REVIEW 4 major objections 7 minor 14 references
When Feedback Empowers the Uplink: Integrating Adaptive Coding with Wireless Power Transfer
T0 review · 4 major / 7 minor · reviewed 2026-08-07 · deepseek-v4-flash
Pith's one-line read This paper claims that a single downlink feedback signal can simultaneously guide adaptive uplink coding and charge IoT devices, extending device lifetime by up to 13.4×.
desk verdict Interesting system concept, but the central optimization derivation is wrong, so the quantitative claims are unsupported. read the letter →
The pith
A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.
The reading
What carries the argument
The load-bearing model is the feedback-coding SNR curve of Eq. (4): $\eta_{\ell,\mathrm{dB}} = u_0(\epsilon^*) + 1/(\exp\{u_1(\epsilon^*) + u_2(\epsilon^*)\tilde\eta_{\ell,\mathrm{dB}}\} + u_3(\epsilon^*))$, where $\tilde\eta_{\ell,\mathrm{dB}}$ is the downlink feedback SNR. The curve converts feedback quality into required uplink power, and its saturation plateau is what makes the dual-purpose split worthwhile. The second mechanism is the power-splitting ratio $\rho_\ell$: the device harvests fraction $\rho_\ell$ of received RF power and decodes feedback with fraction $1-\rho_\ell$, which couples the two objectives through the effective feedback SNR. The paper's optimization machinery—Hungarian assignment for subcarriers, then alternating dual updates with KKT-derived closed forms—keeps the coupling tractable.
What would settle it
Recompute the 13.4× and 1.83× lifetime gains using a measured SNR-vs-feedback-SNR curve for a concrete feedback code (for example, a trained Attentioncode or Deepcode) in place of the undisclosed coefficients in Eq. (4); if the measured curve lacks a clear saturation plateau within the simulated $P_{\text{total}}$ range, the energy-harvesting portion of FACET's advantage disappears.
Extended reading notes
Core claim
The paper claims that the informational and energetic roles of feedback can be merged. With a feedback channel code, the critical uplink SNR needed for a target block error rate decreases exponentially with feedback SNR and then saturates at $u_0(\epsilon^*)$; the paper uses this saturation to justify diverting part of the received downlink power, controlled by a splitting ratio $\rho_\ell \in [0,1]$, from decoding to an RF energy harvester. It then solves a min-max problem that minimizes $\max_\ell (P_\ell - \rho_\ell \kappa \tilde P_\ell)$, the worst-case net energy consumption, using Hungarian subcarrier assignment, alternating optimization, and Lagrangian duality, with closed-form updates for each variable block. The reported result is that the optimal policy operates in a coding-dominated regime at low feedback power and a WPT-dominated regime at high feedback power, producing the claimed lifetime gains.
Load-bearing premise
The entire gain rests on the empirical feedback-coding curve in Eq. (4), whose fitted coefficients $u_0,u_1,u_2,u_3$ are not disclosed; if that curve does not accurately describe real feedback codes, the reported lifetime extensions are unsupported.
Editorial extensions
If this is right
- In high-power regimes, surplus feedback power that would otherwise be wasted can be redirected to energy harvesting, lowering net device energy consumption without hurting decoding reliability.
- Devices experience max-min fairness: the worst-case net energy consumption is minimized, so edge devices with poor channels are not starved.
- The closed-form per-block solutions make the resource allocation computationally light enough for practical IoT deployment.
- FACET self-adapts between coding-dominated and WPT-dominated operation as the total feedback power budget varies.
- Under the reported simulation settings, device lifetime extends by 1.83× over feedback-only and 13.4× over WPT-only schemes.
Reading between the lines
- Beyond the paper: because Eq. (4) is code-specific, the same framework should transfer to any feedback code that exhibits saturation; the crossover power where WPT begins to dominate could be precomputed from the code's saturation point and used as a simple switching rule.
- Beyond the paper: if the coefficients $u_0,u_1,u_2,u_3$ were disclosed, the optimization could be reproduced and compared against a fixed-split policy, and a hardware testbed measuring harvested DC power versus decoding error rate would directly test the Pareto optimality claim.
- Beyond the paper: the min-max formulation suggests a natural extension to energy-neutral operation, where the optimization targets zero net consumption rather than minimum worst-case, which the paper does not explicitly consider.
Editorial analysis
A structured set of objections, weighed in public.
