{"id":"503fb46a-6160-4b6e-9e8b-d8cb5dbaa442","arxiv_id":"2509.08378","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":5.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":2,"one_line_summary":"A multi-objective genetic algorithm plans deployment of heterogeneous EM devices (EMSs, repeaters, IABs) on candidate urban sites, recovering up to 100% of blind-spot coverage in simulated scenarios while trading off cost and energy.","lead":"This paper presents an optimization-based planning method for deploying a mix of passive and active wireless devices, such as electromagnetic skins, smart repeaters, and integrated access and backhaul nodes, to close coverage gaps in urban networks. It reports numerical experiments on two real city maps showing large reductions in blind-spot area at lower cost and energy than increasing base-station power.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Eq. (9)'s isolated-scatterer superposition is untested; inter-SEE shadowing and mutual coupling could change the reported blind-spot reductions.","rationale":"The paper's central claim is that the optimized heterogeneous SEE deployment recovers or enhances coverage in real urban blind spots at lower cost/energy than alternatives. All of the reported performance numbers are obtained by evaluating the coverage cost function with Eq. (10), which depends directly on Eq. (9)'s superposition of independently simulated SEE fields. If inter-SEE interactions—shadowing, multiple scattering, or active-device feedback—are non-negligible, then the predicted received power maps, the Pareto front, and the selected χ_BCS could all be different from what would occur in the field. This is therefore the most load-bearing assumption in the argument. The reader identified the same assumption, and I agree it is the key weakness. The proposed test would settle it by comparing the superposition-based prediction against a simultaneous multi-device simulation for the actual optimized deployment; if the coverage metrics are stable, the concern does not land. I do not see a reason to move the verdict: the paper is already conditional, and this concern reinforces the condition rather than overturning the framework. The Eq. (7) normalization issue is worth fixing but is less central because the raw cost/energy comparisons (ξ and ν) are reported independently of the normalized Φ_CS values.","tokens_in":32831,"tokens_out":6085,"duration_ms":72082,"concrete_test":"For the χ_BCS deployment of Test Case 1 (6 SP-EMS, 1 RP-EMS, 3 SRs at the sites in Fig. 8), run a full ray-tracing simulation with all devices present simultaneously in the same WinProp/OSM model, using the same BTS settings and Pth=−65 dBm. Compare the resulting P(r,t) against the Eq. (9) superposition of the single-device database maps. Report ΔΩ_BCS(t1), ΔΩ_BCS(t2), and the CDF at Pth from both; if the blind-spot reduction changes by more than ~5 percentage points or the mean |ΔP| over the RoIs exceeds ~1 dB, the isolated-superposition assumption is material to the central claim. Also rerun with SRs/IABs disabled to assess active-device feedback sensitivity.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The core quantitative claims (e.g., ΔΩ_BCS(t1)=86.1%, CDF below Pth of 4.9–5.3%) are computed from Eq. (10), which uses Eq. (9) to form the total field as the sum of fields E^(s,n) precomputed for each SEE deployed alone in the scenario, plus the BTS field. This is an isolated-scatterer superposition. In a real deployment, the SEEs are not independent: the 4.58 m^2 EMSs are electrically large at 3.5 GHz (~25λ), so one surface can shadow another; the induced currents on passive EMSs depend on the total incident field including energy scattered by the other SEEs; and SRs/IABs can re-receive signals retransmitted by other active nodes, creating feedback paths. None of these multiple-scattering or mutual-coupling terms appear in Eq. (9). Because the NSGA-II ranks every candidate deployment using this predicted field, any material error in the superposition directly changes the selected solution and the claimed coverage gain. The paper provides no comparison of Eq. (9) against a simultaneous multi-SEE simulation, a full-wave model, or measurement. A smaller but real issue: Eq. (7) as printed makes ξ_max depend on χ, which would force Φ_CS=1 for every nonempty solution, inconsistent with Table IV; presumably a normalization typo, but it should be corrected or clarified.