{"id":"f5ad506b-3ba2-41b7-9767-64553801c198","arxiv_id":"2602.01385","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":5.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":6,"one_line_summary":"TriphiBot is a quadcopter with two passive wheels that flies, drives, and swims by shifting its center of gravity and using FOC motor drivers in all three media.","lead":"This paper describes TriphiBot, a drone-like robot that can fly, drive on land or the seabed, and swim using only its four rotors and two free wheels. It shifts the robot's center of gravity to make ground motion efficient and uses encoder-based field-oriented motor control to work in both air and water.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Ground-mode flatness and the 100%-thrust claim rely on a no-slip, flat-floor assumption (Eqs. 9, 14, 17) that the paper's own 9-cm obstacle and seabed tests violate; no slip/terrain model or uncertainty is included.","rationale":"The reader's weakest_assumption identifies exactly the most load-bearing vulnerability: every headline claim—eccentric-CoG efficiency, HNMPC terrestrial tracking, and smooth cross-domain transitions—passes through the terrestrial flatness model. The prototype and FOC comparison are genuine evidence and are not disputed; the issue is not that the robot fails on flat floors, but that the claimed deployment envelope ('land or seabed', obstacle crossing) is broader than the model supports. The 9-cm obstacle test is especially telling because it is the paper's own explicit violation of θ_T=0, while the seabed demo occurs in a tank with a flat bottom but loose sediment, where the no-skid assumption is already questionable. A conditional verdict remains appropriate: the concern is addressable by adding a slip/terrain model, by restricting claims to flat no-slip surfaces, or by providing robustness data under slip. No change to the reader's verdict is needed.","tokens_in":14964,"tokens_out":14705,"duration_ms":171997,"concrete_test":"Independently re-derive §III.B after replacing Eq. (14) with a standard side-slip model, e.g., ψ_T = arctan2(κ(ṗ_y − v_slip_y), κ(ṗ_x − v_slip_x)) with slip velocity determined by lateral force balance at the wheels, and check whether all states/inputs are still expressible from σ_t=[p_x,p_y,θ_T] and derivatives. If a slip state or friction coefficient enters non-algebraically, the terrestrial system is not differentially flat with the proposed flat output, and the HNMPC's ground-mode guarantees lack support outside the no-slip idealization.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The terrestrial model in §III is the load-bearing foundation for both the efficiency claim and the HNMPC. Eq. (9) assumes motion in a horizontal plane with no lateral wheel velocity; Eq. (14) sets yaw from the velocity direction under a no-skid assumption; Eq. (17) derives thrust from longitudinal acceleration under that same assumption and with θ_T=0. Because the HNMPC (Eqs. 18–20) uses this flat mapping, any violation invalidates the planned command sequence. The paper itself presents two cases that violate it: the 9-cm obstacle crossing (§V.D, Fig. 14c) deliberately uses θ_T≠0 and contact with an obstacle, so Eq. (9) and the flat-output relation no longer hold; seabed operation (§V.E, Fig. 14a) occurs on a loose, sedimented bottom where no-skid rolling is implausible. No slip model, terrain model, or disturbance/uncertainty term is provided. If the no-skid relation fails, ψ_T is no longer a function of the flat output, and the terrestrial flatness proof collapses; the '100% of thrust used for propulsion' claim then holds only on a flat, high-friction, horizontal surface, which is a much narrower claim than land/seabed capability.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"TriphiBot is a triphibious robot built from a 1.1 kg quadcopter frame plus two passive wheels, with no additional actuators. The paper claims aerial, terrestrial, and aquatic locomotion, smooth transitions between these modes, and an eccentric center-of-gravity design that aligns rotor thrust with ground motion so that, in the nominal terrestrial configuration, 100% of thrust contributes to propulsion. A second contribution is a Field-Oriented Control (FOC) propulsion system intended to provide fast, precise bidirectional thrust and automatic torque matching in air and water. The paper derives aerial and terrestrial dynamics, proves differential flatness for both modes, and proposes a hybrid NMPC-PID control architecture. Experiments include propulsion bench tests, aerial and terrestrial figure-eight tracking, a 9 cm obstacle crossing, underwater attitude tracking, and water-to-air takeoff. The hardware demonstrations support the existence claim that the robot can fly, roll, swim, and transition among these modes; however, the strength of the efficiency and modeling claims is limited by the assumptions in the terrestrial model and by the absence of repeated trials and uncertainty quantification.","tokens_in":15329,"tokens_out":7443,"duration_ms":87353,"significance":"If the design and modeling claims hold, this is a useful minimalist contribution to cross-domain robotics: the eccentric-CoG idea is elegant, the FOC-based unified propulsion unit is plausible and experimentally supported, and the hybrid control architecture is a practical step toward seamless air-land-water operation. The paper clearly demonstrates the central