{"id":"a8042e4d-03b2-46b6-9b1f-246419073d68","arxiv_id":"2502.06362","paper_version":1,"verdict":"CONDITIONAL","confidence":"HIGH","novelty_score":4.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":3,"one_line_summary":"A three-tendon origami manipulator uses conductive threads as both actuators and resistance-based length sensors, feeding reconstructed tendon lengths into a forward kinematic model to estimate its end-effector path without external vision.","lead":"This paper builds a bendable origami arm whose metal threads act as both the pulling tendons and the electrical sensors that report how the arm is bent. The arm can therefore report its own shape without cameras, which could help origami robots operate in tight, dark, or cluttered spaces.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The published resistance-to-length calibration in Sec. III-B is unphysical as written, and the end-effector reconstruction is supported only by a qualitative 'reasonable agreement' with no error metric.","rationale":"I read the paper in good faith. The hardware concept is plausible and the paper includes useful supporting measurements: the free-thread resistance is linear in length, cyclic loading up to 10 N produces only 3.9 mΩ variation, the sensor output is repeatable over 20 cycles, and a separate dataset with increasing amplitude is used to validate the length reconstruction, which reduces circularity. However, the central claim is a specific quantitative one: that resistance readings, converted by a calibrated polynomial, plus PCC kinematics reconstruct the end-effector path. That claim has two unaddressed problems. First, the calibration equation is internally inconsistent with the stated resistance range: the coefficients listed in Sec. III-B produce negative lengths for the reported Ri values, so as written the pipeline cannot even be executed. This may be a typographical or normalization error, but the manuscript does not define the normalization, making the result unreproducible. Second, the validation against OptiTrack is only qualitative; no error metric is reported, so 'reasonable agreement' cannot be checked. The reader correctly identified the resistance-to-length mapping as load-bearing and the absence of error analysis as a weakness, but the paper's calibration was in fact performed on the assembled manipulator (Fig. 4C), so the concern is less about force independence inside the structure and more about the unphysical published mapping and missing endpoint error. These issues do not prove the approach false; they mean the claim is not yet substantiated in a checkable form. I therefore keep the reader's CONDITIONAL verdict unchanged, with the added condition that the calibration mapping be corrected and quantified, and that endpoint reconstruction error be reported.","tokens_in":7593,"tokens_out":6857,"duration_ms":59419,"concrete_test":"Request the raw resistance, tracked-length, and OptiTrack trajectory datasets, and have the authors re-specify the calibration equation with explicit variable definitions (raw Ω or normalized ΔR/R0 with the normalization formula). Evaluate the published polynomial over the full resistance range used in Sec. III-B: if it produces negative lengths, the mapping is invalid and the reconstructed path in Fig. 5B must be recomputed. Independently, compute the per-axis RMSE and maximum error between the reconstructed and OptiTrack end-effector paths; if the mean error exceeds roughly 10% of the manipulator length, the phrase 'reasonable agreement' should be replaced with a quantitative, bounded error claim.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The path reconstruction in Fig. 5B depends entirely on converting the measured tendon resistance Ri into a tendon length li via the third-order polynomial in Sec. III-B. As printed, with Ri in the reported 9.5–10.5 Ω range, li = 0.102 − 0.172Ri − 0.205Ri^2 − 0.173Ri^3 evaluates to large negative values (e.g., about −195 at 10 Ω), so either the coefficients, the independent variable (raw ohms vs. a normalized resistance), or the stated equations are misreported. Because these li values are the inputs to the forward kinematic model in Eq. (3), the central claim that the sensor readout reconstructs the end-effector path cannot be independently reproduced from the manuscript. The only reported validation is the sentence \"the reconstructed path shows a reasonable agreement\" with no RMSE, no maximum error, and no raw data or code. The reader's concern about drift or contact effects is secondary: until the calibration mapping is correctly specified and the reconstruction error is quantified, the proprioceptive-accuracy claim is unfalsifiable.