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The manuscript proposes a multi-objective planning strategy for deploying heterogeneous smart electromagnetic entities (SEEs) — SP-EMSs, RP-EMSs, SRs, and IABs — to recover and enhance wireless coverage in urban blind-spot regions. The heterogeneous planning problem (HPP) is encoded as an integer chromosome over candidate installation sites, with three objectives: coverage mismatch (Φ_CV), installation cost (Φ_CS), and energy consumption (Φ_EC). A database of per-SEE ray-traced coverage maps is precomputed, and an integer NSGA-II explores the solution space. Two real-world OSM-based scenarios are considered (Trento Nord, N=20; San Martino, N=25), with T=2 BTS configurations each. The paper reports Pareto fronts and four representative trade-off solutions (χ_BC, χ_BCS, χ_CC, χ_CE), claiming that the best-compromise solution reduces blind-spot area by 86.1%/88.9% in Trento Nord and achieves below-threshold probabilities of 4.9–5.3% in San Martino, at substantially lower cost and energy than the full-coverage solution or a 525–550% increase in BTS transmit power.","tokens_in":33171,"tokens_out":7148,"duration_ms":77902,"significance":"If the simulation methodology is accepted, the paper makes a useful contribution: it integrates four SEE classes into a single planning framework, precomputes coverage maps to make the NSGA-II optimization tractable, uses realistic urban geometry from OSM with full-wave/ray-tracing models, and provides concrete cost/energy trade-offs. The comparison against increasing BTS transmit power is instructive and strengthens the practical motivation. However, the quantitative claims are conditional on the isolated-scatterer superposition in Eq. (9), which is not validated against a simultaneous multi-SEE simulation or measurement. Since that assumption is load-bearing for every coverage result, the numerical conclusions are not yet established at the claimed level. The paper also contains several correctable but important inconsistencies in the formulation and in the key numerical tables.","major_comments":[{"comment":"Equation (9) forms the total field for a candidate deployment as the sum of fields E^(s,n) computed for each SEE deployed alone, plus the BTS field. This isolated-scatterer superposition is the basis for Φ_CV, for every NSGA-II evaluation, and therefore for all reported coverage gains (ΔΩ_BCS, CDF values). It neglects inter-SEE shadowing, mutual coupling, and interactions among active repeaters/IABs. At 3.5 GHz, the 4.58 m² EMSs are roughly 25λ across, so shadowing is physically plausible. The paper provides no comparison against a simultaneous multi-SEE ray-tracing simulation, a full-wave model, or measurement. Since the Conclusions explicitly defer multi-hop links to future work, the current model assumes independent single-hop contributions. Please justify this assumption quantitatively (e.g., run a joint simulation for χ_BCS and χ_BC and report the change in coverage) or weaken the c","section":"Sect. 3.4, Eq. (9)-(10)"},{"comment":"There is a numerical inconsistency in a key reported result. The χ_BCS row lists #SEEs=9, but the component counts are 6 SP-EMS + 1 RP-EMS + 3 SRs = 10. Using Table I costs, the installation cost is 6×$500 + 1×$750 + 3×$3000 = $12,750, not the reported $6,750. The energy value ν=62 W is consistent with the listed components. Since the paper's headline comparison of χ_BCS with χ_BC and with increased BTS power relies on this row, the corrected cost and the resulting comparison must be reported.","section":"Table IV, BCS row"},{"comment":"Equation (7) defines ξ_max = Σ_n max_s {δ_{χn s} ξ_s}. Because δ_{χn s}=1 only for the class selected at site n, the maximum over s of the δ-weighted cost equals the selected class cost, making the denominator in Eq. (6) equal to the numerator for any nonempty χ. Hence Φ_CS ≡ 1 for every nonempty solution, which is inconsistent with the nonzero but less-than-one values in Tables IV and VII. The same issue affects ν_max in Eq. (8). This is presumably a normalization typo, but it must be corrected in the formulation; for example, ξ_max should be the maximum possible cost over all classes and sites, independent of χ.","section":"Sect. 3.3, Eqs. (6)-(7)"},{"comment":"Candidate sites are generated from free-space path-loss ellipses/circles with heuristic exclusion rules. The optimizer is restricted to these N sites, so any location that would be effective under multipath but lies outside the free-space geometry is never considered. No sensitivity analysis is given for the candidate-site selection. Since the planning strategy's final deployments are constrained to this set, the optimality of the reported solutions is contingent on the free-space site model. Please quantify how many candidate sites are removed/retained by the environmental rules, and ideally compare the optimized Pareto front with a candidate set expanded by ray-traced strong-coverage locations.","section":"Sects. 