hardware capability in real experiments, which is the strongest part of the work. It also provides quantitative comparisons with prior propulsion systems (e.g., specific thrust 0.825 N/W vs. 0.059 N/W for ESC, and vs. 0.265 N/W for TJ-FlyingFish), which are valuable even if single-trial. The main caveat is that the terrestrial flatness proof and the associated efficiency claims are derived under a flat, no-skid, horizontal-surface assumption that the paper's own obstacle and seabed experiments violate, so the theoretical guarantees are narrower than the claimed operational capability. The paper would be acceptable after substantial revision that honestly delimits these assumptions or extends the model.","major_comments":[{"comment":"The terrestrial differential-flatness proof and the HNMPC that relies on it are built on assumptions that the paper's own experiments violate. Eq. (9) constrains the robot to motion in a horizontal plane with p_z=0 and no lateral wheel velocity; Eq. (14) sets ψ_T from the velocity direction under a no-skid assumption; Eq. (17) then derives the required thrust from longitudinal acceleration. The 9 cm obstacle test in §V.D deliberately uses θ_T≠0 and contact with the obstacle, and the seabed experiment in §V.E operates on loose sediment where no-skid rolling is implausible. No slip model, terrain model, or disturbance/uncertainty term is provided. As a result, the flat-output mapping used by the HNMPC in Eqs. (18)–(20) is not valid in these demonstrated scenarios, and the '100% of thrust used for propulsion' claim holds only on a flat, high-friction, horizontal surface. This is load-bearin","section":"§III.B, Eq. (9), (14), (17); §V.D, Fig. 14(c); §V.E, Fig. 14(a)"},{"comment":"The 95.37% energy-saving figure is not measured for TriphiBot; it is taken from prior work by overlapping authors (Lai et al. [14]) on a different, transformable robot. The statement 'Existing research has confirmed that robots using this method for ground movement can achieve energy savings of up to 95.37%' may be technically true, but it is presented in the introduction of the current paper as if it supports the eccentric-CoG design. The paper provides no onboard energy-consumption comparison between ground and flight modes for the actual TriphiBot prototype. Since ground-mode efficiency is a central contribution, the authors should either report a direct efficiency measurement for this platform, or clearly state that the figure is a literature result for a related mechanism and should not be attributed to the present design without further evidence.","section":"§II.A and §V.B"},{"comment":"The hybrid dynamics f(x,u)=η f_t(x,u)+(1−η) f_a(x,u) is not well-defined as written. The state vector X in §III.A contains both aerial states (v_A, Θ_A, ω_A) and terrestrial states (v_l, θ_T, ψ_T, ω_T), but f_t and f_a are defined on different state spaces and propagate different subsets of X. To be a valid single transition function for the NMPC constraint x_{i+1}=f(x_i,u_i), the unused states must be explicitly carried or their evolution during the transition must be defined. In addition, Eq. (20) sets η from the reference height p_W,z,k, whereas §V.F says the switch is based on 'real-time altitude and pitch angle.' If the switching is reference-based, a tracking error can trigger the wrong mode; if it is state-based, Eq. (20) is incorrect. This is a modeling/control consistency issue that should be clarified and reconciled.","section":"§IV.A, Eq. (18)–(20)"},{"comment":"The quantitative claims of superiority are based on single trials with no error bars, repeated runs, or statistical analysis. For example, the specific-thrust comparison (0.825 vs. 0.059 N/W, a 14× margin), the underwater startup time (0.14 s), and the trajectory-tracking RMSE values (aerial 0.096 m, terrestrial 0.074 m, hybrid 0.084 m) are reported without any indication of variance across runs. The underwater attitude-tracking maximum error of 31° is likewise a single observation. Given that the paper makes strong comparative claims, at least a few repeated trials and a measure of spread (or a clear statement that the data are representative) should be provided.","section":"§V.B–§V.E"}],"minor_comments":[{"comment":"Typo: 'devided' should be 'divided'.","section":"Fig. 2 caption"},{"comment":"The text states that underwater thrust T relates to rotational speed ω by T = c_a ω²; the coefficient should be the underwater thrust coefficient c_t,w, and the sign/direction dependence should be specified. Also, Table I lists c_t,a and c_t,w, but the notation in the text is inconsistent.","section":"§V.B"},{"comment":"The legend reads 'TrofyBot', which is not defined; it should refer to 'Lai et al. [14]' and the comparison should be mentioned in the caption for clarity.","section":"Fig. 13(a)"},{"comment":"If θ_T=0 is assumed for maximum energy efficiency, writing cosθ_T in the denominator is misleading because the expression is then just T = m ˙v_l. If θ_T is allowed to be nonzero in the flat-output parameterization, the text should state that Eq. (9) is modified accordingly, since a nonzero θ_T generally changes the vertical geometry and ground-contact constraints.","section":"§III.B, Eq. (17)"},{"comment":"The phrase 'maximum speed of up to 200,000 eRPM' refers to the electrical RPM capability of the driver electronics, not necessarily the mechanical rotor speed. Please clarify to avoid a misleading hardware