{"id":"c85dc6cc-a7fe-4edc-a40b-6f3e13973c57","arxiv_id":"2504.20313","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":3.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":2,"one_line_summary":"The paper fits a thrust equation with two empirical constants and measures the KV380 motor torque constant as 0.0311 Nm/A, but the thrust validation depends on fitted parameters and a coarse 200-node CFD.","lead":"This engineering thesis fits a thrust equation and measures the torque constant for Harpy, a bipedal robot with two electric ducted fan thrusters. It reports a motor constant of 0.0311 Nm/A and a thrust model the authors say predicts within about 5% of simulation, though the fitted and coarse-validation basis is weaker than the prose suggests.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Eq. 3.9 is printed without the kgf-to-Newton conversion: its RHS is in Newtons, and Table 3.2 only matches after dividing by g=9.81; a verbatim implementation will overestimate kgf by ~9.8x.","rationale":"The reader's weakest assumption concerned the physical validity of pitch speed and the ad hoc (K1,K2) form; that is a legitimate modeling concern, but the manuscript has a more basic, checkable defect. The central validation claim (5.22% error) depends on Eq. (3.9), and that equation as printed cannot produce the table's kgf values. This is not merely a typesetting issue: the thesis explicitly says the output is in kgf and uses these values for EDF selection and moment estimates in Chapter 4. The torque-constant result (Kt = 0.0311 Nm/A) is a separate, reasonably supported measurement and is not affected by this attack; that is why the appropriate outcome remains a conditional acceptance rather than rejection. The fix is straightforward: print the gravity conversion, relabel units, and ideally release the calculation script. I therefore keep the reader's CONDITIONAL verdict, but for a different and more concrete reason than the pitch-speed assumption.","tokens_in":32887,"tokens_out":12190,"duration_ms":123043,"concrete_test":"Reproduce Table 3.2 row 1 using Eq. (3.9) in SI units: d=8 in, Pitch=5 in, RPM=20,000, vo=0, rho=1.225 kg/m^3, and (d/(3.29546*Pitch))^1.5. Compute the RHS exactly as printed. It should evaluate to about 24.1 N. Then divide by 9.81 and confirm this equals 2.455 kgf. If the printed equation does not contain the /9.81 step, the manuscript must be corrected to include it, or all outputs must be relabeled as Newtons and the validation table regenerated.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The final thrust equation, Eq. (3.9), is dimensionally inconsistent with its stated output unit. The RHS is 1.225 kg/m^3 * pi * (0.0254d)^2/4 * [RPM*0.0254*Pitch/60]^2 * (d/(3.29546*Pitch))^1.5. With d and Pitch in inches and RPM in rev/min, the bracketed term is m/s, so the product has units kg*m/s^2 = N. The text and Table 3.2 label the output as 'kgf', but no factor 1/g appears. Reproducing the first row (d=8 in, Pitch=5 in, RPM=20,000, static vo=0) gives RHS approx 24.1 N; dividing by 9.81 yields 2.455 kgf, exactly the tabulated value. Thus Table 3.2 was computed with an unprinted gravity conversion, and a reader who implements Eq. (3.9) verbatim while trusting the 'kgf' label will command roughly 9.8 times the intended force. This is first-order relative to the reader's concern about pitch-speed physics: even if pitch speed and the (K1,K2) form are accepted, the printed equation fails to reproduce the claimed validation. The 5.22% average error against an ANSYS model with 200 nodes and a single-node thrust readout cannot resolve this unit ambiguity, and the hardware comparison is a single max-thrust point without an RPM sweep or error bars.