3.1.1-3.1.2"}],"minor_comments":[{"comment":"The non-domination condition says there exists β for which Φ_β{χ_i^(p)} ≤ Φ_β{χ_i^(q)}; for strict non-domination this should be a strict inequality (<). As written, identical objective vectors could be classified as non-dominated.","section":"Sect. 3.5, Step 2"},{"comment":"In the χ_BCS row for w=4, ΔP_avg(t1)=30.4 dB exceeds ΔP_max(t1)=18.1 dB, which is impossible; this appears to be a typo (likely 3.4 dB). Similarly, the χ_CE row for w=1 lists ΔP_avg(t2)=20.5 dB with ΔP_max(t2)=6.2 dB; please check all average/max entries.","section":"Table VIII"},{"comment":"The #SEEs column in Table IV is inconsistent for χ_BCS as noted in the major comments; please also verify the counts for the other rows for completeness (the BC row is consistent: 1 SR + 2 IABs = 3 SEEs).","section":"Sect. 4, Table IV"},{"comment":"The BTS power-increase comparison (PTX=125 W and 130 W) is informative, but please clarify explicitly whether these curves are generated with the same ray-tracing framework and the same BTS model apart from transmit power, so the comparison is apples-to-apples.","section":"Fig. 15 / Sect. 4"},{"comment":"The abbreviations PSE and ASE are introduced but the text mostly uses SEE; please keep the terminology consistent throughout.","section":"General"}],"recommendation":"major_revision","confidential_remarks":"The paper is within the scope of the journal and the proposed framework is potentially useful for SEME planning. The main obstacle is the unvalidated superposition assumption in Eq. (9), which underpins all quantitative coverage claims. The table and normalization inconsistencies are easily fixed, but the missing joint multi-SEE validation requires additional work. I therefore recommend major revision rather than acceptance at this stage."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The genuinely new thing here is that the planning problem is broadened from single-type EMS deployment to a heterogeneous mix of SP-EMS, RP-EMS, SR, and IAB, with a three-objective integer NSGA-II giving a Pareto front across coverage, cost, and energy. That is a practical step beyond the authors' earlier EMS-only work [37,38], and the two real-map test cases (Trento Nord, San Martino) are a reasonable proof of concept. The candidate-site rules are sound engineering heuristics, the cost/energy figures come from datasheets, and the comparison against a 525–550% BTS power increase makes a fair point: smart deployment can beat brute force.\\n\\nThe big soft spot is Eq. (9). The total field is taken as a superposition of fields precomputed for each SEE deployed alone, with no check against a joint multi-SEE simulation, a full-wave model, or measurements. At 3.5 GHz the EMSs are around 25 wavelengths long, so inter-device shadowing and mutual coupling could be real; active SRs/IABs can also create feedback paths. Because this predicted field is what the NSGA-II uses to rank every candidate deployment, any error here changes the selected solution and the headline ΔΩ and CDF numbers. I do not think this is fatal—the method still has value as a planning tool—but the authors need to justify the approximation or test it on at least one scenario with a simultaneous multi-SEE simulation.\\n\\nThere is also a clear typo in Eq. (7): as printed, ξ_max depends on χ, which makes Φ_CS equal to 1 for every nonempty solution, inconsistent with Table IV. It should be a constant bound. Minor points: only one of 50 seeds is reported, with no spread; and the free-space path-loss formulas for candidate-site regions are explicitly filtered by feasibility rules, so I would not count that as a flaw. No code or data is released, which is common in this literature.