claim.","section":"§II.B"}],"recommendation":"major_revision","confidential_remarks":"The paper is within the scope of the journal and reports a real, working triphibious platform with a clean mechanical design. The main risk is not the hardware but the over-stated theoretical guarantees: the terrestrial flatness proof and the '100% thrust' efficiency claim rest on ideal assumptions that the paper's own experiments contradict. The self-citation overlap with Lai et al. [14] is not, in my view, an integrity issue, but the transferability of the 95.37% energy-saving figure should be handled more carefully. With a revision that disambiguates the hybrid dynamics, adds repeated experiments or explicit limitations, and narrows the terrestrial claims to the idealized case, the paper could become a solid contribution."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Colleague—\n\nWorth a scan. TriphiBot is a quadcopter with two passive wheels and an intentionally shifted CoG so that when it rests on the wheels, the thrust axis is roughly horizontal. That lets the rotors push the robot along the ground without any steering or drive actuators. Add a FOC-based motor controller with encoders, and you have a single propulsion system that works in air and water. The robot actually flies, rolls on the ground, swims, crosses a 9-cm obstacle, and lifts off from water. The central design idea is simple and effective, and the hardware demonstrations are believable. That alone is a legitimate contribution: most triphibious platforms need transformable mechanisms or extra actuators; this one doesn't.\n\nThe paper also does a few things well. The terrestrial differential-flatness analysis is clean, given its assumptions. The pitch-stability comparison against two prior passive-wheel robots shows the eccentric-CoG trick helps. The FOC propulsion unit, with encoder feedback, beats a typical ESC in low-speed and reversal tests—not surprising, but the numbers are concrete. The control architecture, mixing HNMPC for air/ground and PID for water, is sensible for the sensing constraints.\n\nThe soft spots are in the scope of the claims. The 100% thrust-efficiency claim and the flat-output mapping both assume flat ground, no wheel slip, and θ_T=0. The paper's own obstacle-crossing experiment deliberately uses θ_T≠0, and the 'seabed' test is just driving on the bottom of a lab tank. So the strong efficiency claim holds only on a flat, high-friction surface, which is a much narrower claim than the abstract suggests. The 95.37% energy saving is borrowed from the authors' earlier paper, not measured here. There are no repeated trials or error bars, and parameters like added mass and drag are hand-tuned. The 'seamless transitions' language also oversells what is a finite-state switch with PID during water takeoff. None of this is fatal: the core existence claim, that this simple design can operate in all three domains, is solid. But the paper would be stronger if it admitted the efficiency result is conditional and reported variances.\n\nWho's this for? Robotics systems people who care about minimalism in cross-domain platforms. It deserves a serious referee—not a desk reject—but the authors should be pushed to narrow the claims and add repeatability data.\n\nRecommendation: send to peer review, with the expectation of major revision.","headline":"A credible triphibious hardware demo with a genuinely simple eccentric-CoG design; the efficiency and 'seamless' claims are narrower than the evidence, but the prototype deserves a serious referee.","tokens_in":15839,"tokens_out":2171,"would_cite":true,"duration_ms":22317,"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":"A quadcopter with two passive wheels and an eccentric center of gravity can fly, drive, and swim using a single set of rotors.","keywords":["triphibious robot","eccentric center of gravity","field-oriented control","hybrid nonlinear model predictive control","cross-domain transition","passive wheels","underwater propulsion","torque matching"],"falsifier":"Run the robot on a surface with measurable slip or a gentle slope and compare trajectory-tracking RMSE to the flat-ground case; if RMSE degrades sharply or the HNMPC diverges when θT deviates from zero, the central claim that thrust is '100% aligned with motion by default' and that the flat-output mapping is valid would be falsified. Alternatively, measure the actual thrust direction with a load cell while the robot pitches during obstacle crossing: the 100% utilization claim depends on θT = 0.","tokens_in":1308,"feed_emoji":"🚁","tokens_out":2023,"duration_ms":64824,"temperature":0.7,"pith_summary":"This paper argues that a single minimalist platform—a quadcopter frame with two freely rotating wheels and no extra actuators—can operate across air, land, and water. The key efficiency claim is that by deliberately shifting the center of gravity toward the front, the rotor thrust vector aligns with the direction of ground motion by default, so nearly all thrust contributes to propulsion instead of lifting the robot against gravity. To handle the vastly different torque demands of air and water, the authors propose a unified propulsion system based on field-oriented control (FOC) that gives precise, fast bidirectional rotor thrust. A hybrid nonlinear model predictive controller, combined with PID for underwater entry and exit, coordinates seamless transitions between modes. If