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"This MS thesis reports system-identification work for the Harpy thruster-assisted bipedal robot. It derives a semi-empirical thrust equation for electric ducted fans (EDFs) that relates thrust to fan diameter, blade pitch, RPM, air density, and ambient velocity through empirical coefficients K1 and K2, and it attempts to validate the model against ANSYS Fluent CFD and a limited load-cell measurement. The thesis also characterizes the KV380 joint motor's torque constant Kt by three methods: datasheet estimation, back-EMF measurement, and a calibrated reaction torque sensor, recommending Kt = 0.0311 Nm/A from the torque-sensor method for use in closed-loop force-based impedance control. In addition, the manuscript documents the EDF/ESC test platform, thruster mounting, and integration into Harpy.","tokens_in":33229,"tokens_out":6394,"duration_ms":62906,"significance":"The torque-sensor characterization is the strongest contribution: it is a direct calibrated measurement with 28 trials, and the reported value of 0.0311 Nm/A is plausible and useful for current-based joint torque control. The paper also provides a useful integration record for a thruster-assisted biped. However, the thrust-model contribution is not presently validated: K1 and K2 are fit to vendor data, the ANSYS model uses 200 nodes with a single-node thrust readout, the hardware comparison is a single maximum-thrust point without uncertainty, and Eq. (3.9) has a load-bearing units error. If the thrust model is reframed as a calibrated curve fit and the reproducibility errors are corrected, the thesis can be a useful engineering contribution; as written, the central thrust-model claim is not supported.","major_comments":[{"comment":"Equation (3.9) is printed with 'kgf' as the output unit, but its right-hand side evaluates in Newtons: substituting d=8 in, Pitch=5 in, RPM=20,000, and vo=0 gives approximately 24.1 N, and Table 3.2 lists 2.455 kgf, which is exactly 24.1/9.81. The table can only be reproduced with an unprinted division by g. A reader implementing Eq. (3.9) verbatim while trusting the kgf label will overestimate thrust by roughly a factor of 9.8; this is a load-bearing reproducibility error in the central thrust-model claim.","section":"§3.1, Eq. (3.9) and Table 3.2"},{"comment":"The two printed forms of the thrust equation do not agree: Eq. (3.8) has K1=0.30 in (K1·d/Pitch)^K2, while Eq. (3.9) uses d/(3.29546·Pitch), i.e., a coefficient of approximately 0.30345, and the placement of the correction factor relative to the subtracted ambient-velocity term is ambiguous. The bracket structure in Eq. (3.9) is difficult to parse, and the text does not explain the change from 0.30 to 0.30345; this must be reconciled for the equation to be implementable.","section":"§3.1, Eqs. (3.8)–(3.9)"},{"comment":"The validation is not independent: K1 was fitted to 149 vendor data points and K2 was set by trial and error in Section 3.1, and Section 3.3 then compares that same fitted model to ANSYS. The claimed average error of 5.22% therefore measures agreement of an empirical fit with one coarse simulation, not predictive accuracy on unseen conditions. The manuscript should either state that the model is a calibrated curve fit or validate it on held-out data and against direct thrust measurements.","section":"§3.1 and §3.3"},{"comment":"The ANSYS model uses 200 nodes and reads thrust from a single node at the propeller output, with no mesh-convergence or domain-independence study reported. With this resolution, the 5.22% agreement cannot be distinguished from discretization error, so the CFD comparison does not provide quantitative validation of the thrust equation. A mesh-convergence study and thrust integration over the outlet or rotor surface are required.","section":"§3.2, Table 3.1"},{"comment":"The hardware comparison is a single maximum-thrust point (2.4 kgf) without RPM measurement, error bars, or a stated number of repeated trials, and the load-cell plot in Fig. 3.5 appears to come from a separate tabletop test. This is insufficient to support the thesis's 'less than 6% error' claim; a sweep over PWM/RPM with repeated trials and reported uncertainty is needed.","section":"§3.3, Table 3.3"}],"minor_comments":[{"comment":"Equation (3.7) gives pitch speed in mph (RPM·Pitch/1056), while Eq. (3.8) uses RPM·Pitch·0.0254/60 in m/s; the two expressions are inconsistent, and Eq. (3.7) is not used in the subsequent derivation.","section":"§3.1, Eq. (3.7)"},{"comment":"The sentence 'using the maximum thrust of the EDF, which is 2.7 kgf (approximately 27 N), applied at a distance of 0.2 m... the resulting moment is slightly over 0.5 kgf·m' mixes kgf and N; the moment should be expressed in N·m for consistency.","section":"§4.3"},{"comment":"The