Pith. sign in

REVIEW 4 major objections 5 minor 15 references

Integrated Sensing, Computing and Semantic Communication for Vehicular Networks

T0 review · 4 major / 5 minor · reviewed 2026-08-07 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

Pith's one-line read This paper proposes a joint beamforming and semantic-extraction design that lets a roadside unit track vehicles, deliver semantic messages, and deter eavesdroppers with one antenna array.

desk verdict A legitimate but incremental combination of ISAC, semantic communication, and secure beamforming, with a concrete BTI omission in the eavesdropper constraint that the single-user simulation conveniently hides. read the letter →

arxiv 2506.00522 v1 pith:B62Q7PHD submitted 2025-05-31 eess.SP

Yinchao Yang , Zhaohui Yang , Chongwen Huang , Wei Xu , Zhaoyang Zhang , Dusit Niyato , Mohammad Shikh-Bahaei This is my paper
classification eess.SP
keywords integratedsensingandcommunicationsemanticvehicularnetworkstransmitbeamformingphysicallayersecurityextendedKalmanfilterposteriorCramér-RaoboundBernstein-typeinequality
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

This paper argues that the sensing, computing, and security functions of a vehicular network can be designed as one problem rather than three. The roadside unit uses a uniform linear array to track every vehicle with an extended Kalman filter while simultaneously beaming semantic messages to intended vehicles, and the paper's claim is that choosing the beamformers and the semantic extraction ratios together maximizes the worst-case semantic secrecy rate while minimizing the sum of posterior Cramér-Rao bounds on the tracked angles. Since semantic extraction consumes computing power that must come from the same budget as transmission and sensing, the optimization binds all three resources. A Bernstein-type inequality plus alternating optimization solves the non-convex problem, and the reported simulations show a higher average transmission rate and a better semantic secrecy rate than non-semantic or perfect-CSI benchmarks with nearly unchanged tracking accuracy.

What carries the argument

Four objects carry the argument: the semantic transmission rate $S_{k,t} = (\iota/\rho_{k,t}) \log(1+\gamma_{k,t})$ with $\rho_{k,t}$ the semantic extraction ratio; the BLEU-based lower bound (10) on $\rho$; the worst-case semantic secrecy rate $\mathrm{SSR}_{k,t} = \min_{l\in L} [S_{k,t} - S_{l|k,t}]^+$; and the posterior Cramér-Rao bound $\mathrm{PCRB}(\theta_{i,t})$ formed from observation Fisher information plus the EKF's predicted MSE matrix. These are coupled through the objective (24), where the beamformers $\mathbf{W}$ and $\mathbf{R}$ and the ratios $\rho$ are chosen jointly, with the computing power $P_{\mathrm{comp}} = -F \sum \ln \rho$ inside the same power budget as the transmitted waveform. The Bernstein-type inequality converts the two outage probability constraints into convex restrictions, and alternating optimization over beamformers and extraction ratios, followed by Gaussian randomization for rank-one recovery, produces the final solution.

What would settle it

Build a small testbed with a real semantic encoder, a shared knowledge base, and an eavesdropper, then measure the received semantic rate as $\rho$ and SINR vary; if the rate does not follow $(\iota/\rho)\log(1+\gamma)$, or if the semantic secrecy gain disappears once the eavesdropper holds the same knowledge base, the central claim is refuted.

Watch

Extended reading notes

Core claim

The central claim is that semantics can be added to integrated sensing and communication almost for free, as long as the extraction ratio is treated as an optimization variable. The paper models the semantic rate as $S_{k,t} = (\iota/\rho_{k,t}) \log(1+\gamma_{k,t})$, treats the semantic secrecy rate as the worst-case gap between the intended vehicle's rate and an eavesdropper's rate, and uses the PCRB on the vehicle angle as the sensing metric. The joint optimization (24) maximizes $\kappa_1 \min_k \mathrm{SSR}_{k,t}$ minus $\kappa_2 \sum_i \mathrm{PCRB}(\theta_{i,t})$ under a total power-and-computing constraint and rank-one beamforming, with outage probability constraints handled by Bernstein-type inequalities. The reported results: semantic communication brings the average rate to 5.4945 bps/Hz from 3.7862 bps/Hz, the robust design beats a perfect-CSI baseline in semantic secrecy rate, and angle and distance tracking remain accurate throughout, with only a slight degradation when a vehicle is directly in front of the roadside unit.

