{"id":"87b6892e-005b-44ae-89c3-4f57e7c799d4","arxiv_id":"2412.02649","paper_version":1,"verdict":"CONDITIONAL","confidence":"HIGH","novelty_score":5.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":1,"one_line_summary":"A sequential mode-selection algorithm, which picks communication-only transmitters first and sensing receivers second, gives the best balance of fewer active access points and low runtime in mmWave cell-free ISAC networks.","lead":"This paper designs algorithms that switch each access point in a millimeter-wave cell-free network between transmitting joint sensing and communication signals, receiving reflected sensing signals, or switching off, to minimize active access points while meeting communication and sensing targets. The work matters because energy-efficient operation is a key goal for 6G networks that combine communication with radar-like sensing.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The headline performance claim is only demonstrated under an unconditional LOS sensing channel, which contradicts the paper's blockage motivation; if AP-target links are blocked the CRLB constraint collapses and sequential mode selection may be infeasible.","rationale":"I agree with the reader's weakest assumption. The paper is internally consistent under the stated LOS model, and the optimization machinery (binary A/B lifting, linear-fractional constraints) is sound. However, the central claim is a comparative systems claim about mmWave ISAC in blockage-prone environments, and the only sensing channel model used is LOS-only. The manuscript explicitly flags this assumption, which is exactly the kind of self-identified limitation that should bind the verdict. A secondary reproducibility concern: Algorithm 3's pseudocode has an outer 'while CRLB > eta' loop whose body only re-checks the communication subproblem, so as printed it does not implement the prose description of adding TX APs until the sensing threshold is met; this affects the heuristic baseline, not the main sequential-versus-alternation comparison. Because the LOS concern directly undermines applicability but not the internal logic, CONDITIONAL remains the right verdict; no change is needed.","tokens_in":7800,"tokens_out":11939,"duration_ms":132502,"concrete_test":"Re-run the Rosslyn experiment with a ray-traced or probabilistic blockage map: for each AP-target pair mark the link as LOS or blocked, set the sensing channel gain alpha_{m,n} to zero for blocked links, and recompute the CRLB constraint (13c). Compare active AP counts, feasibility, and runtime of Alg. 1, Alg. 2, and Alg. 3 at blockage probabilities p = 0, 0.2, 0.5. If the sequential scheme becomes infeasible or its AP reduction relative to alternation shrinks substantially, the headline claim is conditional on the LOS assumption.","verdict_should_be":"UNCHANGED","load_bearing_attack":"Section II-B states 'We assume an available line-of-sight (LOS) link between the APs and the target' and Eq. (9) models every sensing channel as the rank-one LOS response a_m(phi_m)a_n^T(phi_n); the CRLB constraint (13c) inherits this assumption. The Introduction, by contrast, motivates cell-free massive MIMO precisely as a way to cope with mmWave blockage. Under blockage, an AP selected as an ISAC transmitter for communication may have no usable sensing path to the target, so the CRLB expression and the sensing guarantee in (13c) no longer hold. The sequential scheme is especially exposed because Algorithm 2 selects a from (19) using only communication constraints and then solves (18) for b; it never revises a when the sensing constraint is infeasible. The simulations keep the target position fixed and do not include blocked AP-target links, so the claimed ~50% AP reduction and the 'best performance-complexity balance' are untested in the very blockage-prone scenarios the paper targets.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"This paper studies AP mode selection in a mmWave cell-free massive MIMO-ISAC system, where each AP is assigned to be an ISAC transmitter, a sensing receiver, or shut down. The objective is to minimize the number of active APs subject to downlink SINR constraints and a Cramér-Rao lower bound (CRLB) constraint for passive target tracking. Because the joint problem is large and non-convex, the authors propose three suboptimal algorithms: an alternating optimization, a sequential optimization that first chooses ISAC transmitters based only on communication constraints and then chooses sensing receivers, and a heuristic based on channel gain and distance ordering. The paper claims that the sequential scheme achieves an AP reduction comparable to alternation while running about eight times faster, and that both optimization-based schemes reduce the number of active APs by about 50% relative to the heuristic. The derivations are built on external