symbols 'KGF' and 'kgf' are used inconsistently; the manuscript should standardize on either kgf or N throughout.","section":"Figures 3.4 and 3.5, Table 3.2"},{"comment":"The statement that 'all three methods fall within five sigma of the Back EMF method' is not supported because the sigma values for the three methods are not reported; the distributions should be given numerically.","section":"§5.2"},{"comment":"There are numerous typos and grammar issues, including 'coeficients', 'radio' for 'ratio', and 'menuvers'; a careful proofread is needed before publication.","section":"Throughout"}],"recommendation":"major_revision","confidential_remarks":"The torque-sensor characterization is the most defensible and publishable part of this work. The thrust-model section needs a major reframing and a corrected, SI-consistent equation before the central claims can be accepted; if the authors provide an independent load-cell validation sweep and properly report uncertainty, a future revision could be viable."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Quick take on 2504.20313: the torque-sensor half is real, usable hardware data; the thrust half has a unit bug that undermines the printed equation, and the validation is too weak to support the headline accuracy.\n\nWhat's new: a direct torque-sensor measurement of Kt for the T-Motor MN4006 KV380 giving 0.0311 Nm/A, with 28 trials and a plausible setup. That's a concrete number the Harpy project can use for current-based force control, and the comparison of three methods (datasheet, back-EMF, torque sensor) is a fair exercise. The thrust model itself is a momentum-theory expression with two fitted constants; not novel, but the fitted values for this EDF are new data.\n\nSoft spots, in order of severity. First, Eq. 3.9 does not reproduce Table 3.2 as printed. The RHS comes out in Newtons, but the table says kgf. I checked the first row: 24.1 N divided by 9.81 gives 2.455 kgf, exactly the tabulated value. So the table was computed with an unprinted g conversion, and any reader who implements the equation verbatim will overestimate thrust by about 9.8x. That's a load-bearing error, not a typo in a footnote.\n\nSecond, the validation is not independent. K1=0.30 is fitted to 149 vendor points, K2=1.5 chosen by trial and error, then the equation is compared to an ANSYS model with 200 nodes and a single-node thrust readout. The 5.22% average error is CFD-vs-equation, not prediction-vs-experiment; the empirical comparison is one max-thrust point with no RPM sweep and no error bars.\n\nThird, the abstract and contributions claim less than 6% error and closed-loop force-based impedance control integration, but the thesis doesn't actually demonstrate closed-loop impedance control; it characterizes the motor so that such control is possible. That overstatement should be tempered.\n\nThe torque part is solid enough to be useful. The thrust part needs a corrected equation, release of the fitting dataset and CFD files, and error bars on Kt before the accuracy claims are credible. If those changes are made, this is a reasonable system-identification report for the Harpy platform.\n\nRecommendation: send to peer review, but expect revision. The Kt measurement alone justifies referee time; the thrust model needs fixing first.","headline":"Torque-sensor Kt is solid hardware data; the thrust equation as printed has a unit bug that breaks reproducibility, and the validation is too weak to support the headline accuracy.","tokens_in":33762,"tokens_out":3251,"would_cite":false,"duration_ms":31764,"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":"The thesis shows that a fitted thrust equation predicts electric ducted fan thrust within 5.22% on average against CFD, and that torque-sensor measurement gives the joint motor a torque constant of $K_t = 0.0311\\ \\mathrm{N\\,m/A}$.","keywords":["system identification","electric ducted fan","thrust modeling","torque constant","bipedal robot","impedance control","CFD validation","thruster-assisted locomotion"],"falsifier":"Mount the actual EDF on a thrust stand with a tachometer, sweep the throttle from minimum to maximum, and compare measured thrust with Eq. 3.9 at each measured RPM; systematic growth of error outside the fitted range, or