Load-bearing premise

The load-bearing premise is that the formulas $S_{k,t}=(\iota/\rho_{k,t})\log(1+\gamma_{k,t})$, $P_{\mathrm{comp}}=-F\sum\ln\rho$, and the BLEU-based lower bound (10) accurately describe real knowledge-base semantic systems; the simulation bypasses (10) by setting $\rho_{LB}=0.65$, so a mismatch between these abstractions and actual semantics would undermine the objective.

Editorial extensions

If this is right

  • One roadside array can serve as radar, communication transmitter, and secrecy engine at once: the sensing waveform is part of the transmitted signal, so no extra spectrum is used for tracking.
  • Semantic communication changes the throughput calculation: the reported average rate rises from 3.7862 to 5.4945 bps/Hz when semantic extraction is used, because information rather than raw bits is delivered.
  • Channel-prediction error can be absorbed into the design: the Bernstein-type inequality ensures the semantic secrecy rate constraint holds with high probability, and the resulting robust design outperforms a perfect-CSI benchmark.
  • Security has a geometric limit: when an unintended vehicle is close to the roadside unit, the semantic secrecy rate drops toward zero, so physical proximity can overwhelm the knowledge-base advantage.

Reading between the lines

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

  • An extension left implicit is that the same joint design applies to any ISAC node with a knowledge base—drones, ships, or smart infrastructure—provided the semantic rate model $S=(\iota/\rho)\log(1+\gamma)$ stays valid in those channels.
  • The security margin is conditional on the eavesdropper's knowledge base being no better than the legitimate user's; if an eavesdropper obtains the same KB, the semantic secrecy rate reduces to a physical-layer secrecy problem and the reported gains would shrink.
  • A testable refinement is to replace the fixed simulation value $\rho_{LB}=0.65$ with the actual lower bound (10) evaluated from real BLEU and precision scores, which would verify that the feasible region of the optimization is not an artifact of the shortcut.
  • The rank-one Gaussian randomization step is a heuristic; its gap could be quantified by comparing objective values before and after randomization across many channel realizations.
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 proposes an integrated sensing, computing, and semantic communication (ISCSC) framework for vehicular networks, in which a roadside unit (RSU) tracks vehicles with an extended Kalman filter (EKF), transmits semantic messages to intended vehicles, and uses sensing signals as useful interference or artificial noise. The authors formulate a joint optimization problem that maximizes a probabilistically constrained worst-case semantic secrecy rate while minimizing the sum of posterior Cramér–Rao bounds, subject to a total power budget that includes a semantic-computing power term. They solve the non-convex problem by dropping the rank-one constraint, applying Bernstein-type inequality (BTI) to convert outage constraints into convex restrictions, and using alternating optimization (AO). Numerical results illustrate EKF tracking gains and semantic secrecy rate improvements over a perfect-CSI benchmark, with one unintended and one intended vehicle in the simulated scenario.

Significance. The topic is timely and the system model is coherent: combining ISAC, semantic communication, and physical-layer security is a reasonable extension of recent work. If the optimization were solved exactly as stated, the contribution would be useful to the vehicular-networking community. The paper does not provide machine-checked proofs or reproducible code, and it relies on the authors' prior work [10] for the core semantic rate model and computing-power model. The main strengths are a clear problem statement, a sensible EKF-based tracking backbone, and a concrete algorithmic pipeline. However, the numerical validation is limited to a single intended vehicle, which hides a structural mismatch in the BTI reformulation, and the simulation bypasses the semantic-extraction-ratio lower bound in Eq. (10). The significance is therefore conditional: the framework is promising, but the current evidence does not fully establish the claims made in the abstract.

