{"id":"f6cad945-8a72-4aa7-b432-96cdb9cfb1cb","arxiv_id":"2509.01223","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":5.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"A super multidimensional scaling variant estimates a rigid body's translation and rotation from distance and angle measurements using only a subset of the edge kernel.","lead":"This paper adapts super multidimensional scaling to rigid body localization, combining distance and angle measurements to estimate a vehicle's translation and rotation. The authors report mean square error close to the Cramer-Rao lower bound in simulations.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"K3 block is unobservable under the declared AA/TT information model; Eq. (14) is not derivable from Eq. (13) without accessing v_AT.","rationale":"The reader identified the same load-bearing concern: Eq. (14) depends on K3, which is quadratic in the unknown anchor-to-target edges, making it unavailable under the declared 'anchor-to-anchor and target-to-target information' claim. This is indeed the most critical issue because it strikes at the central novelty claim of the paper—the assertion that only a minor part of the kernel, based on known AA/TT information, is used. The concern is not merely a wording problem: either the algorithm silently uses AT measurements to build K3, which contradicts the stated information model, or it attempts to infer K3 from estimates, in which case the fixed-point iteration requires a convergence proof that is absent. The paper has other technical issues (e.g., C^{-1} in Eq. (12) is invalid because C is non-square; the pseudo-inverse is needed), but those are more easily fixed. The K3 gap is the crux of the method's justification. It is nevertheless fixable by explicitly stating that the kernel block K3 is computed from noisy AT measurements, which are the raw inputs according to the first sentence of the abstract. With that clarification and a derivation of Eq. (14), the method could be made sound. Thus the reader's CONDITIONAL verdict is appropriate; the concern reinforces it rather than overturning it.","tokens_in":7677,"tokens_out":12517,"duration_ms":127170,"concrete_test":"Independently re-derive Eq. (14) from the rank-1 relation w = u^* v_AT^T (Eq. (13)) without assuming access to v_AT. Determine explicitly whether K3 can be expressed using only v_AA, v_TT, and measured distances/AoA. Then re-implement the algorithm twice: (i) with K3 formed from the measured noisy AT edges, and (ii) with K3 formed from the current estimate of v_AT (i.e., recomputing the left-hand side of (13) from u^(n)). Compare convergence and final MSE against the CRLB in Figs. 2–3. If variant (ii) diverges or degrades, the iterative update is circular and the reported near-CRLB performance cannot be attributed to the described sub-block approach.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim is that an SMDS sub-block can recover the anchor-to-target edge vector, and thus the rigid body translation and rotation, using only a minor of the kernel based on anchor-to-anchor and target-to-target information (Abstract; Conclusion). However, the update Eq. (14) iterates over v_AT using the block column w = [K1; K3; K4^H] defined in Eq. (13). This block column contains K3 = v_AT^* v_AT^T (block (2,2) in Eq. (10)), which is quadratic in the very unknown v_AT the algorithm seeks. If K3 were built only from the known noise-free AA and TT edges, it would be identically zero or undefined, since those edges never involve the AT pairs. Therefore, the declared information model cannot supply K3. The only way to form K3 is to use noisy anchor-to-target distance/AoA measurements directly, contradicting the statement that the relevant minor 'depends on the known noise-free measurements... anchor-to-anchor and target-to-target' (Conclusion). Even if the intended implementation uses AT measurements, this is unstated, and the abstract's 'key point' mischaracterizes the method. Moreover, the update's convergence depends on the spectral structure of K3, and no derivation or convergence proof is given.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper proposes a rigid body localization (RBL) method based on an SMDS-style rank-one complex edge kernel. The method is intended to estimate the anchor-to-target edge vector v_AT, and then the target positions, using only a sub-block of the kernel (Eq. (13)). From v_AT and the known rigid-body shape, translation and rotation are recovered via a standard weighted least-squares fit (Section III.B). Simulation results compare translation and rotation MSE against CRLBs for distance-only and distance-plus-angle variants. The central claim is that the sub-block depends on anchor-to-anchor and target-to-target information only, and that the resulting estimates approach the CRLB.","tokens_in":8077,"tokens_out":8004,"duration_ms":90630,"significance":"If the method worked as claimed, it would be a useful low-complexity alternative to full SMDS for RBL: it would exploit the known structure of the anchors and the rigid body while estimating only the anchor-to-target block. The rank-one kernel construction is correct in the noiseless model, and the rotation/translation least-squares step is a standard and sound