\\n\\nOverall, the paper is worth a serious referee. The planning framework is useful to people working on SEME deployment, and the flaws are fixable: fix the normalization, validate the superposition, and show seed variance. I would send it out and ask for those three things.","headline":"Useful extension of EMS planning to heterogeneous devices with a credible cost/energy trade-off, but the superposition assumption in Eq. (9) needs validation before trusting the coverage numbers.","tokens_in":33693,"tokens_out":2891,"would_cite":true,"duration_ms":33553,"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 mixed deployment of passive surfaces and active repeaters, chosen by a multi-objective planner, restores urban wireless coverage at far lower cost and energy than raising base-station power.","keywords":["smart electromagnetic environment","wireless coverage planning","multi-objective optimization","NSGA-II","electromagnetic skins","smart repeaters","integrated access and backhaul","urban propagation"],"falsifier":"Install, or full-wave simulate, the best-compromise deployment - 6 SP-EMSs, 1 RP-EMS, and 3 SRs in Trento Nord - with all devices present at once, and measure the received-power CDF inside the original blind-spot region at Pth = -65 dBm. The paper's superposition model predicts 7.8 percent of samples below threshold at t1 and 5.8 percent at t2; if the measured CDF is materially higher, the neglect of inter-device shadowing and re-radiation is the cause.","tokens_in":1842,"feed_emoji":"📡","tokens_out":2191,"duration_ms":83676,"temperature":0.7,"pith_summary":"This paper argues that a city's dead zones need not be fixed with more base stations or more transmit power. Instead, a planner can deploy a heterogeneous mix of cheap passive reflectors and modest active repeaters, and a multi-objective search can choose where each device goes. The authors formalize the choice as a planning problem, solve it with an integer genetic algorithm, and test it on two real districts of Trento, Italy, at 3.5 GHz. In the first district, the best-compromise deployment shrank the blind-spot area by 86-89 percent while costing about a third of the full-coverage solution and using a tenth of its energy; matching that coverage by raising base-station power would have required a 525-550 percent increase in transmit power. The takeaway is that the propagation environment itself - walls, facades, and poles - can be treated as a design resource rather than an obstacle.","feed_headline":"Smart surface mix erases 86-89% of urban dead zones","feed_subtitle":"Planner picks passive skins and repeaters that beat a 525% power hike on cost and energy.","key_machinery":"The load-bearing mechanism is the Heterogeneous Planning Problem (HPP): an integer chromosome of length N encodes which of four device types (SP-EMS, RP-EMS, SR, IAB) occupies each admissible site, with zero meaning no device. Feasibility is pre-computed geometrically - an ellipse for passive skins traced from the base station and each blind-spot barycenter under a single-reflection path-loss bound, and the intersection of two visibility circles for active devices. Each device's coverage contribution is pre-computed once with a ray-tracing tool and stored, so any deployment's field is evaluated as a superposition of stored maps (Eq. 9). An integer NSGA-II then searches the resulting space un","core_discovery":"The central claim is that a heterogeneous smart electromagnetic environment - static and reconfigurable electromagnetic skins (passive reflective surfaces), smart repeaters, and integrated access-backhaul nodes placed on building facades and poles - can recover coverage in user-defined regions of interest of a real urban scenario, and that the recovery can be planned as a three-objective optimization. Each candidate installation site is assigned one device type or none; the optimizer minimizes the time-averaged coverage shortfall, the installation cost, and the energy consumption simultaneously, producing a Pareto front of trade-off deployments rather than a single answer. The authors report","pith_inferences":["The reported gains rest on a superposition assumption (Eq. 9) that ignores device-to-device interactions; a field trial or full-wave simulation of a dense deployment would show whether real gains close to 86-89 percent survive mutual shadowing and re-radiation.","The same formulation could be re-targeted almost unchanged at indoor scenarios or Wi-Fi access points - the paper lists these as future work - since the cost terms and the integer encoding do not depend on the 3.5 GHz outdoor setting.","The reconfigurable skins are used here in a single-bit static configuration; using their switching capability across time instants could let one deployment track the moving blind spots the paper models