major comments (4)
  1. [§IV-B, Eq. (27)] The BTI reformulation of the eavesdropper outage constraint omits the intended-user multiuser interference term. In Eq. (11), the SINR Γ_{l|k,t} has denominator h_l^H(Σ_{k'≠k} W_{k',t} + Σ_{i∈L∪K} R_{i,t}) h_l + σ_c^2, but in Eq. (27) the matrix χ_{l|k,t} is defined as Σ_{i∈L∪K} R_{i,t} − (1/\hat{Γ}_{l|k,t}) W_{k,t}, dropping the positive-semidefinite term Σ_{k'≠k} W_{k',t}. Since the dropped term is positive semidefinite, the implemented constraint is stricter than the stated constraint in (25c); consequently Algorithm 1 solves a conservative restriction of problem (24), not the problem itself. The numerical section uses only K=1 intended vehicle, so the omitted term is exactly zero in the reported simulations and the mismatch is invisible. The author should either include the missing term to match (11) or explicitly state and justify that the restriction is intended, and quantify the resulting optimality gap.
  2. [§IV-A/B, Eqs. (24)-(25), Algorithm 1] The transformation from (24a) to (25a) is not an equivalence. The objective in (25a), κ_1(λ−ϱ), replaces the worst-case semantic secrecy rate min_k SSR_{k,t} with a difference of two separate outage-threshold variables, but the constraints (25b) and (25c) only guarantee Pr(S_k ≥ λ) and Pr(S_{l|k} ≤ ϱ) individually; a joint guarantee on S_k − S_{l|k} requires a union bound that is not stated. In addition, Algorithm 1 updates λ and ϱ by fixed increments Δλ and Δϱ in step 6 rather than optimizing them, and the inner-loop stopping criterion in step 7 checks only changes in W and R, not the objective or the λ/ϱ variables. No convergence proof for the AO loop is provided. These issues undermine the claim that the joint optimization problem (24) is solved as stated.
  3. [§III-A, Eq. (10), and §V] The semantic communication model is imported from the authors' prior paper [10] without independent validation, and the numerical section does not instantiate the key bound in Eq. (10). The simulation sets ρ_LB = 0.65 directly, without computing Q_t, w_{g,k,t}, or p_{g,k,t} from any knowledge-base or language-model measurement. Since the semantic secrecy rate in Eq. (8) and the computing power in Eq. (14) both depend on this model, the quantitative claims of throughput and security gains rest on an unvalidated abstraction. The paper should either provide a self-contained justification or experimental calibration of Eqs. (8)-(14), or clearly mark the entire semantic model as an assumption and present a sensitivity analysis with respect to ρ_LB and the model parameters.
  4. [§IV-B, Algorithm 1 step 8] The rank-one constraint (24d) is dropped and then handled by Gaussian randomization, but no theoretical guarantee or empirical validation of the randomization step is provided. No number of randomizations is reported, and no bound is given on the suboptimality of the resulting rank-one solution. Because the final beamformers are obtained from this step, the paper needs at least a description of the randomization procedure, the number of trials, and a comparison with the SDP upper bound, or an explicit statement that the rank-one recovery is heuristic.
minor comments (5)
  1. [§III-A, Eq. (10)] The terms Q_t, w_{g,k,t}, and p_{g,k,t} in Eq. (10) are not defined in this paper, and the reader is directed to [10]; the authors should either define these terms or state clearly that Eq. (10) is a result from [10] and is not derived here.
  2. [§II-A and §II-B] The symbol r_{i,t} is used both for the sensing beamforming vector in Eq. (3) and for the observation variable in the state-space model in Eq. (6). This collision makes the model ambiguous and should be fixed by renaming one of the two quantities.
  3. [§IV-B, Eq. (27)] In the definition of s_{l|k,t}, the notation \bar{h}_{l|k} should be \bar{h}_{l,t}; the eavesdropper channel does not depend on the intended user index k.
  4. [§IV-B, complexity statement] The claimed complexity O(I_1(N + log(I_2))) is not plausible for an algorithm that solves an SDP with matrix variables of size N×N; SDP solution complexity is typically polynomial of degree three or more in N. The complexity expression should be revised or justified.
  5. [§V] The simulation description reports Δλ = Δϱ = 0.1 but does not state the number of Gaussian randomization trials in step 8 of Algorithm 1, nor does it state whether the reported points satisfy the outage constraints empirically; adding these details would improve reproducibility.