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The manuscript reports a tendon-driven origami continuum manipulator whose stainless-steel conductive threads serve both as actuation tendons and as resistive proprioceptive sensors. A Wheatstone bridge measures each tendon's resistance, a third-order polynomial maps measured resistance to tendon length, and a piecewise-constant-curvature forward kinematic model reconstructs the manipulator configuration and end-effector path. The authors characterize axial and bending stiffness, thread strength and cyclic durability, sensor repeatability over 20 cycles, and compare a reconstructed closed path against OptiTrack motion capture, claiming 'reasonable agreement.' The paper concludes that this embodied sensing platform can provide proprioceptive feedback for closed-loop control without external vision.","tokens_in":7749,"tokens_out":3919,"duration_ms":33100,"significance":"If the sensing–kinematics pipeline is validated quantitatively, the contribution is useful for soft and continuum origami robots: it removes the need for external vision or motor-side encoders and preserves the flexibility of the origami backbone. The paper has several solid experimental components: mechanical characterization of the origami structure, tensile and cyclic testing of the conductive thread, a force-independence check in Sec. III-B, and repeatability data over 20 cycles. However, the central accuracy claim currently rests on a calibration equation that is unphysical as printed and on a qualitative path comparison, so the significance cannot be fully assessed until these issues are corrected.","major_comments":[{"comment":"The resistance-to-length mapping is stated as li = 0.102 − 0.172Ri − 0.205Ri^2 − 0.173Ri^3 with parameters a = 0.102, b = −0.172, c = −0.205, d = −0.173. With Ri in the reported range 9.5–10.5 Ω, this expression evaluates to roughly −170 to −195 mm, not the positive tendon lengths used in the forward kinematics or the reported RMSE of 1.25 mm. As printed, the mapping cannot be the one used to produce Fig. 4C or Fig. 5. Please provide the correct coefficients, clarify whether Ri is a normalized resistance rather than raw ohms, and report the normalization range. Until this is fixed, the sensing-to-kinematics pipeline cannot be independently reproduced.","section":"Section III-B"},{"comment":"The central claim that proprioceptive reconstruction yields the end-effector path rests solely on the sentence 'The reconstructed path shows a reasonable agreement with the tracked positions' with no quantitative error metric. Please report the reconstruction error over the full trajectory (e.g., RMSE, maximum error, and normalized path error), specify the number of trials, and, if possible, provide per-axis errors. Without these numbers the central accuracy claim is not falsifiable.","section":"Section III-C, Fig. 5B"},{"comment":"The force-independence check shown in Fig. 4B is performed on a free thread stretched between tensile-testing grippers, not inside the folded origami structure under bending, contact, and friction. Since tendon routing in the assembled manipulator can introduce contact pressure, localized bending, and friction that may alter the resistance–length relationship, the conclusion that actuation forces do not affect resistance needs verification in the assembled manipulator or an explicit statement of this as a limitation. This is load-bearing because the calibration is reused in the final path reconstruction.","section":"Section III-B"},{"comment":"The calibration described in Sec. III-B is a single fitted polynomial across all tendons, but the paper does not report the number of calibration samples, per-tendon residuals, or the exact normalization procedure. Additionally, the calibrated length RMSE (1.25 mm) is not propagated through Eq. (3) to give an end-effector uncertainty. Please provide the calibration dataset details and, if feasible, propagate the length uncertainty through the forward kinematic model so that the reconstructed path has an associated confidence interval.","section":"Section III-B"}],"minor_comments":[{"comment":"In Eq. (2), the sums over j = 1..n appear inside the vector with arguments sin((2j−1)θ/2n) and cos((2j−1)θ/2n); please make the indexing and parentheses unambiguous and define n explicitly in the text.","section":"Eq. (2)"},{"comment":"The text in Sec. III-B says the repeatability test in Fig. 4D was performed on tendon 1, while the figure caption says 'Tendon 2'; please harmonize this discrepancy.","section":"Fig. 4D"},{"comment":"The conclusion states that the resistance change is 'small (1 − 2 Ω)', but the ranges reported in Sec. III-B are 9.65–10.2 Ω, 9.5–10.5 Ω, and 9.65–10.35 Ω, corresponding to variations of 0.55–1.0 Ω; please reconcile these numbers.","section":"Section IV"}],"recommendation":"major_revision","confidential_remarks":"The core idea is promising and the mechanical characterization is useful, but the printed calibration equation is not internally consistent with the reported resistance range, and the final validation is qualitative. If the authors can provide the corrected mapping (including normalization) and quantitative path-reconstruction errors, I would be willing to reconsider the manuscript."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Quick take: this is a legitimately useful integration—conductive threads doing double duty as tendons and resistance sensors on a Yoshimura-pattern manipulator—and the mechanical characterization is careful. But the paper as submitted has a load-bearing arithmetic problem. In Sec. III-B, the third-order polynomial li = 0.102 − 0.172Ri − 0.205Ri^2 − 0.173Ri^3, with Ri in the reported 9.5–10.5 ohm range, gives large negative lengths (around −195 at 10 ohm). Either the coefficients, the variable (raw vs. normalized resistance), or the printed equation is wrong. Since those li values are the inputs to the forward kinematics in Eq. (3), the central claim—reconstructing the end-effector path from resistance readings—cannot be independently checked from the manuscript. That's the main thing to fix.