{"id":"dafa4352-907e-44f6-9d8e-390c3275b0fe","arxiv_id":"2507.17136","paper_version":1,"verdict":"REJECT","confidence":"MODERATE","novelty_score":3.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":5,"one_line_summary":"A hydraulic curtain wall installation robot is modeled with D-H kinematics and a Stribeck friction model, and its 18 minimal inertial parameters are fitted by least squares, yielding in-sample torque residuals below 0.4 Nm.","lead":"This paper applies standard robot dynamics identification to a new hydraulically driven curtain wall installation arm, fitting friction and inertial parameters to measured torques. The reported sub-0.4 Nm torque residuals likely measure in-sample fit, not independent prediction, so the accuracy claim is overstated.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Validation is in-sample: no separate trajectory is described, so Table 7 residuals are fitted residuals and cannot support the sub-0.4 Nm predictive claim.","rationale":"I looked for the weakest load-bearing link between evidence and conclusion. The conclusion says experimental validation confirms high precision. The only quantitative evidence is Table 7. To confirm predictive accuracy, those residuals must be computed on data not used in the fit. The paper's own text does not say that; in fact the phrase 'actual data were input into the observation matrix' is naturally read as the same identification run. The reader identified this same weakness. I agree. I also noticed Eq. (28) is a normalized RMS error, not a standard deviation in Nm, which would independently invalidate the literal claim even if validation were out-of-sample. But the in-sample issue is more fundamental because it attacks whether the experiment can confirm anything. The paper has real engineering value (a working curtain wall installation arm, a full pipeline, and plausible parameter curves), and none of this implies misconduct; the fix is reporting an out-of-sample validation with a correct metric. Because the current manuscript lacks that, the rejection stands as presented, and no additional concern is needed.","tokens_in":10241,"tokens_out":3766,"duration_ms":43053,"concrete_test":"Perform an explicit out-of-sample validation. Run a new trajectory with different Fourier coefficients (or a manual trajectory) within the Table 2 joint limits, record pressures and displacements, compute theoretical torques from the already identified parameters in Tables 5 and 6 with no refitting, and report both the actual RMSE in Nm and the normalized residual of Eq. (28). If the RMSE on this hold-out trajectory remains below 0.4 Nm, the central claim is supported. If only one run exists, split the recorded time series in half: fit on the first half and evaluate on the second half. Additionally, re-derive Eq. (28) with the 1/K factor so the reported metric has units of Nm.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim requires that the identified parameters predict joint torques on trajectories not used for fitting. Section V.C describes only that 'actual position, velocity, and acceleration data were input into the observation matrix' and compared with measured torques; it never states that these data come from a different trajectory than the one used to build the least-squares estimate in Eq. (24). Because Eq. (24) minimizes the squared residual over the very data collected for identification, small in-sample residuals are expected even for a model with poor generalization. With 18 independent inertia parameters and six friction curves fitted per joint, the reported residuals near 0.2–0.4 Nm do not by themselves evidence high precision. A secondary unit problem strengthens the concern: Eq. (28) as printed is normalized by sum_k tau_ik^2, making xi dimensionless (a relative RMS error), while Table 7 gives units of Nm; so either the formula is misprinted or the reported numbers are not residual standard deviations in Nm. Both problems target the same load-bearing validation claim.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The manuscript presents a holistic dynamic parameter identification framework for a six-degree-of-freedom hydraulically driven curtain wall installation robot. The authors construct a Denavit-Hartenberg model from measured structural parameters, integrate a linearized Stribeck friction model for hydraulic cylinders, design Fourier-series excitation trajectories under joint constraints, and identify cylinder friction and rigid-body minimal inertial parameters through a hierarchical least-squares procedure. Full-scale experiments are reported, and the paper claims residual standard deviations below 0.4 Nm between theoretical and measured joint torques (Table 7), which is presented as evidence of high-precision identification.","tokens_in":10445,"tokens_out":5187,"duration_ms":53920,"significance":"If the validation claim were properly supported, the work would be practically valuable: it provides a complete pipeline for identifying physically interpretable dynamic parameters for a non-standard hydraulically actuated manipulator, which is needed for model-based control and simulation. The paper has strengths: it performs separate cylinder friction identification before rigid-body identification, uses a minimal parameter set via SymPyBotics, and reports full-scale experiments on all six joints. However, the