residuals that trend with RPM, would falsify the pitch-speed-plus-loss-factor model. A locked-rotor test applying known currents and directly measuring torque would settle whether $K_t = 0.0311\\ \\mathrm{N\\,m/A}$ holds at stall.","tokens_in":32639,"feed_emoji":"🚀","tokens_out":9122,"duration_ms":81729,"temperature":0.7,"pith_summary":"The thesis tries to establish that a simple closed-form thrust equation, fitted with two empirical constants, can predict the thrust of the electric ducted fans on the Harpy bipedal robot well enough for design and control, and that a direct torque-sensor measurement yields a trustworthy torque constant for the robot's joint motors. The practical motivation is that thrusters give a biped roll stabilization, obstacle clearance, and stumble recovery, but only if the propulsion and actuation hardware are well characterized. The paper reports the fitted equation agreeing with CFD to an average 5.22% error, and a torque constant of $K_t = 0.0311\\ \\mathrm{N\\,m/A}$ that will be used for closed-loop force-based impedance control.","feed_headline":"Thrust formula matches ducted-fan CFD within 5.22%","feed_subtitle":"The same system ID yields a 0.0311 N·m/A torque constant for closed-loop force-based impedance control","key_machinery":"The load-bearing object is a closed-form thrust equation, $F = 1.225 \\cdot \\pi (0.0254 d)^2/4 \\cdot (\\mathrm{RPM}\\cdot 0.0254\\cdot \\mathrm{Pitch}/60)^2 \\cdot (K_1 d/\\mathrm{Pitch})^{K_2}$ with $K_1 \\approx 0.30$ and $K_2 = 1.5$, where the first speed factor is the blade pitch speed and the parenthetical ratio is an empirically tuned loss term. It carries the thrust argument by predicting output from diameter, pitch, RPM, and air density without running CFD. The second mechanism is the torque-sensor testbed: a reaction torque sensor in the 0–2 N·m range, paired with a load-cell transmitter and its calibration curve, converts measured current into torque, and the slope of torque versus motor current supplies the $K_t$ used in the control law.","core_discovery":"The paper's central claim is that electric ducted fan thrust can be captured by a momentum-based equation in which the effective exit velocity is the pitch speed, RPM times blade pitch, and all remaining aerodynamic losses are absorbed by an empirically fitted factor. The constants $K_1 = 0.30$ and $K_2 = 1.5$ were selected by trial and error against 149 vendor data points, and the resulting equation reproduces CFD results across four diameter/pitch combinations with an average absolute error of 5.22%. For the selected EDF unit, the same equation predicts about 2.64 kgf of thrust, consistent with CFD and bench results near 2.4 kgf and the manufacturer's 2.75 kgf. On the motor side, the thesis claims that the KV380 motor's torque constant is most reliably obtained with a calibrated reaction torque sensor, giving $K_t = 0.0311\\ \\mathrm{N\\,m/A}$, which becomes the conversion factor from motor current to joint torque for impedance control.","pith_inferences":["The 5.22% error is reported against CFD, not against a dense set of direct thrust measurements on the actual EDF; a realistic next test is a full throttle sweep on a thrust stand to see whether the fitted form holds across the whole RPM range.","The fitting procedure could be extended to make air density and ducted-fan hub blockage explicit, which would likely improve fidelity at altitude or with the annular-area correction already noted in the thesis.","Using a single measured $K_t$ for all six joints assumes the motors are built alike; per-joint torque calibration would be a natural follow-up before relying on current-based torque commands in control.","With $K_t$ known, current feedback becomes torque feedback, so the same hardware could support contact-force estimation without adding joint torque sensors."],"forward_implications":["A designer can estimate EDF thrust from diameter, pitch, and RPM with a hand calculation, bypassing CFD, at an average error around 5% for the tested configurations.","The dual-EDF configuration on Harpy delivers roughly 2.4 kgf in simulation and on the bench, enough to support a thrust-to-weight ratio above one for the robot.","With the thrusters mounted about 0.2 m from the center of mass, maximum