rather than just the union of them.","The paper's comparison with a 525-550 percent transmit-power increase suggests an operator-side test: measure the marginal cost of extra BTS power against the amortized cost of a skin-and-repeater deployment, a trade-off the installation-cost and energy terms do not fully capture."],"forward_implications":["Network operators can restore QoS in identified blind spots without new base stations or fiber backhaul, using the planner's Pareto front to pick a deployment matched to their budget.","Passive electromagnetic skins do substantial work on their own: in the San Martino test, skins alone removed up to roughly 75 percent of one region's blind spot, showing that zero-energy devices can carry part of the load.","The same planning pipeline carries over to any class of device, since the chromosome encoding and cost terms are defined generically over the device alphabet.","The full-coverage benchmark quantifies what complete restoration costs, and every cheaper Pareto point makes explicit the coverage the operator trades away for cost and energy savings.","The reported best-compromise solutions use far fewer active devices than the full-coverage solution (e.g., 3 SRs and 0 IABs vs. 2 IABs and 1 SR in Trento Nord), indicating that the planner is favoring a mostly passive infrastructure."],"supporting_citations":[{"why":"Supplies the NSGA-II multi-objective genetic optimizer used to sample the Pareto front of deployments.","marker":"[61]"},{"why":"Provides the EM-skin planning approach, including the incidence/reflection geometry and the base-station model reused here.","marker":"[37]"},{"why":"Gives the SP-EMS synthesis procedure used in the SEED block to design the static skins.","marker":"[14]"},{"why":"Gives the RP-EMS design method with PIN-diode unit cells used in the SEED block.","marker":"[25]"},{"why":"Defines the heterogeneous SEE ecosystem and the device classes the alphabet is built from.","marker":"[11]"},{"why":"Supplies the smart-repeater model (two dual-polarized phased arrays) adopted for SRs.","marker":"[48]"},{"why":"Supplies the IAB model as a micro base station with in-band wireless backhaul.","marker":"[55]"},{"why":"The ray-tracing simulator used to compute the reference coverage and the per-device coverage database.","marker":"[64]"},{"why":"The geographic database from which the two real urban scenarios are modeled.","marker":"[63]"}],"fun_headline_variants":["Smart EM mix cuts urban dead zones 86-89%","Heterogeneous EM devices recover coverage in cities at low energy","Planner optimizes skins, repeaters, IABs to erase 86-89% of blind spots","Optimized EM mix: 86-89% fewer dead zones with lower cost and power","Smart surfaces and repeaters cut dead zones 86-89% in urban areas"],"cache_read_input_tokens":35328,"weakest_assumption_plain":"The planning tool predicts coverage by adding up each device's contribution as if it acted alone in the scene; if installed devices interact - one skin shadowing another or re-radiating onto it - the real coverage could fall short of the predicted 86-89 percent reductions.","fun_headline_variants_meta":{"raw":{"variants":["Smart EM mix cuts urban dead zones 86-89%","Heterogeneous EM devices recover coverage in cities at low energy","Planner optimizes skins, repeaters, IABs to erase 86-89% of blind spots","Optimized EM mix: 86-89% fewer dead zones with lower cost and power","Smart surfaces and repeaters cut dead zones 86-89% in urban areas"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.001626,"raw_usage":{"total_tokens":6262,"prompt_tokens":661,"completion_tokens":5601,"prompt_tokens_details":{"cached_tokens":256},"prompt_cache_hit_tokens":256,"prompt_cache_miss_tokens":405,"completion_tokens_details":{"reasoning_tokens":5495}},"tokens_in":405,"tokens_out":5601,"duration_ms":40125,"temperature":1.0,"reasoning_tokens":5495,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-04T20:38:29.240661+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Install, or full-wave simulate, the best-compromise deployment - 6 SP-EMSs, 1 RP-EMS, and 3 SRs in Trento Nord - with all devices present at once, and measure the received-power CDF inside the original blind-spot region at Pth = -65 dBm. The paper's superposition model predicts 7.8 percent of samples below threshold at t1 and 5.8 percent at t2; if the measured CDF is materially higher, the neglect of inter-device shadowing and re-radiation is the cause.","supporting_citations":[],"review_version":1}