Referee Report
Summary. The paper proposes a Feedback-Aided Coding and Energy Transfer (FACET) framework for IoT uplinks, in which downlink feedback signals serve both as real-time coding guidance and as a wireless power source. The authors formulate a min-max optimization problem to minimize the worst-case net energy consumption over transmit power, power-splitting ratios, and subcarrier assignments, then propose a two-stage solution: a Hungarian-based subcarrier assignment followed by alternating optimization with Lagrangian duality and closed-form updates. Simulations report lifetime gains of up to 1.83× over a feedback-only baseline and 13.4× over fixed power-splitting WPT baselines.
Significance. The conceptual idea of exploiting the saturation of feedback-coding gain to repurpose surplus downlink power for energy harvesting is timely and potentially valuable for energy-constrained IoT systems. The problem formulation is reasonable in spirit, and the use of a min-max objective for fairness is appropriate. However, the paper's central algorithmic claims are not supported: the transformed problem is asserted to be convex without justification, the Lagrangian written for the t_l subproblem does not match the stated constraint, and the resulting closed-form solution is invalid for nonnegative dual multipliers. Since the reported lifetime gains are produced by this algorithm, they do not establish the paper's claims. The manuscript also does not disclose the fitted parameters of the empirical coding curve in Eq. (4), which is the quantitative foundation of the energy-saving mechanism, nor does it provide code or validation data. If these issues were fixed, the framework could be a useful contribution, but in its current form the central results are not reproducible or internally consistent.
major comments (4)
- [Section III.C.1] The Lagrangian written for the t_l subproblem does not correspond to constraint (10) or to the original feedback-SNR constraint (9). The manuscript's L(r,t,γ,ν) contains the term ν_l(t_l - (1/exp(u1))[Σ_s U_s,ℓ^(2)(1-ρ_l)δ_s,ℓ ep_s]^{10u2/ln10} + u3 + u0), which is neither constraint (10) nor an equivalent rearrangement of (9). Consequently, the stationary condition leading to t_l* = 10 log10(-10ν_l/(γ_l U_l^(0) ln10)) is not the KKT condition of the stated problem. Moreover, for nonnegative dual multipliers in a minimization problem, the argument of the logarithm is negative, so t_l* is not a real solution. This invalidates the closed-form update, the alternating optimization, and the claimed convergence.
- [Section III.B, Eq. (11)] The claim that the transformed problem (11) is convex is not correct. The objective and the energy constraints contain bilinear products ρ_l ep_s (for example, the term U_l^(0)10^{t_l/10} - ρ_l Σ_s U_s,ℓ^(1) ep_s), whose Hessian with respect to (ρ_l, ep_s) is indefinite. In addition, constraint (10) has a left-hand side ln(1/(t_l - u0) - u3), which is not convex over its domain for positive u3. Thus the transformed problem is not a convex optimization problem, and the later assertion of guaranteed convergence to a local optimum is unsupported.
- [Section II, Eq. (4) and Section IV] Equation (4) is the quantitative foundation for all claimed energy savings, but the coefficients u0, u1, u2, u3 and the fitting procedure from [6] are not disclosed, and no validation error is reported. The lifetime extensions of 1.83× and 13.4× in Section IV are therefore contingent on an unverified empirical curve and are not reproducible from the information given. This issue is load-bearing because the entire coding-gain mechanism and the saturation effect that motivates FACET are encoded in Eq. (4).
- [Section III.B, Eq. (7)] The two-stage decomposition is not shown to preserve the min-max fairness objective. The subcarrier assignment in Eq. (7) maximizes the total weighted channel gain across all devices, which does not generally minimize the worst-case net energy consumption; a device with poor channels may be assigned a subcarrier that worsens its net energy. No argument or proof is provided that this first stage is compatible with the fairness criterion in the original problem (6), so the fairness claim is not established.
minor comments (7)
- [Section III, opening] The sentence "This section, we introduce our FACET framework" should read "In this section, we introduce our FACET framework."
- [Eq. (4)] The placement of parentheses in Eq. (4) is ambiguous: the denominator should be explicitly written as exp{u1 + u2 eη_ℓ,dB} + u3 so that the reader knows u3 is outside the exponential.
- [Figure 2] The caption says "left and right subfigures" but the figure contains three panels (a), (b), and (c); the text's reference to "the right subfigure" for the κ variation is therefore ambiguous.
- [Table I] The entry for FACET's optimal ρ is given as "Dynamic ( ρ)", which is incomplete; please specify the range or the value used in the reported simulation.
- [Section III.C.1] The Lagrangian expression has unbalanced parentheses and inconsistent use of t vs. t_l; please rewrite it carefully so that the reader can verify the KKT derivation.