\n\nWhat the paper does well: the design is sound, the flat-prismatic-ring modification to the Yoshimura pattern is a sensible way to stabilize contraction, and the force-independence check on the conductive thread (Fig. 4B) is a useful negative result. The repeatability data over 20 cycles and the increasing-amplitude reconstruction in Fig. 5A are reasonable supporting evidence. The authors also honestly note the small resistance range (1–2 ohm) and fabrication sensitivity in the conclusion, which are real limitations.\n\nThe soft spots beyond the misreported polynomial: the final validation in Fig. 5B is a single sentence—'reasonable agreement'—with no RMSE, no maximum error, no error bars, and no raw data or code. For a sensing paper, that's thin. The force-independence test was done on a free thread, not inside the folded structure under contact and bending, so drift under real operating conditions is not addressed. The PCC/no-torsion assumption is reasonable for this geometry but untested against stereo vision or a model. These are fixable with more reporting rather than new theory.\n\nNet: the concept is a genuine contribution for vision-free proprioception in origami manipulators, and the flaws are mostly presentational and quantitative. Send it to peer review, but the referee should insist on a corrected, physically consistent calibration equation, quantitative reconstruction error, and ideally a data/code release. Right now the headline result is not reproducible from the text.","headline":"Clever integration of conductive-thread sensing into an origami manipulator, but the printed calibration equation cannot reproduce the reported lengths and the final path validation is only qualitative.","tokens_in":8322,"tokens_out":2095,"would_cite":false,"duration_ms":18696,"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":"Conductive tendons let an origami arm sense its own bending without cameras.","keywords":["origami manipulator","continuum robot","proprioceptive sensing","conductive thread","tendon-driven actuation","Wheatstone bridge","forward kinematics","soft robotics"],"falsifier":"Run the three-tendon phase-shifted cyclic protocol from Fig. 5B, compare the reconstructed end-effector positions to OptiTrack, and compute the mean and maximum error over cycles; if the maximum error grows with cycle number or exceeds a few centimeters without any vision correction, the resistance--length mapping is not stable under coupled loading.","tokens_in":7354,"feed_emoji":"🦾","tokens_out":4122,"duration_ms":35518,"temperature":0.7,"pith_summary":"This paper establishes that the tendons that actuate a flexible origami manipulator can also serve as its position sensors. Because each tendon's resistance grows with its active length, measuring resistance with a Wheatstone bridge yields a reading of how much that side of the arm has contracted; a polynomial fit converts those readings into tendon lengths, and a piecewise-constant-curvature forward model turns the three lengths into an estimated end-effector position. The reconstructed path agrees reasonably with a motion-capture reference during a three-tendon cyclic trajectory. If this works generally, an origami arm could sense and eventually control its own shape without external vision, which matters in cluttered or visually obstructed environments.","feed_headline":"Origami arm reads its own bend from tendon resistance","feed_subtitle":"Conductive tendons double as sensors, letting the arm reconstruct its own path without cameras or external tracking.","key_machinery":"The central object is the conductive thread used as both tendon and sensor, exploiting the relation $R = \\rho l/A$ so that active length is encoded in resistance. The measurement chain is a Wheatstone bridge that records each tendon's resistance, a normalized third-order polynomial $l_i = f(R_i)$ that calibrates resistance to tendon length, and the piecewise-constant-curvature forward kinematic model of Eqs. (2)--(3) that converts the three tendon lengths into segment length $L$, bending angle $\\theta$, and deflection angle $\\phi$, and then to the end-effector position. The calibration polynomial is what carries the sensing claim: its fitted coefficients and RMSE quantify how faithfully resistance tracks length.","core_discovery":"On this design, a Yoshimura-pattern origami tube with flat prismatic rings is pulled by three conductive stainless-steel threads, each thread simultaneously acting as the actuating tendon and as a resistive strain sensor. Rolling a tendon shortens its active length, lowering its resistance; the paper measures each resistance through a Wheatstone bridge, normalizes the readings, and maps them to tendon lengths with a third-order polynomial (R$^2 = 0.97$, RMSE $= 1.25$ mm). These reconstructed lengths feed a Piecewise Constant Curvature forward kinematic model that outputs the end-effector coordinates. During a closed cyclic trajectory with a $2\\pi/3$ phase offset between tendons, the reconstructed path is reported to show reasonable agreement with OptiTrack motion capture, demonstrating configuration reconstruction from onboard resistance alone.","pith_inferences":["The resistance--length