{"id":"9f721b3e-3a5f-491a-aee7-e56f49d6a94a","arxiv_id":"2511.21271","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":5.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":5,"one_line_summary":"An adaptive VLC framework that partitions the room into activity and non-activity zones and switches LED power-allocation objectives based on user location achieves 53.59% simulated energy savings.","lead":"This paper proposes an adaptive indoor visible-light system that switches between energy-saving and signal-uniformity modes based on where the user is. The authors report 53.59% energy savings and a 57.79% improvement in SNR uniformity in simulation.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Energy-savings claim is not reproducible: baseline power and trajectory dwell times are unspecified; the 53.59% figure may be an artifact of arbitrary choices.","rationale":"After reading the paper, the optimization formulations in Section III are internally consistent, and the SNR uniformity improvement (57.79%) follows directly from the reported variance reduction (1.99 to 0.84). The localization self-consistency issue raised by the reader is real, but it is a standard simulation limitation that could be addressed with experimental validation. In contrast, the energy-savings figure—the paper's primary objective—is not even well-defined in principle because neither the baseline power nor the trajectory dwell times are specified. This makes the 53.59% claim non-reproducible and potentially misleading. The concern is not about correctness of the math but about the interpretability of the key numerical result. A concrete recalculation with explicit parameters would settle whether the number is meaningful. Since the reader already assigned a CONDITIONAL verdict, this concern reinforces that condition rather than changing it.","tokens_in":8425,"tokens_out":5065,"duration_ms":55951,"concrete_test":"Obtain the exact baseline LED power and per-phase trajectory durations from the authors. Recompute time-averaged power for the adaptive versus non-adaptive systems under a grid of baseline values (e.g., each LED at Pmin=10 W, Pmax=80 W, or a total of 361.60 W) and dwell-time ratios (e.g., no-user fraction 10%, 30%, 50%). If the savings percentage shifts by more than 10 percentage points across this grid, the 53.59% headline is not robust.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim's 53.59% energy-savings figure depends on two unspecified inputs: (1) the power level of the non-adaptive baseline described as 'all LEDs operating at fixed normal power' (Section IV-C), and (2) the time durations of each phase in the user trajectory of Fig. 3. The paper only reports the optimized total power (300.06 W) and an unoptimized total (361.60 W) for the enhanced mode, but never states the baseline LED power nor the dwell times for the no-user, non-activity, and activity phases. Because no-user mode can drive all LEDs to Pmin = 10 W (Section III-A) and uniformity mode can lower power substantially, the savings percentage can vary widely with these choices. Without explicit values, the 53.59% headline is not reproducible and could be inflated by an excessively high baseline or by spending most of the trajectory in the no-user mode. This is a load-bearing gap in the paper's primary quantitative claim, not a mere presentation issue.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper proposes an adaptive integrated sensing, communication, and illumination (ISCI) framework for indoor visible-light systems. The receiving plane is first restricted to the intersection of the minimum enclosing circle of the LED projection convex hull and the physical room boundary, and then partitioned into an activity area (maximum inscribed circle) and a surrounding non-activity area. A user is localized through NLOS reflected-power variations, and this location is used to switch among three modes: a low-power no-user mode, a uniformity mode that minimizes SNR variance, and an enhanced mode that minimizes total transmit power subject to SNR and illuminance constraints in the activity area. Numerical simulations report 53.59% energy savings over a non-adaptive baseline, a 57.79% improvement in SNR uniformity, and a mean localization error of 0.071 m. The optimization formulations (convex QP and LP) are algebraically consistent, but the quantitative claims rest on several unreported parameters and on a self-consistent simulation methodology.","tokens_in":8720,"tokens_out":3733,"duration_ms":39999,"significance":"If the reported gains are robust, the framework is a useful contribution to energy-efficient VLC-ISAC: the geometric construction of the receiving plane from the LED deployment is original and tractable, the mode-switching policy is practically motivated, and the QP/LP formulations are clearly derived with a sound convexity