central validation is currently in-sample, the stated error metric is mis-specified, and several unit and consistency issues affect the quantitative claims.","major_comments":[{"comment":"Section V.C states only that \"actual position, velocity, and acceleration data were input into the observation matrix\" and compared with measured torques; it does not state that these data come from a trajectory separate from the one used to form H and Γ in Eq. (24). Since Eq. (24) minimizes ∥ρ∥² over exactly the data used for validation, Table 7 reports fitting residuals, not out-of-sample prediction errors. The sub-0.4 Nm values therefore do not, by themselves, support the abstract's \"high-precision\" claim. Please add a hold-out validation trajectory (or a cross-validation analysis) and report residuals on data not used in the least-squares estimate.","section":"V.C, Eq. (24)"},{"comment":"Equation (28) defines ξ_RSD as sqrt(Σ(τ_ik−τ′_ik)² / Σ τ_ik²), which is a dimensionless normalized RMS error, not a residual standard deviation in Nm. Table 7 labels the column ξ_RSD(Nm) and the conclusion states \"below 0.4 Nm\". Either the formula is missing a factor of 1/K and the square root of the mean squared residual, or the reported numbers are relative errors (0.18–0.38, i.e., 18–38%), which would not support the stated accuracy. The authors should correct the formula, the units, and the numerical claim accordingly.","section":"Eq. (28), Table 7"},{"comment":"The recursive least-squares gain in Eq. (11) is written K_k = P_{k−1}λ_k (λ_k^T P_{k−1}λ_k − 1)^{-1}. In standard RLS the denominator is λ_k^T P_{k−1}λ_k + 1 (or a variant with a forgetting factor), and the minus sign can make the gain singular or negative for small λ_k^T P_{k−1}λ_k. As the recursive update is the basis for the cylinder friction identification in Section IV.A, the authors should verify the equation and, if it is a typo, correct it and state whether the implemented algorithm uses the printed form.","section":"Eq. (11)"},{"comment":"There are unit inconsistencies in the trajectory data. In Table 3, a_i^l, b_i^l and q_i,0 are given without units; if they are radians or rad/s, values such as a_1^1 = −8.996 and b_1^1 = 8.600 produce joint excursions far outside the limits in Table 2 (e.g., joint 1 limit is [−0.0523, 1.0472] rad), while if they are degrees, Eq. (27) and the frequency terms in Eq. (25) are not in consistent units. Additionally, the nominal θ_i values in Table 1 (130° for joint 2, −60° for joint 3) lie outside the corresponding limits in Table 2. Please clarify the units and verify the consistency of the numerical values.","section":"Tables 2, 3; Eq. (25)"}],"minor_comments":[{"comment":"The name \"Strubeck\" is used in Section IV.A and in the Figure 7 caption; the correct spelling is \"Stribeck\".","section":"Throughout"},{"comment":"The friction parameters f_c, f_v, f_s in Eq. (4) appear as forces in the hydraulic cylinder force balance Eq. (5), but Table 5 lists units of N·m; clarify whether these are forces or joint torques.","section":"Table 5"},{"comment":"Equation (16), the parallel axis theorem, is garbled: the coordinate-frame notation is inconsistent and the equation contains duplicated or misplaced terms. Please rewrite it cleanly.","section":"Eq. (16)"},{"comment":"Reference [26] is cited as the source of the SymPyBotics tool package, but the reference is a paper on autonomous construction robots; provide the correct citation for SymPyBotics.","section":"References"},{"comment":"The text preceding Eq. (24) says \"Equation (24) can be expressed in the form of a least squares estimate,\" but Eq. (24) is the least-squares estimate itself; please rephrase to avoid circular wording.","section":"Section IV.B"}],"recommendation":"major_revision","confidential_remarks":"The paper is within scope for the journal. The main obstacle is the validation methodology: the reported sub-0.4 Nm residuals are fitting residuals unless a separate validation trajectory is added. This is fixable with additional analysis or a new experiment, so I recommend major revision rather than rejection. Please also check the reference list carefully, as at least one citation does not match its claimed source."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Dear colleague,\n\nQuick read of Liu et al. The headline: this is a competent system-identification writeup for a real hydraulically driven curtain-wall robot, but the Table 7 validation is almost certainly in-sample, so the sub-0.4 Nm claim does not mean what the abstract says. The paper is not a wash; it needs an out-of-sample validation pass before the headline is supportable.\n\nWhat is new: the hardware, plus the specific two-stage identification pipeline — separate Stribeck friction identification on the hydraulic cylinders, then least-squares identification of the 18-parameter minimal rigid-body set with Fourier excitation. The Stribeck parameter tables and minimal inertia parameters are new for this arm. Separating cylinder friction before rigid-body identification is a sensible engineering move. The methods are standard, but they are applied to a platform that has not been modeled in the cited literature.