mapping was fitted on one calibration set and its force-independence was tested on a free thread, not inside the folded bellow under bending and contact; drift there would directly corrupt the reconstructed path, so a practical system needs periodic recalibration or compensation.","Since the resistance swing is only 1--2 $\\Omega$, the paper's own signal-conditioning suggestion (amplifying before filtering) is the lever that determines whether this method reaches sub-millimeter accuracy; the current RMSE of 1.25 mm in length already propagates to end-effector error.","A testable extension is to run the same technique with two or more origami modules in series, where coupled tendon routing will make the resistance--length mapping nonlinearly interdependent; whether the single-segment calibration survives that coupling is an open question."],"forward_implications":["An origami continuum manipulator can be closed-loop controlled using only onboard resistance readings, removing the need for cameras or motion capture in cluttered environments.","The same conductive-tendon sensing could extend to multi-module manipulators and to detecting contact or payload-induced changes, since the paper lists those as natural next steps.","Because the tendon resistance was shown to be insensitive to tension up to 10 N in the free-thread test, actuation effort and sensing can be multiplexed without an extra sensor layer.","The flat-ring origami body provides the stiffness needed for dynamic motions while keeping the structure flexible, so proprioception is added without sacrificing the manipulator's compliance."],"supporting_citations":[{"why":"Supplies the Piecewise Constant Curvature assumption used for the forward kinematic model.","marker":"[22]"},{"why":"Provides the kinematic relations converting tendon lengths into segment length, bending angle, and deflection angle.","marker":"[24]"},{"why":"Tracker software used to extract tendon lengths from video, the ground truth for calibrating the resistance--length polynomial.","marker":"[25]"},{"why":"Demonstrates the torsionally stiff origami continuum body and inverse kinematics that this work builds on, providing the baseline need for integrated sensing.","marker":"[9]"},{"why":"Presents the PCC-based kinematic modeling approach for extensible continuum actuators used in the manipulator model.","marker":"[23]"}],"fun_headline_variants":["Origami arm senses its own bends via thread resistance","Self-aware origami manipulator uses tendon resistance as sensors","No cameras needed: Origami arm tracks itself with conductive tendons","Conductive tendons give origami robot a sense of its own shape","Origami manipulator reads its own position from tendon resistance"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The mapping from measured resistance to tendon length, fitted on one calibration dataset, stays accurate during arbitrary coupled actuation of all three tendons inside the folded structure, even though force independence was only tested on a free thread.","fun_headline_variants_meta":{"raw":{"variants":["Origami arm senses its own bends via thread resistance","Self-aware origami manipulator uses tendon resistance as sensors","No cameras needed: Origami arm tracks itself with conductive tendons","Conductive tendons give origami robot a sense of its own shape","Origami manipulator reads its own position from tendon resistance"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000637,"raw_usage":{"total_tokens":2890,"prompt_tokens":852,"completion_tokens":2038,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":468,"completion_tokens_details":{"reasoning_tokens":1955}},"tokens_in":468,"tokens_out":2038,"duration_ms":13399,"temperature":1.0,"reasoning_tokens":1955,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-08T15:40:05.297138+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Run the three-tendon phase-shifted cyclic protocol from Fig. 5B, compare the reconstructed end-effector positions to OptiTrack, and compute the mean and maximum error over cycles; if the maximum error grows with cycle number or exceeds a few centimeters without any vision correction, the resistance--length mapping is not stable under coupled loading.","supporting_citations":[{"cited_title":"Kinematics for multisection continuum robots,","cited_arxiv_id":null,"evidence_quote":"Provides the kinematic relations converting tendon lengths into segment length, bending angle, and deflection angle."},{"cited_title":"Tracker introduction to video modeling,","cited_arxiv_id":null,"evidence_quote":"Tracker software used to extract tendon lengths from video, the ground truth for calibrating the resistance--length polynomial."},{"cited_title":"An origami continuum robot capable of precise motion through torsionally stiff body and smooth inverse kinematics,","cited_arxiv_id":null,"evidence_quote":"Demonstrates the torsionally stiff origami continuum body and inverse kinematics that this work builds on, providing the baseline need for integrated sensing."},{"cited_title":"A geometric approach towards inverse kinematics of soft extensible pneumatic actuators in- tended for trajectory tracking,","cited_arxiv_id":null,"evidence_quote":"Presents the PCC-based kinematic modeling approach for extensible continuum actuators used in the manipulator model."}],"review_version":1}