- [Notation] The symbol η is used both for SNR and as a Lagrange multiplier in Section III.C.2; please use distinct symbols to avoid confusion.
- [Section IV] The total power notation is inconsistent: the text uses Ptotal while the optimization problem in (6b) uses ePtotal; please unify.
Circularity Check
The numerical lifetime gains inherit a same-author fitted SNR curve, but the WPT/power-splitting optimization is an independent contribution.
-
self citation load bearing
[Section II, Eq. (4); Section IV performance evaluation]
"To characterize the power reduction capability of feedback-aided adaptive channel coding, we consider the feedback codes in [6], where the critical uplink SNR to achieve ϵ∗ is characterized as ηℓ,dB = u0(ϵ∗) + 1 exp {u1(ϵ∗) +u2(ϵ∗)eηℓ,dB} + u3(ϵ∗) , (4)"
All quantitative lifetime claims, including the 13.4× and 1.83× gains, are computed from this SNR-reduction curve, which is imported directly from the authors' own prior work [6]. The coefficients u0..u3 are not disclosed or independently validated in this paper, so the numerical predictions are evaluations of a same-author fitted/assumed model rather than first-principles predictions. The FACET optimization itself (the ρ, eps tradeoff) has independent content, so the circularity is partial rather than total.
full rationale
The paper's central algorithmic contribution—formulating and solving the min-max net-energy problem with power splitting ρℓ and feedback-power allocation eps—is not circular: the optimization objective, constraints, and closed-form structure do not assume the paper's own conclusions. The main circularity burden is that the quantitative magnitude of FACET's benefit rests on Eq. (4), a feedback-coding SNR curve taken from a self-citation ([6], Y. Shao, DEEP-IoT) with undisclosed fitted coefficients. The saturation behavior and the resulting 'redirect surplus feedback power to WPT' conclusion are built into that assumed curve, so the reported lifetime extensions are downstream of the self-cited model. However, the comparison against FCA-IoT (feedback-only) shares the same curve, and the WPT-versus-decoding tradeoff is genuinely optimized, so the central claim still contains independent content beyond the imported curve. The Lagrangian/KKT sign issue noted by the skeptic is an internal correctness flaw rather than a circularity, because it concerns whether the stated algorithm solves the stated problem, not whether the conclusion is identical to an input assumption. Overall score 4 reflects load-bearing self-citation without full reduction by construction.
Assumptions & free parameters
free parameters (3)
- u0, u1, u2, u3 (feedback coding curve coefficients) =
not disclosed
- kappa (RF-to-DC energy harvesting efficiency) =
0.15 in Table I
- Simulation scenario parameters (G, Q, M, S, alpha, noise powers) =
not specified
assumptions (3)
- domain assumption Eq. (4) accurately models the critical uplink SNR for real feedback codes as a function of feedback SNR.
- ad hoc to paper The transformed optimization problem (11) is convex in the continuous variables.
- domain assumption Each IoT device can harvest energy from all downlink subcarriers while decoding only its own, using a power-splitting factor rho_l.
Cite this review
Pith. "Pith review of When Feedback Empowers the Uplink: Integrating Adaptive Coding with Wireless Power Transfer." pith.science (2026). https://pith.science/paper/T436YNDG
@misc{pith2026250521951,
author = {Pith},
title = {Pith review of: When Feedback Empowers the Uplink: Integrating Adaptive Coding with Wireless Power Transfer},
year = {2026},
howpublished = {\url{https://pith.science/paper/T436YNDG}},
note = {Machine review of arXiv:2505.21951}
}
read the original abstract
Energy consumption and device lifetime are critical concerns for battery-constrained IoT devices. This paper introduces the Feedback-Aided Coding and Energy Transfer (FACET) framework, which synergistically combines adaptive feedback channel coding with wireless power transfer. FACET leverages the saturation effect of feedback coding, where increasing downlink power yields diminishing returns, to design a dual-purpose feedback mechanism that simultaneously guides uplink coding and replenishes device energy. We characterize the inherent tradeoff between feedback precision and harvested power, and formulate a fairness-constrained min-max optimization problem to minimize worst-case net energy consumption. An efficient algorithm based on alternating optimization and Lagrangian duality is developed, with each subproblem admitting a closed-form solution. Simulations show that FACET nearly triples device lifetime compared to conventional feedback coding architectures, and remains robust across a wide range of power regimes. These results suggest that FACET not only improves communication efficiency but also redefines the role of feedback in energy-constrained IoT systems.