component. The paper also provides a comparison against CRLBs from a companion paper, which is appropriate if the measurement model matches. However, the central algorithmic claim is not currently supported: the selected sub-block is not constructed from AA/TT information alone, and Eq. (12) and Eq. (14) contain serious technical gaps. The contribution therefore needs substantial revision before it can be evaluated as stated.","major_comments":[{"comment":"The paper's key claim is that the used minor of the kernel depends only on anchor-to-anchor and target-to-target information. This is not correct. The block column in Eq. (13) contains K1 = v_AA^* v_AT^T, K3 = v_AT^* v_AT^T, and K4^T = v_TT^* v_AT^T. Every one of these blocks is quadratic in the unknown v_AT. If only AA and TT edges are known, K1, K3, and K4 cannot be computed. The initial estimate in Eq. (15) also requires K1 and K4. Thus the algorithm, as described, cannot be executed under the information model stated in the Abstract and Conclusion. The authors must either state explicitly that noisy AT measurements are used to form these blocks, or show how they are obtained from AA/TT information alone. If K3 is formed from the current iterate, the noiseless update becomes circular and the claimed 'key point' must be revised.","section":"Section III.A, Eqs. (13)-(15), and Conclusion"},{"comment":"Eq. (12) writes x_hat = C^{-1} v_hat, but C in Eq. (6) is rectangular: v has P = T(T-1)/2 entries while x has T entries. Even disregarding the rank deficiency from the translation/rotation gauge, C^{-1} does not exist. A pseudo-inverse or a constrained least-squares inversion with anchor coordinates must be specified. The reader cannot reproduce the reconstruction step from the text as written. Citing [21] is not sufficient, because this paper's v_AT-only formulation changes the problem.","section":"Section II.B, Eq. (12)"},{"comment":"Eq. (14) is asserted without derivation. The notation is ambiguous: the numerator appears to be a row-vector product, the denominator is unclear, and no convergence proof is given. The noiseless fixed point may reproduce v_AT by construction if K3 is built from v_AT, but this does not establish convergence from a noisy initial estimate or uniqueness. A rigorous derivation of the update, an explicit statement of which quantities are known, and a convergence analysis are required before the method can be assessed.","section":"Section III.A, Eq. (14)"},{"comment":"The simulations do not describe how the kernel sub-blocks are formed in practice. In particular, it is not stated whether K1, K3, and K4 are computed from noisy AT measurements, from true v_AT, or from the current iterates. This makes it impossible to distinguish a genuine estimator from one that uses oracle knowledge of the unknown edges. Please provide a complete algorithmic specification, including how each block of Eq. (13) is obtained from the measurements, and ideally release code or pseudocode.","section":"Section IV.B, Figs. 2-3"}],"minor_comments":[{"comment":"There are grammatical errors, e.g., 'an dthe' in the Conclusion. More importantly, the wording 'based only on a set of measurements of the distances, as well as the angles between sensors of the vehicle to the anchor landmark points' seems to describe AT measurements, while the subsequent 'anchor-to-anchor and target-to-target information' describes known geometry. The information model should be stated unambiguously.","section":"Abstract/Conclusion"},{"comment":"The symbol M is used both for the number of anchors and for the total number of edges. This is confusing; use a distinct symbol such as P for the edge count, as introduced in the text.","section":"Section II.B, Eq. (6)"},{"comment":"Figure 2 is labeled RMSE while Eq. (23) and the text define MSE. Also the text says that the distance-only SMDS performs slightly better than full SMDS in the small-range-error regime, but this is not explained; if SMDS is not optimal in that regime, a comment is needed.","section":"Section IV.B"},{"comment":"The notation g'|^d_t and lambda_na is not defined in enough detail to reproduce the CRLB. Since the CRLB is taken from the companion paper [25], either summarize the relevant definitions or include them in an appendix.","section":"Section IV.A, Eqs. (24)-(25)"}],"recommendation":"major_revision","confidential_remarks":"The central algorithmic claim is not supported by the information model stated in the paper. However, the underlying idea may be salvageable if the authors re-frame the method as using AT measurements and provide the missing derivations. If the revision cannot clarify how K3 is constructed without oracle knowledge, the contribution should not be accepted."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"You should know: this paper extends the Turbo MRC-SMDS idea to rigid body localization and shows simulations that track the CRLB, but the written method cannot work on the declared information model. The block column in Eq. (13) includes K3 = v_AT^* v_AT^T, which is quadratic in the unknown edge vector v_AT, and Eq. (14) iterates on that block. If only anchor-to-anchor and target-to-target edges are known, K3 is unavailable. That confirms the stress-test note: the abstract and conclusion mischaracterize what the minor actually needs. The update stands on the unstated premise that noisy anchor-to-target measurements are used to form K1, K3, and K4.