\n\nWhere it falls short. Section V.C says the actual position, velocity, and acceleration data were input into the observation matrix to compute theoretical torques, but it never says these data come from a trajectory different from the one used to fit Eq. (24). With 18 inertia parameters plus six friction curves, in-sample residuals near 0.2–0.4 Nm are expected even if the model generalizes poorly. The stress-test note is right: the RSD formula in Eq. (28) is a normalized relative error (dimensionless), while Table 7 reports Nm; either the formula is misprinted or the units are wrong. Both problems land on the same central claim. I also see a likely sign error in the recursive gain in Eq. (11): the denominator should be λ_k^T P_{k-1} λ_k + 1, not minus 1; as written the gain can blow up. That is a minor issue because the paper ultimately uses batch least squares, but it is still sloppy. The friction parameters in Table 5 are labeled N·m, which is odd for cylinder forces; could be a typo, but it matters for anyone reusing them.\n\nThe citation pattern is fine; the relevant dynamic identification literature is cited. The contribution is application-specific, and that is acceptable for the right venue. If the authors add a proper out-of-sample validation — a separate trajectory, or a consistently held-out subset, with a correctly dimensioned residual metric — this becomes a solid engineering paper. As it stands, the central high-precision claim is not supported.\n\nMy recommendation: send it to peer review if the venue publishes engineering applications, with a clear request for out-of-sample validation and corrected formulas. Do not accept the current version.","headline":"Competent application of standard dynamic identification to a new hydraulic curtain-wall arm, but the headline sub-0.4 Nm validation is in-sample and the RSD formula is mislabeled, so the central claim needs out-of-sample evidence.","tokens_in":10987,"tokens_out":2525,"would_cite":false,"duration_ms":28366,"reading_group":"no","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"The paper claims its composite rigid-body plus hydraulic-cylinder dynamic model, identified from optimized Fourier excitation trajectories by hierarchical least squares, reproduces the measured joint torques of a curtain wall installation…","keywords":["dynamic parameter identification","curtain wall installation robot","hydraulic cylinder dynamics","Stribeck friction model","Fourier excitation trajectories","least squares estimation","minimal inertia parameter set","residual standard deviation"],"falsifier":"Collect a fresh dataset from a trajectory whose Fourier coefficients differ from those in Table 3 (or from a manually guided motion), compute the joint torques predicted by the identified parameters, and compare them with measured torques; if the residual standard deviations exceed 0.4 Nm, the high-precision claim would be contradicted.","tokens_in":9991,"feed_emoji":"🏗️","tokens_out":7879,"duration_ms":72842,"temperature":0.7,"pith_summary":"The paper sets out to make a hydraulically driven robotic arm for curtain wall installation usable for model-based control by identifying its dynamic parameters from experiments. It builds a Denavit–Hartenberg rigid-body model of the arm and couples it with a hydraulic-cylinder dynamics model whose friction is described by a linearized Stribeck term. Fourier-series trajectories are designed to excite the minimal 18-parameter inertia set while respecting joint limits, and a two-stage least-squares procedure first identifies friction, then inertia parameters. The claimed outcome is a model whose predicted joint torques match sensor measurements with residual standard deviations below 0.4 Nm for all six joints. This matters because a trustworthy dynamic model is what would let controllers compensate for the robot's own dynamics and lets designers simulate the arm before building it.","feed_headline":"Predicts curtain wall robot joint torques within 0.4 Nm error","feed_subtitle":"A two-stage least-squares identification of rigid-body and Stribeck friction parameters delivers the sub-0.4 Nm match.","key_machinery":"The load-bearing object is the composite parametric dynamic model $\\tau = Y(q,\\dot q,\\ddot q) X$ for the arm, coupled with the hydraulic-cylinder force balance $m\\ddot x + c\\dot x + Kx + f_c\\,\\mathrm{sgn}(\\dot x) + f_v\\dot x + f_s \\dot x^{1/3} = p_1A_1 - p_2A_2 - F$, in which the Stribeck friction model, which captures static, transition, and fluid-dynamic lubrication zones, is linearized so that all unknowns enter affinely. The cylinder and arm parameters are identified in two stages: first the friction and stiffness parameters of each cylinder from pressure and displacement data, then the 18-parameter minimal inertia set of the arm by least squares with the observation matrix built from Fourier-series excitation trajectories that respect joint limits. The reduction from 78 to 18 parameters, generated with a symbolic dynamics toolbox, is what makes the second-stage least squares well-conditioned.","core_discovery":"On the paper's own terms, the central discovery is that the proposed hierarchical identification framework delivers high-precision dynamic parameters for this hydraulically actuated arm on a real platform. The composite model treats the hydraulic cylinder force balance and the rigid-body arm dynamics as one parametric system: the cylinder model contributes mass, stiffness, Coulomb, viscous, and Stribeck friction parameters, and the arm model is reduced from 78 