Figures
Reference graph
Works this paper leans on
-
[6]
DEEP-IoT: Downlink-enhanced efficient-power Internet of Things,
Y . Shao, “DEEP-IoT: Downlink-enhanced efficient-power Internet of Things,” IEEE Transactions on Wireless Communications, vol. 24, no. 2, pp. 1722–1736, 2025
work page 2025
-
[1]
6G Internet of Things: A comprehensive survey,
D. C. Nguyen, M. Ding, P. N. Pathirana, A. Seneviratne, J. Li, D. Niyato, O. Dobre, and H. V . Poor, “6G Internet of Things: A comprehensive survey,” IEEE Internet Things J. , vol. 9, no. 1, pp. 359–383, 2021
work page 2021
-
[2]
A tutorial on NB-IoT physical layer design,
M. Kanj, V . Savaux, and M. Le Guen, “A tutorial on NB-IoT physical layer design,” IEEE Communications Surveys & Tutorials , vol. 22, no. 4, pp. 2408–2446, 2020
work page 2020
-
[3]
Federated edge learning with misaligned over-the-air computation,
Y . Shao, D. G ¨und¨uz, and S. C. Liew, “Federated edge learning with misaligned over-the-air computation,” IEEE Transactions on Wireless Communications, vol. 21, no. 6, pp. 3951–3964, 2021
work page 2021
-
[4]
A survey on LoRa networking: Research problems, current solutions, and open issues,
J. P. S. Sundaram, W. Du, and Z. Zhao, “A survey on LoRa networking: Research problems, current solutions, and open issues,” IEEE Commu- nications Surveys & Tutorials , vol. 22, no. 1, pp. 371–388, 2019
work page 2019
-
[5]
A survey on LPWA technology: LoRa and NB-IoT,
R. S. Sinha, Y . Wei, and S.-H. Hwang, “A survey on LPWA technology: LoRa and NB-IoT,” ICT Express, vol. 3, no. 1, pp. 14–21, 2017
work page 2017
-
[7]
Feedback in the non- asymptotic regime,
Y . Polyanskiy, H. V . Poor, and S. Verd ´u, “Feedback in the non- asymptotic regime,” IEEE Transactions on Information Theory , vol. 57, no. 8, pp. 4903–4925, 2011
work page 2011
-
[8]
The zero error capacity of a noisy channel,
C. Shannon, “The zero error capacity of a noisy channel,” IRE Trans.Inf. Theory, vol. 2, no. 3, pp. 8–19, 1956
work page 1956
Show all 14 references
-
[9]
Deepcode: Feedback codes via deep learning,
H. Kim, Y . Jiang, S. Kannan, S. Oh, and P. Viswanath, “Deepcode: Feedback codes via deep learning,” Adv Neural Inf Process Syst , 2018
2018
-
[10]
Attentioncode: Ultra-reliable feedback codes for short-packet commu- nications,
Y . Shao, E. Ozfatura, A. G. Perotti, B. M. Popovi ´c, and D. G ¨und¨uz, “Attentioncode: Ultra-reliable feedback codes for short-packet commu- nications,” IEEE Trans. Commu. , vol. 71, no. 8, pp. 4437–4452, 2023
2023
-
[11]
All you need is feedback: Communication with block attention feedback codes,
E. Ozfatura, Y . Shao, A. G. Perotti, B. M. Popovi ´c, and D. G ¨und¨uz, “All you need is feedback: Communication with block attention feedback codes,” IEEE J. Sel. Areas Inf. Theory , vol. 3, no. 3, pp. 587–602, 2022
2022
-
[12]
Wireless power transfer: An overview,
Z. Zhang, H. Pang, A. Georgiadis, and C. Cecati, “Wireless power transfer: An overview,” IEEE Transactions on Industrial Electronics , vol. 66, no. 2, pp. 1044–1058, 2018
2018
-
[13]
Fundamentals of wireless information and power transfer: From rf energy harvester models to signal and system designs,
B. Clerckx, R. Zhang, R. Schober, D. W. K. Ng, D. I. Kim, and H. V . Poor, “Fundamentals of wireless information and power transfer: From rf energy harvester models to signal and system designs,” IEEE Journal on Selected Areas in Communications , vol. 37, no. 1, pp. 4–33, 2018
2018
-
[14]
The hungarian method for the assignment problem,
H. W. Kuhn, “The hungarian method for the assignment problem,” Naval Research Logistics Quarterly , vol. 2, no. 1-2, pp. 83–97, 1955
1955
Reviewed August 7, 2026 · model on record in the stance chip above.
Discussion (0). Sign in to comment.