results for distributed radar CRLB and cell-free massive MIMO spectral efficiency, and the binary linearizations of the CRLB constraint in (15)-(17) are presented as exact reformulations for the fixed-vector subproblems.","tokens_in":8035,"tokens_out":6633,"duration_ms":72490,"significance":"If the results hold under the stated assumptions, the sequential optimization scheme is a practically relevant contribution: it avoids solving a joint mixed-integer problem in both mode vectors, and the reported runtime reduction relative to alternation is substantial. The paper does not fit any parameters to data, and the use of standard CRLB and spectral-efficiency expressions from the cited literature is appropriate. The linearization steps in Section III are internally consistent. However, the evaluation is built on an unconditional LOS sensing channel, a single fixed target position, and an underspecified heuristic parameter R, so the headline claims about blockage resilience and the 'best performance-complexity balance' are not yet demonstrated for the scenarios that motivate the work. The contribution is therefore significant but conditional.","major_comments":[{"comment":"The sensing model assumes an available LOS link between every AP and the target, and Eq. (9) represents every sensing channel as the rank-one LOS response a_m(phi_m) a_n^T(phi_n). Since the CRLB constraint (13c) inherits this assumption, the sensing guarantees collapse if an AP-target link is blocked. This directly conflicts with the Introduction's motivation that cell-free massive MIMO overcomes mmWave blockage. The manuscript should either incorporate a blockage-aware sensing model with NLoS links or explicitly restrict all claims to LOS-only sensing scenarios. In the latter case, the sequential Algorithm 2 is especially exposed, because it selects a using only communication constraints and never revises the transmitter set when the sensing constraint is infeasible.","section":"Section II-B, Eq. (9), Eq. (13c)"},{"comment":"The text states that 'the target position is kept fixed' in the simulation experiments, even though the system model and introduction emphasize changing target and blockage scenarios. All reported AP counts therefore correspond to a single geometric configuration of the target relative to the APs, and no blocked AP-target links are included. Consequently, the claimed ~50% AP reduction and the 'best performance-complexity balance' are not tested across the target positions or blockage realizations that the paper targets. Please average over target positions and blockage realizations and report the spread of the results.","section":"Section IV, simulation setup"},{"comment":"As written, Algorithm 3 contains an infinite loop whenever the sensing constraint is not satisfied after the communication constraints become feasible. The outer 'while CRLB > eta' loop contains only the inner communication-feasibility loop, which modifies a; once P1 is feasible, no quantity affecting the CRLB is updated, so the outer condition never changes. The prose says that the highest-gain AP is added as an ISAC transmitter until the sensing threshold is met, but this step is missing from the pseudocode. Please correct the algorithm and specify the exact implementation used to produce Table I and Fig. 2, since the heuristic comparison depends on it.","section":"Section III-C, Algorithm 3"},{"comment":"The heuristic selects the 'closest R APs' as initial sensing receivers, but the value of R is not specified in the input list, the text, or the simulation setup. The performance of the heuristic, and hence the paper's comparison against it, may depend strongly on R. Please report the chosen R and ideally show sensitivity of the comparison to this parameter.","section":"Section III-C, heuristic parameter R"}],"minor_comments":[{"comment":"The claim that the sequential algorithm is 'approximately eight times' faster than alternation holds only for K=4 (201.8 vs 26 seconds); the ratios are about 5.9 for K=6 and 4.5 for K=8. Please qualify the runtime claim accordingly.","section":"Table I, Conclusion"},{"comment":"The alternation algorithm does not specify a maximum number of random restarts when the initial b is infeasible; as written it could loop indefinitely. Please add a stopping criterion and a failure-handling rule.","section":"Algorithm 1"},{"comment":"The notation tr(Ct) is used without defining Ct, and the displayed fraction in Eq. (12) is difficult to parse because the denominator is split across lines. Please clarify the expression and define all variables.","section":"Eq. (12)"},{"comment":"The 'modified version of the LP-MMSE precoder' from [11] is only mentioned in passing. A brief description of the modification would improve reproducibility.","section":"Section II-A"},{"comment":"The assumption that communication and sensing