Circularity Check

0 steps flagged · score 2.0 of 10

No derivation in this paper reduces to its inputs by construction; the only self-citation is the import of the semantic model from the authors' prior work, which is a modeling input rather than a circular conclusion.

full rationale

The paper's claimed new contribution is the outage-constrained joint optimization (24) over beamformers and semantic extraction ratios, solved by Bernstein-type inequality and alternating optimization. The semantic rate expression S_{k,t}=ι/ρ_{k,t} log(1+γ_{k,t}) in (8), the lower bound (10), and the computing power (14) are taken from the authors' own prior work [10]; they are stated as definitions or imported results, not derived in this paper, and the paper does not claim to predict them from the vehicular model. The optimization objective and constraints are formed from these definitions, so no equation is shown to be equivalent to another by construction, and no fitted parameter is renamed as a prediction. The BTI transformation in (27) omits the inter-user interference term in the eavesdropper constraint, making the implemented problem a conservative restriction of (25c), but conservatism is a correctness issue, not circularity. Likewise, setting ρ_LB=0.65 in the simulations instead of evaluating (10) is a validation gap, not a circular step. The EKF and PCRB components are standard external results. Because the central optimization and algorithm are independent of the cited model's derivation, the paper does not exhibit a circular reduction; the score reflects the minor self-citation of [10] for load-bearing model equations that are not independently re-validated here.

Assumptions & free parameters 6 free parameters · 7 assumptions · 0 invented entities

The optimization rests on the authors' prior semantic model [10], the PCRB/EKF framework of [8], and the BTI machinery of [11]. The solved problem differs from the advertised one because the BTI version of the eavesdropper constraint omits the multi-user interference from (11). Several constants (F, ι) are never specified, and the simulation replaces the theoretical ρ bound with an ad hoc value.

free parameters (6)
  • Trade-off coefficients κ1, κ2 = 0.5 each
    Chosen by hand in Section V; they set the balance between secrecy rate and sensing accuracy in (24a).
  • Computing-power coefficient F = Not specified
    Appears in (14); never assigned a value, so the numerical results are not fully reproducible.
  • Word-to-bit ratio ι = Not specified
    Scales the semantic rate in (8) and thresholds in (27); no value given in Section V.
  • Semantic extraction ratio lower bound ρ_LB = 0.65 (simulation)
    Set directly in Section V instead of evaluating the theoretical bound (10); makes the simulation independent of the claimed derivation.
  • Outage probabilities ϵ1, ϵ2 = 0.01
    Simulation choices defining chance constraints (25b)-(25c).
  • State/measurement noise covariances Q1, Q2 = per vehicle, from [8],[12]
    Imported from prior work; values drive the EKF and PCRB and are not derived here.
assumptions (7)
  • domain assumption Semantic rate formula S=(ι/ρ)log(1+γ) and lower bound (10) are valid
    Adopted from [10] in Section III-A; the whole secrecy objective depends on this model.
  • domain assumption Computing power equals -F Σ ln(ρ_k)
    Adopted from [10] in (14); no independent physical or experimental justification.
  • domain assumption CSI error Δh is zero-mean Gaussian with known covariance Ω
    Needed in Section IV-B to apply BTI; actual beam-tracking error statistics are not measured.
  • standard math BTI gives valid convex restrictions of chance constraints
    Imported from [11]; standard result but applied with an incorrect quadratic form for the eavesdropper constraint (27).
  • domain assumption Posterior FIM equations (17)-(23) are valid for the state model
    Taken from [8] without derivation although the state model (5) differs from [8].
  • ad hoc to paper Rank-one relaxation with Gaussian randomization is near-optimal
    No approximation guarantee is proven; it is standard practice but unchecked here.
  • ad hoc to paper Alternating optimization converges to a useful point
    Convergence is asserted in Algorithm 1 without proof for this non-convex problem.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Integrated Sensing, Computing and Semantic Communication for Vehicular Networks." pith.science (2026). https://pith.science/paper/B62Q7PHD