argument. The paper also addresses a real gap in the literature, which has focused on throughput and sensing accuracy rather than energy consumption. However, the numerical evidence is currently not reproducible and the localization validation is circular (same forward model for both the lookup table and the 'actual' measurements). The significance is therefore conditional on supplying the missing parameters and on testing against independent data or a mismatched model.","major_comments":[{"comment":"The headline claim of 53.59% energy savings is not reproducible. The non-adaptive baseline is only described as 'all LEDs operating at fixed normal power', but the per-LED power value and the time durations of the trajectory phases (no-user, non-activity, activity) are never given. Because no-user mode drives all LEDs to P_min = 10 W and uniformity mode can substantially reduce power, the savings percentage depends critically on these choices. Please report the baseline power level, the dwell times of each trajectory phase, and the per-mode energy integrals used to compute the 53.59% figure.","section":"Section IV-C, Fig. 3"},{"comment":"The 0.071 m mean localization error is a self-consistency check rather than an independent validation. The same NLOS channel model (Eqs. (4)-(5)) is used both to build the offline lookup table containing the predicted power variations and to generate the online 'actual' power variations in Eq. (9). No measurement noise, model mismatch, or occlusion is modeled, and the occlusion set K in Eq. (7) is never specified. The paper should either validate with experimental data or explicitly model mismatch/noise, and it must define K.","section":"Section II-B, Eqs. (4)-(10); Section IV-C"},{"comment":"The enhanced-mode SNR threshold Λ_th is never assigned a numerical value, yet it is the key constraint in the LP (Eq. (24c)) and directly affects the reported optimized total power of 300.06 W. Without this value, the feasibility of the enhanced-mode solution cannot be checked. Please give Λ_th and the exact enhanced illuminance bounds E_min^e/E_max^e; the phrase '500 lx above' is ambiguous because it is not clear whether both bounds are shifted.","section":"Section III-C, Eq. (24c)"},{"comment":"The numerical results are based on a single random LED deployment and one 'representative' user trajectory, but the random seed, the exact LED/PD coordinates, and the trajectory waypoints and time stamps are not provided. Consequently, the reported error and energy savings have no error bars and cannot be reproduced. Please provide the deployment parameters and a precise trajectory definition, or run Monte Carlo simulations over multiple deployments and trajectories and report the distribution of results.","section":"Section IV, Figs. 3-4"}],"minor_comments":[{"comment":"The centering matrix M is denoted by the same symbol as the number of LEDs M. Rename one of them (e.g., C_L for the centering matrix) to avoid confusion.","section":"Section III-B, Eq. (20)"},{"comment":"The symbol K is used both for the number of candidate positions and for the set of non-occluded surface elements in Eq. (7). These should be distinguished (e.g., K_cand for the number of candidates).","section":"Section II-B"},{"comment":"The SNR simplification neglects the constant noise terms (shot-noise background and thermal noise) with the justification of a 'high-SNR reliable channel'. Please quantify the resulting approximation error or state the operating regime in which this simplification is accurate.","section":"Section III-B, Eq. (13)-(15)"},{"comment":"Several parameters are only stated as 'following [10], [13]' (e.g., reflection coefficients, detector areas, noise parameters). For a self-contained numerical section, these values should be listed or a table provided.","section":"General"}],"recommendation":"major_revision","confidential_remarks":"The core issue is reproducibility: the central quantitative claims (53.59% energy savings, 57.79% uniformity improvement, 0.071 m localization error) depend on parameters and simulation choices that are not reported. The localization claim is also circular. These are fixable within the manuscript's scope if the authors supply the missing parameters, report a precise trajectory and deployment, and either add an independent validation or reframe the localization results as a simulation self-consistency check. I do not see a fundamental flaw in the optimization framework itself."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The idea at the center of this paper is simple and defensible: partition the floor into an activity area and a surrounding non-activity area, and switch the LED power optimization objective depending on which area the user is in. The MEC/MIC construction is a clean way to define those regions from a random LED layout, and the QP/LP formulations in Section III are standard and correctly derived. The writing is clear. Credit where it's due: the geometric partition plus location-triggered switching is a reasonable extension of known VLC resource allocation work, and the simulation set-up is mostly explained.