signals occupy different frequency bands eliminates all ISAC interference but also removes the spectral-sharing benefit of ISAC. This should be stated as a limitation in the conclusion rather than only in the system model.","section":"Section II"},{"comment":"The conclusion states that the optimization schemes reduce active APs by 50% relative to the heuristic, but the exact numerical values are not given in the text. Please state the measured AP counts or percentages explicitly.","section":"Fig. 2 and Section V"}],"recommendation":"major_revision","confidential_remarks":"The technical core of the paper is sound: the CRLB and spectral-efficiency expressions come from the cited literature, and the binary linearizations for the fixed-vector subproblems are correct. There is no circularity or data-fitting concern. The reason for major revision is the gap between the blockage-motivated framing and the LOS-only, single-target simulation evaluation, together with the underspecified/buggy heuristic pseudocode. If the authors can supply blocked-link or NLoS sensing experiments, or alternatively narrow the paper's claims to LOS sensing, and fix Algorithm 3, the contribution could be acceptable for publication."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"This is a credible systems paper on AP mode selection for mmWave cell-free ISAC. The formulation—minimizing active AP count with a CRLB sensing constraint and binary TX/RX/off modes—is not in the cited literature, and the comparison of three algorithms is a reasonable contribution. The alternating and sequential optimizations are built correctly on top of the authors' earlier work; the binary linearizations in (15) and (17) are valid, and the CRLB/SE expressions come from established sources.\n\nThe main result, that the sequential scheme cuts active APs by roughly half relative to the heuristic and runs eight times faster than alternating, is plausible. The runtime difference is the most convincing part.\n\nNow the soft spots. The LOS sensing assumption is in real tension with the blockage motivation. Every sensing channel is modeled as rank-one LOS, and the CRLB constraint inherits that. In a blocked urban scenario, the sensing performance could be far worse, and the mode choices might be wrong. The simulations sidestep this: the target position is fixed, and no blocked AP-target link is tested. So the headline numbers are only shown for favorable geometry.\n\nMore concretely, Algorithm 3 as printed has a loop bug: once the inner 'while P1 not feasible' exits, the outer 'while CRLB > η' will never add another TX AP even though that is exactly what the surrounding text says should happen. In effect the pseudocode cannot terminate in cases where communication is feasible but sensing is not. Either the pseudocode is wrong or the implementation differs; it needs to be fixed. Related, the sequential algorithm never revises its TX set when the sensing constraint is infeasible, and the paper does not report how often that happens.\n\nThere are also no error bars or optimality-gap bounds, so the claimed 50% reduction is one number from one setup. These are fixable in revision.\n\nOverall, the central idea holds up if you read the LOS assumption as a deliberate simplification, and the paper is a moderate, honest extension of prior work. It is not a breakthrough, but it is clearly written and worth a serious referee. I'd send it out with major revision: fix the heuristic pseudocode, add blocked or at least variable target positions, and give error bars or a bound.","headline":"Solid, moderate systems paper on AP mode selection; the headline 50% claim is untested under blockage and the heuristic pseudocode has a loop bug, but the optimization core is sound enough for peer review.","tokens_in":8548,"tokens_out":6071,"would_cite":true,"duration_ms":63605,"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":"Sequential AP mode selection matches alternating optimization's active-AP count while running eight times faster.","keywords":["ISAC","cell-free massive MIMO","AP mode selection","power allocation","multi-static sensing","mmWave","CRLB","energy efficiency"],"falsifier":"Run the same Rosslyn scenario but insert a single building blockage on the direct path between the target and one of the sequentially selected sensing receivers, then recompute the trace of the CRLB; if it exceeds the sensing threshold $\\eta$ for a nontrivial fraction of target positions, the scheme's sensing guarantee is conditional on LOS.","tokens_in":7607,"feed_emoji":"📡","tokens_out":12220,"duration_ms":97739,"temperature":0.7,"pith_summary":"The paper asks which access points in a cell-free massive MIMO network should act as integrated sensing and communication (ISAC) transmitters, which should only receive the reflected sensing signal, and which should be switched off. It sets up