@IEEEco...

Pith tools

Reviewed August 4, 2026 · model on record in the stance chip above.