\n\nThe problems are in the validation. The headline 53.59% energy savings is computed against a 'non-adaptive system' whose power level is never stated. With no-user mode driving everything to Pmin = 10 W, the savings percentage will depend heavily on how long the user spends outside the room and what 'normal power' the baseline uses. The same goes for the enhanced-mode SNR threshold Λ_th in Eq. (24c), which is never given. None of these values appear in the numerical section, so the central quantitative claim is not reproducible.\n\nThe localization error is also weaker than it looks. The 'actual' power variations in the online phase are generated by the same NLOS channel model that builds the lookup table. That makes the 0.071 m mean error a self-consistency check rather than evidence of performance under real conditions. No measurement noise, model mismatch, or occlusion is modeled; the occlusion set K in Eq. (7) is never specified. And everything rests on one random LED deployment and one trajectory, so the generality of the numbers is unknown.\n\nThese are addressable gaps rather than a broken argument. The optimization logic is coherent, and the idea could be made solid with explicit baseline power, Λ_th, dwell times, noise injection, and ideally an experimental check. As it stands, the paper is a plausible framework with unsupported headline numbers. A referee could get it into shape, but only if the authors are willing to pin down the missing parameters and de-circularize the localization evaluation.\n\nI'd send it to review rather than desk reject — the idea is worth a careful look, and the flaws are omissions, not incoherence. For my own work I wouldn't cite it yet. It might be useful for a reading group discussion about simulation-validation practices, but it's not urgent.","headline":"A plausible adaptive-control formulation for VLC energy savings, but the headline 53.59% figure is underdetermined because the baseline power and trajectory timing are never defined.","tokens_in":9192,"tokens_out":2370,"would_cite":false,"duration_ms":25672,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"A visible-light system that switches its LED power-control objective based on user location reports 53.59% energy savings while keeping communication, illumination, and sensing constraints satisfied.","keywords":["visible light communication","integrated sensing and communication","adaptive lighting control","NLOS sensing","SNR uniformity","energy savings","user localization","illumination constraints"],"falsifier":"Run the same lookup-table localization in a physical indoor VLC testbed with a moving person and compare the estimated trajectory to camera-tracked ground truth; if the mean localization error exceeds roughly the margin between the activity-area boundary and neighboring sensing points, the mode switches will frequently be wrong and the claimed 53.59% energy savings will not materialize.","tokens_in":8314,"feed_emoji":"💡","tokens_out":3450,"duration_ms":36429,"temperature":0.7,"pith_summary":"The paper argues that indoor VLC systems waste power by treating every spot in a room as needing the same high-performance link. It proposes an adaptive integrated sensing, communication, and illumination framework that first divides the floor into an activity area for focused work and a surrounding non-activity area, then uses user location—estimated from non-line-of-sight light reflections off the user's body—as a switch. When the user is in the activity area, the LEDs minimize total transmit power subject to high SNR and illumination requirements; elsewhere, they maximize SNR uniformity; when no user is present, they drop to a low-power sensing mode. Numerical results show 53.59% energy savings over non-adaptive operation, a 57.79% improvement in SNR uniformity, and a mean localization error of 0.071 m. If these results hold under realistic conditions, location-aware lighting control can reconcile energy efficiency with VLC, ISAC, and visual comfort.","feed_headline":"Sensing-driven LED control saves 53.6% power","feed_subtitle":"A visible-light system that switches LED modes by sensing user position keeps illumination, SNR, and localization targets in check.","key_machinery":"The central object is the location-dependent objective switch: user position estimated by matching measured NLOS power variations to an offline lookup table selects among three optimization problems. Supporting geometry: the minimum enclosing circle (MEC) of the LED convex hull defines the receiving plane, and the maximum inscribed circle (MIC) defines the activity