\n\nCredit where it is due: applying the SMDS sub-kernel to RBL is a legitimate extension, and the rank-one kernel construction is correct in the noiseless model. The distance+angle variant performs close to the CRLB in the Monte Carlo results, and the distance-only variant beats the MDS baseline. The rigid-body parameter estimation step is standard but handled cleanly. For a subfield that mostly sees incremental estimators, this one has a plausible core.\n\nSoft spots, in proportion: Eq. (14) is asserted without derivation or convergence proof. Eq. (12) writes C^{-1} where C is rectangular, which needs to be a pseudoinverse or least-squares step. The CRLB benchmark is borrowed from the authors' own companion paper [25], so it is not independent. No code or data artifacts are provided, which makes the near-CRLB results hard to verify. The most load-bearing issue is the K3 observability problem: either the information model is wrong or the method as described cannot be implemented. That is fixable—state that AT measurements are used and rewrite the minor-selection story—but as written it is a real internal contradiction.\n\nThe paper is for researchers working on wireless rigid body localization, 6G/V2X, or SMDS variants. They will find a useful algorithmic variant and a clean simulation setup, but they should read carefully because the core update is not yet justified. I would send this to peer review, not desk reject it, but the revision should be substantial: derive Eq. (14), fix the observability claim, add convergence analysis, and provide code or at least detailed reproducibility info.","headline":"A useful SMDS-to-RBL extension whose central update equation depends on the very unknown edges it claims to estimate; worth a conditional review but needs a rewrite.","tokens_in":8480,"tokens_out":3844,"would_cite":false,"duration_ms":45158,"reading_group":"maybe","serious_thinker":"no","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"The paper proposes an SMDS variant that recovers a rigid body's translation and rotation from distance and angle measurements using a sub-block of the complex edge kernel, with simulated MSE close to the CRLB.","keywords":["rigid body localization","super multidimensional scaling","complex edge kernel","angle of arrival","distance measurements","Cramér-Rao lower bound","wireless localization","sensor networks"],"falsifier":"In an 8-anchor, 8-target simulation, withhold all anchor-to-target distance and angle data, form K3 using only products of anchor-to-anchor and target-to-target edges, and run Eq. (14). If the recovered v_AT does not approach the CRLB in MSE, or if K3 is not uniquely determined by the known blocks, the paper's information premise is false.","tokens_in":7632,"feed_emoji":"📐","tokens_out":8326,"duration_ms":88291,"temperature":0.7,"pith_summary":"This paper tries to establish that a moving rigid body can be localized from mixed distance and angle measurements using only a small part of the complex edge kernel of the SMDS algorithm. If true, the position, shape, and orientation of the body can be recovered with less computational and measurement overhead than constructing and truncating the full kernel. The paper's simulations compare the translation and rotation estimates against a prior MDS-based method and against the CRLB, and report that the estimates reach the fundamental limit. The reason this matters is that rigid body localization underlies navigation, collision detection, and vehicle path prediction, where knowing orientation as well as position is necessary.","feed_headline":"A sub-block of the SMDS kernel recovers rigid body pose","feed_subtitle":"Only a minor block of the edge kernel is needed to estimate translation and rotation near the CRLB.","key_machinery":"The central object is the rank-one complex edge kernel K = v^* v^T, whose entries are products of directed complex edges between pairs of nodes. The method uses a three-block minor of K that couples the unknown anchor-to-target edge vector v_AT with the known anchor-to-anchor and target-to-target edge vectors; the fixed-point update of Eq. (14) is the mechanism that recovers v_AT from that minor. The rank-one structure ensures the kernel contains a consistent geometry, while the minor selection is what makes the method cheaper than full SMDS and tailored to the rigid-body measurement setup.","core_discovery":"The central claim is that a rigid body's landmark positions, translation, and rotation can be recovered by solving for the anchor-to-target edge vector v_AT instead of the full edge vector v. The complete SMDS kernel K = v^* v^T is rank one and would normally be truncated by its largest eigenpair; this paper instead selects the three blocks K1, K3, K4 that couple v_AT with the known anchor-to-anchor and target-to-target edges, and turns them into the fixed-point update of Eq. (14). Numerical simulations with 8 anchors and an 8-point rigid body in a 10 m-by-10 m room are intended to show that the resulting estimate of translation and rotation is close to the CRLB and better than a distance-on","pith_inferences":["The authors