correct, the design offers a lightweight, high-efficiency alternative to existing dual-mode and mechanically complex triphibious robots.","feed_headline":"TriphiBot flies, rolls, and swims with no extra actuators","feed_subtitle":"An eccentric center of gravity aligns rotor thrust with ground motion, and FOC handles both air and water.","key_machinery":"The load-bearing elements are (1) the eccentric CoG: a shift δ of 1.5 cm from the geometric center to the mass center, which creates a gravity restoring torque and aligns thrust with ground motion, eliminating the need for active tilting mechanisms; (2) the FOC-based unified propulsion system, which uses encoder-measured rotor position to generate precise rotating magnetic fields, enabling high torque at low speed, very high speed in air, and fast bidirectional switching; and (3) the differential flatness of the terrestrial dynamics, which lets a single HNMPC controller, switching on altitude, command both aerial and ground modes using the same rotor inputs.","core_discovery":"The paper's central discovery is that an eccentric CoG—a small forward shift of the center of gravity relative to the geometric center, following the roly-poly principle—makes the ground mode of a passive-wheeled quadcopter inherently efficient. When the wheels touch the ground, gravity rotates the body so the propeller plane is vertical; the restoring torque τr = mgδ sinθT passively stabilizes pitch, and the thrust vector is aligned with motion, allowing 100% of thrust to be used for propulsion. The authors show that the terrestrial dynamics are differentially flat with flat output [pW,x, pW,y, θT], permitting a unified HNMPC to plan and track trajectories across aerial and terrestrial mode","pith_inferences":["A natural extension is applying the eccentric-CoG principle to other rotorcraft (e.g., fixed-wing hybrids or tailsitters) to reduce ground rolling resistance without servo-tilt mechanisms.","The flat-ground, no-slip assumption is the main boundary: on uneven or slippery terrain the flat-output mapping breaks, so a slip-aware or terrain-estimating version of the controller would be a logical next step.","The FOC driver's torque matching may make it a drop-in upgrade for existing aerial-aquatic vehicles, but the reported specific-thrust improvements should be replicated independently before broad adoption.","The water-to-air transition logic relies on rotor-speed monitoring to detect leaving the surface; a more robust state estimator using pressure or IMU sensing would likely be needed in waves or currents."],"forward_implications":["Triphibious robots can be built from a standard quadcopter plus two passive wheels and a CoG shift, without added weight or actuators, preserving flight endurance.","The FOC propulsion unit can replace ESC-based drivers in aerial-aquatic robots to enable rapid, low-speed precise thrust control and better underwater energy efficiency.","The differential flatness result means trajectory generation and tracking on flat ground can reuse the same control pipeline as aerial flight, simplifying software and reducing tuning effort.","Smooth air-land-water transitions become possible with a single rotor set, reducing mechanical complexity and transition time compared to transformable designs.","The efficiency advantage of ground mode relative to flight becomes accessible without dedicated mechanical transformations, as the paper notes prior work achieving up to 95.37% energy savings."],"fun_headline_variants":["Eccentric CoG lets quadcopter use 100% thrust on land","TriphiBot: flies, swims, and rolls with no extra parts","Passive wheels plus FOC control enable triphibious motion","One quadcopter, two wheels, three environments, one system"],"cache_read_input_tokens":17024,"weakest_assumption_plain":"The paper's efficiency and flatness claims assume a flat supporting surface, no wheel skidding, no lateral wheel motion, and pitch angle kept near zero during ground travel; the moment the robot pitches to cross an obstacle, the flat-output mapping used by the controller is no longer valid.","fun_headline_variants_meta":{"raw":{"variants":["Eccentric CoG lets quadcopter use 100% thrust on land","TriphiBot: flies, swims, and rolls with no extra parts","Passive wheels plus FOC control enable triphibious motion","One quadcopter, two wheels, three environments, one system"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000611,"raw_usage":{"total_tokens":2698,"prompt_tokens":782,"completion_tokens":1916,"prompt_tokens_details":{"cached_tokens":256},"prompt_cache_hit_tokens":256,"prompt_cache_miss_tokens":526,"completion_tokens_details":{"reasoning_tokens":1836}},"tokens_in":526,"tokens_out":1916,"duration_ms":18936,"temperature":1.0,"reasoning_tokens":1836,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-03T05:39:52.512770+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Run the robot on a surface with measurable slip or a gentle slope and compare trajectory-tracking RMSE to the flat-ground case; if RMSE degrades sharply or the HNMPC diverges when θT deviates from zero, the central claim that thrust is '100% aligned with motion by default' and that the flat-output mapping is valid would be falsified. Alternatively, measure the actual thrust direction with a load cell while the robot pitches during obstacle crossing: the 100% utilization claim depends on θT = 0.","supporting_citations":[],"review_version":1}