area. The uniformity mode minimizes SNR variance via the quadratic form (1/L)||M A p||², a convex QP; the enhanced mode minimizes total LED power as an LP with SNR and illuminance constraints. This combination converts user location into a control signal for LED power allocation.","core_discovery":"The central claim is that the conflict between VLC/ISAC performance, energy consumption, and illumination comfort can be resolved by making the optimization objective depend on the user's activity state. The receiving plane is reduced to the intersection of the minimum enclosing circle of the LED convex hull with the room boundaries, and the activity area is the maximum inscribed circle of that hull. User position estimated from NLOS reflections acts as a switch: enhanced mode solves a linear program minimizing total LED power while enforcing SNR and illuminance thresholds over the activity area; uniformity mode solves a convex quadratic program minimizing SNR variance over the whole plane;","pith_inferences":["Editorial extension: the 0.071 m localization error comes from a self-consistent simulation where the offline NLOS model is identical to the online 'actual' one; real sensors with noise, occlusion, or model mismatch would likely degrade this accuracy, so the practical energy savings may be lower.","Editorial extension: the single-user trajectory result does not address multi-user scenarios or rapid mode transitions; flicker and communication interruption between modes are acknowledged as future work, making the energy number an optimistic bound for realistic deployments.","Editorial extension: a testable extension is to sweep LED/PD placements and room shapes to see how robust the 53.59% savings and 0.071 m error are, since the geometric MEC/MIC regions depend on the LED layout.","Editorial extension: the framework could be combined with occupancy prediction to pre-switch modes before a user crosses the activity-area boundary, reducing mode-switch latency and potential service disruption."],"forward_implications":["If correct, VLC systems can use existing LEDs and photodiodes for both sensing and control without extra hardware, since the same NLOS reflections used for ISAC drive the energy-saving policy.","The no-user low-power mode can extend LED lifetime and reduce standby consumption in addition to saving energy.","SNR-uniformity optimization over the MEC plane removes boundary weak spots: 0% of PDs fall below the deviation threshold, versus 10.91% on the reference plane.","The claimed 53.59% savings is achieved with a non-optimized, random LED deployment, suggesting strategic placement could improve both energy savings and localization accuracy.","Illumination constraints from the ISO/CIE 8995-1 standard are satisfied simultaneously with communication constraints, so visual comfort need not be sacrificed for energy efficiency."],"fun_headline_variants":["Adaptive VLC LEDs cut energy 53.6% via user sensing","User position switches LED modes, saving 53.6% power","Sensing-driven lighting saves 53.6% energy, boosts SNR","Visible light system adapts to user, saves 53.6%","LEDs sense users to save 53.6% power, improve uniformity"],"cache_read_input_tokens":2304,"weakest_assumption_plain":"The entire adaptive policy rests on localization accuracy: the offline NLOS lookup table must match the online measured power variations, and the paper does not model measurement noise, occlusion, or model mismatch—if real reflections deviate from this self-consistent simulation, the reported 0.071 m error and correct mode switches will not transfer to practice.","fun_headline_variants_meta":{"raw":{"variants":["Adaptive VLC LEDs cut energy 53.6% via user sensing","User position switches LED modes, saving 53.6% power","Sensing-driven lighting saves 53.6% energy, boosts SNR","Visible light system adapts to user, saves 53.6%","LEDs sense users to save 53.6% power, improve uniformity"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000322,"raw_usage":{"total_tokens":1642,"prompt_tokens":735,"completion_tokens":907,"prompt_tokens_details":{"cached_tokens":256},"prompt_cache_hit_tokens":256,"prompt_cache_miss_tokens":479,"completion_tokens_details":{"reasoning_tokens":810}},"tokens_in":479,"tokens_out":907,"duration_ms":9227,"temperature":1.0,"reasoning_tokens":810,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-03T20:00:42.216397+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Run the same lookup-table localization in a physical indoor VLC testbed with a moving person and compare the estimated trajectory to camera-tracked ground truth; if the mean localization error exceeds roughly the margin between the activity-area boundary and neighboring sensing points, the mode switches will frequently be wrong and the claimed 53.59% energy savings will not materialize.","supporting_citations":[],"review_version":1}