leave implicit that Eq. (14) needs K3 = v_AT^* v_AT^T, a block quadratic in the very anchor-to-target edges being estimated. If K3 is built from noisy anchor-to-target measurements, part of the reported accuracy may come from those measurements directly rather than from the SMDS minor structure.","A natural ablation would compare the full method, the distance-only variant, and a direct least-squares estimator that uses anchor-to-anchor and target-to-target information without the kernel minor; this would isolate what the SMDS update contributes.","The same sub-block construction should carry over to 3D with a suitable quaternion or rotation-constrained representation, since the rank-one kernel identity itself is dimension-independent.","In settings with sparse target-to-target measurements, the method's dependence on the known v_TT block would make the update degrade; testing the iteration against the fraction of missing target-to-target edges would map its operating range."],"forward_implications":["If the central claim holds, an SMDS-based estimator can output the rigid body's translation and rotation at CRLB-level accuracy while only forming a sub-block of the edge kernel, reducing complexity compared with full SMDS.","The method offers a unified treatment of distance and angle data: when angle measurements are unavailable, a preliminary MDS step supplies angles, and the same kernel update then applies.","The iterative structure of Eq. (14) means the method can be initialized from anchor-to-anchor and target-to-target data alone and refined as anchor-to-target information enters.","The comparison against the CRLB positions the method as practically efficient, since its MSE tracks the theoretical lower bound over the simulated range-error interval."],"supporting_citations":[{"why":"Supplies the SMDS algorithm that the paper modifies, in which a rank-one complex edge kernel is built from distance and angle information and truncated to recover node coordinates.","marker":"[7]"},{"why":"Supplies the Turbo MRC-SMDS idea of iterating on only a minor of the complex edge kernel, which the proposed rigid-body update adapts.","marker":"[21]"},{"why":"Defines the MDS-based rigid body localization baseline and the weighted least-squares pose-estimation formulation used in the second stage.","marker":"[19]"},{"why":"Provides the Fisher-information/CRLB construction for rigid body localization against which the simulations are compared.","marker":"[25]"},{"why":"Supplies the closed-form SVD solution for the rotation estimate in the least-squares step.","marker":"[22]"}],"fun_headline_variants":["Rigid body localization from a minor SMDS kernel block","Sub-kernel SMDS recovers pose to CRLB accuracy","Only part of the edge kernel needed for full pose","Using a kernel corner to estimate translation and rotation","SMDS block-wise update yields pose near the CRLB"],"cache_read_input_tokens":2688,"weakest_assumption_plain":"The load-bearing premise is that the sub-block of the kernel needed to update the anchor-to-target edges can be assembled from anchor-to-anchor and target-to-target measurements alone; if that sub-block actually requires the unknown anchor-to-target measurements themselves, the iteration is circular.","fun_headline_variants_meta":{"raw":{"variants":["Rigid body localization from a minor SMDS kernel block","Sub-kernel SMDS recovers pose to CRLB accuracy","Only part of the edge kernel needed for full pose","Using a kernel corner to estimate translation and rotation","SMDS block-wise update yields pose near the CRLB"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000216,"raw_usage":{"total_tokens":1222,"prompt_tokens":652,"completion_tokens":570,"prompt_tokens_details":{"cached_tokens":256},"prompt_cache_hit_tokens":256,"prompt_cache_miss_tokens":396,"completion_tokens_details":{"reasoning_tokens":490}},"tokens_in":396,"tokens_out":570,"duration_ms":7319,"temperature":1.0,"reasoning_tokens":490,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-05T12:45:56.494140+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"In an 8-anchor, 8-target simulation, withhold all anchor-to-target distance and angle data, form K3 using only products of anchor-to-anchor and target-to-target edges, and run Eq. (14). If the recovered v_AT does not approach the CRLB in MSE, or if K3 is not uniquely determined by the known blocks, the paper's information premise is false.","supporting_citations":[{"cited_title":"Super MDS: Source location from distance and angle information,","cited_arxiv_id":null,"evidence_quote":"Supplies the SMDS algorithm that the paper modifies, in which a rank-one complex edge kernel is built from distance and angle information and truncated to recover node coordinates."},{"cited_title":"Turbo MRC-SMDS: Low-complexity co- operative localization from hybrid information,","cited_arxiv_id":null,"evidence_quote":"Supplies the Turbo MRC-SMDS idea of iterating on only a minor of the complex edge kernel, which the proposed rigid-body update adapts."},{"cited_title":"Least-squares estimation of transformation parameters between two point patterns,","cited_arxiv_id":null,"evidence_quote":"Supplies the closed-form SVD solution for the rotation estimate in the least-squares step."}],"review_version":1}