to 18 independent inertia parameters using a symbolic minimal-set derivation. Excitation trajectories are Fourier series whose coefficients are selected to keep the motion within joint position, velocity, and acceleration bounds and to minimize the condition number of the observation matrix. Least-squares estimates from the measured data yield torque predictions whose residuals, expressed as a normalized standard deviation, stay under 0.4 Nm for every joint; the paper reads this as confirmation that the identified parameters describe the robot's dynamics accurately enough for control and simulation.","pith_inferences":["If the validation used the same trajectories as identification, the residual standard deviations only attest to fitting quality; a hold-out trajectory test would reveal whether the model predicts novel motions.","The pipeline could transfer to other hydraulically driven construction equipment, such as excavators or aerial lifts, since the cylinder model and excitation design do not depend on the specific arm geometry.","Because the paper tabulates piston masses, friction parameters, and the 18 inertia parameters, a reader could reconstruct a complete simulation model of the arm without access to the hardware.","If one wanted to push accuracy further, the linearized Stribeck term $\\dot x^{1/3}$ could be replaced by the full exponential Stribeck or a LuGre model and the change in residuals measured; the paper's reported residuals provide a baseline."],"forward_implications":["Model-based controllers can use the identified parameters for feedforward torque compensation and computed-torque control of this curtain wall robot.","The reported sub-0.4 Nm residuals indicate the minimal 18-parameter model captures the dominant rigid-body and friction dynamics of the arm, supporting its use in simulation.","The separation of cylinder friction identification from rigid-body inertia identification, followed by joint calibration, provides a template for other hydraulic manipulators.","The Fourier excitation trajectory design with joint constraints can be reused for re-identifying parameters after maintenance or payload changes."],"supporting_citations":[{"why":"Cited for the hydraulic cylinder force balance equation, which is the starting point of the friction parameter identification.","marker":"[23]"},{"why":"Provides the Stribeck friction model used to describe the cylinder's static, transition, and fluid-dynamic friction zones.","marker":"[24]"},{"why":"Supplies the linearized Stribeck parameterization that makes the friction terms linear in the unknown parameters.","marker":"[25]"},{"why":"Cited as the source of the symbolic dynamics tool used to derive the minimal 18-parameter inertia set.","marker":"[26]"}],"fun_headline_variants":["Hydraulic curtain-wall robot: torque predictions within 0.4 Nm","Sub-0.4 Nm torque prediction for hydraulic robotic arm","Curtain wall robot arm identified to 0.4 Nm precision","Hierarchical least squares sharpens hydraulic arm to 0.4 Nm","Stribeck friction model improves hydraulic arm torque to 0.4 Nm"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The validation residuals are presented without stating whether the test trajectories were the same ones used for identification, so the sub-0.4 Nm numbers may reflect how well the least-squares fit matches its own training data rather than how well the model predicts new motions.","fun_headline_variants_meta":{"raw":{"variants":["Hydraulic curtain-wall robot: torque predictions within 0.4 Nm","Sub-0.4 Nm torque prediction for hydraulic robotic arm","Curtain wall robot arm identified to 0.4 Nm precision","Hierarchical least squares sharpens hydraulic arm to 0.4 Nm","Stribeck friction model improves hydraulic arm torque to 0.4 Nm"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000952,"raw_usage":{"total_tokens":4065,"prompt_tokens":952,"completion_tokens":3113,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":568,"completion_tokens_details":{"reasoning_tokens":3014}},"tokens_in":568,"tokens_out":3113,"duration_ms":23221,"temperature":1.0,"reasoning_tokens":3014,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-06T14:57:20.150994+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Collect a fresh dataset from a trajectory whose Fourier coefficients differ from those in Table 3 (or from a manually guided motion), compute the joint torques predicted by the identified parameters, and compare them with measured torques; if the residual standard deviations exceed 0.4 Nm, the high-precision claim would be contradicted.","supporting_citations":[{"cited_title":"Review and comparison of dry friction force models,","cited_arxiv_id":null,"evidence_quote":"Cited for the hydraulic cylinder force balance equation, which is the starting point of the friction parameter identification."},{"cited_title":"Friction and rigid body identification of robot dynamics,","cited_arxiv_id":null,"evidence_quote":"Provides the Stribeck friction model used to describe the cylinder's static, transition, and fluid-dynamic friction zones."},{"cited_title":"An improved parameter identification algorithm for the friction model of electro-hydraulic servo systems,","cited_arxiv_id":null,"evidence_quote":"Supplies the linearized Stribeck parameterization that makes the friction terms linear in the unknown parameters."}],"review_version":1}