a joint AP mode selection and power allocation problem that minimizes the number of active APs under two constraints: each user's downlink SINR must stay above a threshold, and the trace of the Cramér-Rao lower bound (CRLB) for target position estimation must stay below a threshold. Because the original combinatorial problem is too large, the paper proposes three suboptimal algorithms: alternating optimization, a sequential scheme, and a heuristic. Numerical results show that the sequential scheme, which first selects ISAC transmitters using only the communication constraints and then assigns sensing receivers to meet the sensing constraint, delivers the best performance-complexity balance: it matches the alternation scheme's active-AP count (about half the heuristic's) while running roughly eight times faster.","feed_headline":"Sequential AP selection halves active APs and runs eight times faster","feed_subtitle":"Energy savings come from switching off half the access points without losing communication or sensing quality.","key_machinery":"The load-bearing object is the trace of the Cramér-Rao lower bound for target position, written as a bilinear function of the binary mode vectors $\\mathbf{a}$ (ISAC transmitters) and $\\mathbf{b}$ (sensing receivers) through matrices $G_a$, $G_b$, and $G_c$ that encode AP-target geometry and reflected signal parameters. The sensing constraint is non-convex, but by introducing binary matrices $A = \\mathbf{a}\\mathbf{a}^T$ and $B = \\mathbf{b}\\mathbf{b}^T$ the problem becomes a mixed-integer program that is convex except for the binary variables. The sequential algorithm exploits this two-layer structure: it solves the transmitter subproblem with the communication constraint converted to a second-order cone, and then solves the receiver subproblem for the sensing constraint alone. This decomposition is what makes the scheme run about eight times faster than alternating between transmitter and receiver selection until convergence.","core_discovery":"The paper's central claim is that, for mmWave cell-free massive MIMO-ISAC with a single passive target, AP mode selection can be solved almost as well by a sequential decomposition as by alternating optimization, and much better than by a greedy heuristic. The joint problem of minimizing active AP count under SINR and CRLB constraints is non-convex and combinatorial; the paper reformulates the sensing constraint into a convex (except for binary variables) form using $A = \\mathbf{a}\\mathbf{a}^T$ and $B = \\mathbf{b}\\mathbf{b}^T$, then decomposes the problem. The sequential algorithm first solves the communication-only transmitter selection problem, then fixes those transmitters and selects sensing receivers to satisfy the CRLB constraint. In the simulated Rosslyn, Virginia scenario with 12 APs, this scheme and the alternation scheme both reduce the average number of active APs by 50% relative to the heuristic, while the sequential scheme's runtime is about eight times shorter than alternation. These results support the abstract's conclusion that the sequential scheme achieves the best performance-complexity balance.","pith_inferences":["A natural reading is that the sensing constraint may not need to influence which APs transmit at all; if so, ISAC sensing can be layered onto existing communication planning with only an additional receiver-selection step, making deployment cheaper in practice.","The LOS assumption suggests a concrete stress test: in blockage-heavy mmWave deployments, one could feed a blockage map into the CRLB matrices and let the receiver-selection step favor AP pairs with unblocked paths, a modification the paper's formulation could absorb.","The same convexification and two-stage decomposition could extend to multiple targets or to jointly optimized sensing power, provided the CRLB keeps a bilinear form in the mode vectors.","One measurable prediction follows: in a scenario with wide angular diversity around the target, the receiver-selection step should pick APs that reduce the geometric dilution of precision captured in $G_a$ and $G_b$, rather than just the closest APs."],"forward_implications":["If the sequential scheme is deployed, a cell-free ISAC network can halve its active access points without sacrificing the required communication SINR or sensing CRLB thresholds, cutting the dominant share of energy consumption.","Because the runtime is about eight times shorter than alternation, the mode selection can be repeated frequently enough to follow a moving target or a changing user population.","Since transmitter selection is driven only by communication constraints, the sensing function can be added to an already-running communication network by solving just the receiver-selection subproblem.","The 50% reduction relative to the heuristic shows that optimization-based mode selection matters: a simple channel-gain and proximity greedy rule leaves substantial energy savings on the table."],"supporting_citations":[{"why":"Supplies the CRLB-based power allocation strategy for distributed radar target localization that the paper's sensing constraint (13c) is built on.","marker":"[12]"},{"why":"Establishes the multi-static target detection and power allocation setup for ISAC in cell-free massive MIMO that this work extends with AP mode selection.","marker":"[3]"},{"why":"Provides the binary-variable transformation and convexification technique used to handle the non-convex AP activation constraints.","marker":"[4]"},{"why":"Supports the premise that AP activation, not transmit power, dominates energy consumption, motivating the active-AP count as the objective.","marker":"[6]"},{"why":"The prior BS mode selection work in cooperative cell-free ISAC that this paper contrasts with and extends by adding passive target tracking and CRLB constraints.","marker":"[7]"},{"why":"Supplies the SINR and spectral efficiency expressions (Corr. 6.3 and Sec. 7.1.2) used to write the user communication constraint.","marker":"[10]"},{"why":"Provides the site-specific ray tracing channel model used for the Rosslyn, Virginia numerical evaluation.","marker":"[9]"},{"why":"Gives the modified local partial MMSE precoder used to compute the effective channel statistics in the SINR expression.","marker":"[11]"}],"fun_headline_variants":["Sequential AP selection halves active APs, runs eight times faster","Half the APs, 8x faster: sequential AP selection for mmWave ISAC","Sequential AP mode selection beats alternating on speed and AP use","In mmWave ISAC, sequential AP selection halves AP count and speeds up 8x","AP selection: sequential scheme halves active APs, 8x faster than alternation"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The sensing model assumes an unobstructed line-of-sight path between every candidate access point and the target, so if a blocker interrupts that path the CRLB constraint no longer describes the actual estimation error and the selected AP modes may fail to sense.","fun_headline_variants_meta":{"raw":{"variants":["Sequential AP selection halves active APs, runs eight times faster","Half the APs, 8x faster: sequential AP selection for mmWave ISAC","Sequential AP mode selection beats alternating on speed and AP use","In mmWave ISAC, sequential AP selection halves AP count and speeds up 8x","AP selection: sequential scheme halves active APs, 8x faster than alternation"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000296,"raw_usage":{"total_tokens":1726,"prompt_tokens":958,"completion_tokens":768,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":574,"completion_tokens_details":{"reasoning_tokens":664}},"tokens_in":574,"tokens_out":768,"duration_ms":7798,"temperature":1.0,"reasoning_tokens":664,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-11T23:14:02.400812+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Run the same Rosslyn scenario but insert a single building blockage on the direct path between the target and one of the sequentially selected sensing receivers, then recompute the trace of the CRLB; if it exceeds the sensing threshold $\\eta$ for a nontrivial fraction of target positions, the scheme's sensing guarantee is conditional on LOS.","supporting_citations":[{"cited_title":"Power allocation strategies for target localization in distributed multiple-radar architectures,","cited_arxiv_id":null,"evidence_quote":"Supplies the CRLB-based power allocation strategy for distributed radar target localization that the paper's sensing constraint (13c) is built on."},{"cited_title":"Multi-static target detection and power allocation for integrated sensing and communication in cell-free massive MIMO,","cited_arxiv_id":null,"evidence_quote":"Establishes the multi-static target detection and power allocation setup for ISAC in cell-free massive MIMO that this work extends with AP mode selection."},{"cited_title":"Optimal joint access point placement and resource allocation for indoor mmwave com- munications,","cited_arxiv_id":null,"evidence_quote":"Provides the binary-variable transformation and convexification technique used to handle the non-convex AP activation constraints."},{"cited_title":"Cooperative cell-free ISAC networks: Joint BS mode selection and beamforming design,","cited_arxiv_id":null,"evidence_quote":"The prior BS mode selection work in cooperative cell-free ISAC that this paper contrasts with and extends by adding passive target tracking and CRLB constraints."},{"cited_title":"Evaluation of cell-free millimeter-wave massive MIMO systems based on site-specific ray tracing simulations,","cited_arxiv_id":null,"evidence_quote":"Provides the site-specific ray tracing channel model used for the Rosslyn, Virginia numerical evaluation."}],"review_version":1}