@misc{pith2026250600522,
  author       = {Pith},
  title        = {Pith review of: Integrated Sensing, Computing and Semantic Communication for Vehicular Networks},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/B62Q7PHD}},
  note         = {Machine review of arXiv:2506.00522}
}
read the original abstract

This paper introduces a novel framework for integrated sensing, computing, and semantic communication (ISCSC) within vehicular networks comprising a roadside unit (RSU) and multiple autonomous vehicles. Both the RSU and the vehicles are equipped with local knowledge bases to facilitate semantic communication. The framework incorporates a secure communication design to ensure that messages intended for specific vehicles are protected against interception. In this model, an extended Kalman filter (EKF) is employed by the RSU to accurately track all vehicles. We formulate a joint optimization problem that balances maximizing the probabilistically constrained semantic secrecy rate for each vehicle while minimizing the sum of the posterior Cram\'er-Rao bound (PCRB), subject to the RSU's computing capabilities. This non-convex optimization problem is addressed using Bernstein-type inequality (BTI) and alternating optimization (AO) techniques. Simulation results validate the effectiveness of the proposed framework, demonstrating its advantages in reliable sensing, high data throughput, and secure communication.

Figures

Figures reproduced from arXiv: 2506.00522 by the authors.

Figure 1
Figure 1. System model of ISCSC in a vehicular network. [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Angle and distance tracking performances. [PITH_FULL_IMAGE:figures/full_fig_p005_2.png] view at source ↗
Figure 4
Figure 4. Sensing performance versus time. is designed to overcome channel uncertainties. The non-convex optimisation problem is tackled by applying the Bernstein￾type inequality and the alternating optimisation methods. Simulation results demonstrate that the proposed framework and algorithm balance the sensing accuracy, communication throughput, and security in vehicular networks. REFERENCES [1] N. Su, F. Liu, Z. Wei, Y.-F.… view at source ↗

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

15 extracted references · 12 canonical work pages

  1. [10]

    Secure design for integrated sensing and semantic communication system,

    Y . Yang, M. Shikh-Bahaei, Z. Yang, C. Huang, W. Xu, and Z. Zhang, “Secure design for integrated sensing and semantic communication system,” in2024 IEEE Wireless Communications and Networking Con- ference (WCNC), pp. 1–7, IEEE, 2024

    work page 2024

  2. [1]

    Secure dual- functional radar-communication transmission: Exploiting interference for resilience against target eavesdropping,

    N. Su, F. Liu, Z. Wei, Y .-F. Liu, and C. Masouros, “Secure dual- functional radar-communication transmission: Exploiting interference for resilience against target eavesdropping,”IEEE Transactions on Wireless Communications, 2022

    work page 2022

  3. [2]

    Bayesian predictive beamforming for vehicular networks: A low-overhead joint radar-communication approach,

    W. Yuan, F. Liu, C. Masouros, J. Yuan, D. W. K. Ng, and N. Gonz ´alez- Prelcic, “Bayesian predictive beamforming for vehicular networks: A low-overhead joint radar-communication approach,”IEEE Transactions on Wireless Communications, vol. 20, no. 3, pp. 1442–1456, 2020

    work page 2020

  4. [3]

    Sensing as a service in 6g perceptive networks: A unified framework for isac resource allocation,