thrust produces roughly 0.5 kgf·m of roll moment, which the paper argues is sufficient for roll stabilization.","Using $K_t = 0.0311\\ \\mathrm{N\\,m/A}$, joint torque can be commanded through motor current, enabling closed-loop force-based impedance control on the existing hardware.","Among the three torque-characterization methods, the torque-sensor approach gives the most consistent and precise values, while the back-EMF method is noisier."],"supporting_citations":[{"why":"Supplies the empirical-plus-CFD propeller thrust modeling approach the thesis extends.","marker":"[13]"},{"why":"Provides blade element momentum theory as the analytical basis for the thrust model.","marker":"[14]"},{"why":"Source for empirically tuned thrust-equation constants of the K1/K2 type.","marker":"[15]"},{"why":"Gives the Clarke/Park transforms and the torque-constant to speed-constant relation used in the motor analysis.","marker":"[16]"},{"why":"Describes the reaction-torque-sensor method adopted for direct torque measurement.","marker":"[17]"},{"why":"Vendor specifications for the selected EDF unit: maximum thrust, diameter, and mass used in the comparisons.","marker":"[74]"}],"fun_headline_variants":["Thrust equation matches CFD to 5.22% for Harpy","Harpy thruster ID: 5.22% thrust error, 0.0311 N·m/A","EDF thrust model validated within 5.22% against CFD","Bipedal thruster characterization: 5.22% CFD match","Torque and thrust ID for Harpy: 0.0311 N·m/A, 5.22% error"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The thrust equation assumes the air leaving the fan moves at pitch speed, RPM times blade pitch, and that every unmodeled effect is captured by the fitted factor $(K_1 d/\\mathrm{Pitch})^{K_2}$ with $K_2 = 1.5$ chosen by trial and error; if the pitch-speed assumption or that functional form is wrong, the predicted thrust is wrong no matter how well $K_1$ was fitted.","fun_headline_variants_meta":{"raw":{"variants":["Thrust equation matches CFD to 5.22% for Harpy","Harpy thruster ID: 5.22% thrust error, 0.0311 N·m/A","EDF thrust model validated within 5.22% against CFD","Bipedal thruster characterization: 5.22% CFD match","Torque and thrust ID for Harpy: 0.0311 N·m/A, 5.22% error"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.001034,"raw_usage":{"total_tokens":4344,"prompt_tokens":929,"completion_tokens":3415,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":545,"completion_tokens_details":{"reasoning_tokens":3299}},"tokens_in":545,"tokens_out":3415,"duration_ms":23964,"temperature":1.0,"reasoning_tokens":3299,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-16T05:32:24.517661+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Mount the actual EDF on a thrust stand with a tachometer, sweep the throttle from minimum to maximum, and compare measured thrust with Eq. 3.9 at each measured RPM; systematic growth of error outside the fitted range, or residuals that trend with RPM, would falsify the pitch-speed-plus-loss-factor model. A locked-rotor test applying known currents and directly measuring torque would settle whether $K_t = 0.0311\\ \\mathrm{N\\,m/A}$ holds at stall.","supporting_citations":[{"cited_title":"Bryant, ``Development of an empirical propeller thrust model using experimental measurements and computational fluid dynamics,'' Ph.D","cited_arxiv_id":null,"evidence_quote":"Supplies the empirical-plus-CFD propeller thrust modeling approach the thesis extends."},{"cited_title":"Analysis of the Blade Element Momentum Theory","cited_arxiv_id":"2004.11100","evidence_quote":"Provides blade element momentum theory as the analytical basis for the thrust model."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Source for empirically tuned thrust-equation constants of the K1/K2 type."},{"cited_title":"Spong, S","cited_arxiv_id":null,"evidence_quote":"Gives the Clarke/Park transforms and the torque-constant to speed-constant relation used in the motor analysis."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Describes the reaction-torque-sensor method adopted for direct torque measurement."},{"cited_title":"[Online]","cited_arxiv_id":null,"evidence_quote":"Vendor specifications for the selected EDF unit: maximum thrust, diameter, and mass used in the comparisons."}],"review_version":1}