    F. Dong, F. Liu, Y . Cui, W. Wang, K. Han, and Z. Wang, “Sensing as a service in 6g perceptive networks: A unified framework for isac resource allocation,”IEEE Transactions on Wireless Communications, 2022

    work page 2022

  5. [4]

    Physical layer security optimization with cram´er-rao bound metric in isac systems under sensing-specific imper- fect csi model,

    H. Jia, X. Li, and L. Ma, “Physical layer security optimization with cram´er-rao bound metric in isac systems under sensing-specific imper- fect csi model,”IEEE Transactions on V ehicular Technology, 2023

    work page 2023

  6. [5]

    S-ran: Semantic-aware radio access networks,

    Y . Sun, L. Zhang, L. Guo, J. Li, D. Niyato, and Y . Fang, “S-ran: Semantic-aware radio access networks,”IEEE Communications Mag- azine, 2024

    work page 2024

  7. [6]

    Semantic communication-based dynamic resource allocation in d2d vehicular networks,

    J. Su, Z. Liu, Y .-a. Xie, K. Ma, H. Du, J. Kang, and D. Niyato, “Semantic communication-based dynamic resource allocation in d2d vehicular networks,”IEEE Transactions on V ehicular Technology, vol. 72, no. 8, pp. 10784–10796, 2023

    work page 2023

  8. [7]

    xurllc- aware service provisioning in vehicular networks: A semantic commu- nication perspective,

    L. Xia, Y . Sun, D. Niyato, D. Feng, L. Feng, and M. A. Imran, “xurllc- aware service provisioning in vehicular networks: A semantic commu- nication perspective,”IEEE Transactions on Wireless Communications, 2023

    work page 2023

Show all 15 references
  1. [8]

    Radar-assisted predictive beamforming for vehicular links: Communication served by sensing,

    F. Liu, W. Yuan, C. Masouros, and J. Yuan, “Radar-assisted predictive beamforming for vehicular links: Communication served by sensing,” IEEE Transactions on Wireless Communications, vol. 19, no. 11, pp. 7704–7719, 2020

    2020

  2. [9]

    S. M. Kay,Fundamentals of statistical signal processing: estimation theory. Prentice-Hall, Inc., 1993

    1993

  3. [11]

    Outage constrained robust transmit optimization for multiuser miso downlinks: Tractable approximations by conic optimization,

    K.-Y . Wang, A. M.-C. So, T.-H. Chang, W.-K. Ma, and C.-Y . Chi, “Outage constrained robust transmit optimization for multiuser miso downlinks: Tractable approximations by conic optimization,”IEEE Transactions on Signal Processing, vol. 62, no. 21, pp. 5690–5705, 2014

    2014

  4. [12]

    Ve- hicular connectivity on complex trajectories: Roadway-geometry aware isac beam-tracking,

    X. Meng, F. Liu, C. Masouros, W. Yuan, Q. Zhang, and Z. Feng, “Ve- hicular connectivity on complex trajectories: Roadway-geometry aware isac beam-tracking,”IEEE Transactions on Wireless Communications, vol. 22, no. 11, pp. 7408–7423, 2023

    2023

  5. [13]

    Optimal energy-efficient transmit beamforming for multi-user miso downlink,

    O. Tervo, L.-N. Tran, and M. Juntti, “Optimal energy-efficient transmit beamforming for multi-user miso downlink,”IEEE Transactions on Signal Processing, vol. 63, no. 20, pp. 5574–5588, 2015

    2015

  6. [14]

    Particle filtering,

    P. M. Djuric, J. H. Kotecha, J. Zhang, Y . Huang, T. Ghirmai, M. F. Bugallo, and J. Miguez, “Particle filtering,”IEEE signal processing magazine, vol. 20, no. 5, pp. 19–38, 2003

    2003

  7. [15]

    Bayesian filtering: From kalman filters to particle filters, and beyond,

    Z. Chenet al., “Bayesian filtering: From kalman filters to particle filters, and beyond,”Statistics, vol. 182, no. 1, pp. 1–69, 2003

    2003

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.