{"id":"1ffc30f5-1221-48df-8247-9243d606d8e4","arxiv_id":"1908.02243","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":5.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":4,"one_line_summary":"A two-actuator spherical robot driven by a pendulum and a sliding mass is modeled with Euler-Lagrange dynamics and shown in simulation to track smooth paths with a pure-pursuit plus PID controller.","lead":"The paper derives the equations of motion for a spherical robot that rolls using a pendulum and steers by sliding a weight along a diagonal shaft, then shows in MATLAB simulation that a simple controller can track a smooth path. A generalist might read it for a worked example of how a nonholonomic robot with internal moving masses can be modeled and controlled, though no physical prototype or experimental data are provided.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The turning angle ψ is imposed by the cone relation (Eq. 6) but is absent from the Euler-Lagrange dynamics; the claimed complete model cannot predict or control turning.","rationale":"The reader's verdict is CONDITIONAL, with the rolling-cone model identified as the weakest assumption. I agree the cone model is central, but the more precise problem is that the yaw angle ψ never appears in the derived equations of motion. The kinematic model (Eqs. 3-6) is used to generate the world trajectory, while the dynamics (Eq. 47) is derived from angular velocities that omit ψ̇. This is an internal inconsistency rather than merely an unvalidated geometric approximation. The no-slip condition for a sphere rolling on a plane does not constrain the spin about the vertical, so the cone relation is not a necessary kinematic identity; at best it is a control specification. Therefore the simulated path tracking in Figs. 6-10 does not demonstrate that the Euler-Lagrange control scheme would steer a physical robot, even if the derived M, C, G matrices are correct for the local rolling/tilting dynamics. The abstract's reference to 'experimental results' is unsupported, as no experiments are reported. The paper could be revised by adding ψ as a dynamic state (or by explicitly modeling the shell's vertical spin and its actuation), which would change the order and the mass matrix, and by validating or replacing the cone relation. As written, the central claim of a complete dynamic model with turn capability is not secure.","tokens_in":21796,"tokens_out":23388,"duration_ms":253804,"concrete_test":"Re-derive the turning kinematics for Norma from a full 3D no-slip rolling model: include the sphere's yaw angle ψ (or full orientation) as a state, impose the contact-point no-slip constraint, and compute the kinetic energy and Euler-Lagrange equations in the reduced variables θ, α, φ, δ, ψ. Check (1) whether ψ̇ enters the kinetic energy; (2) whether the full dynamics reduces to Eq. (6) under a stated approximation (e.g., quasi-static rolling with negligible yaw inertia). If ψ̇ appears in the mass matrix or the ψ equation contains Γ and Φ, then Eq. (6) is an extra constraint and the paper's dynamics omits a degree of freedom. A cheaper partial check: substitute Eq. (6) into the z-component of Eq. (8) and verify dimensional consistency.","verdict_should_be":"REJECT","load_bearing_attack":"The paper's central claim is a complete Euler-Lagrange model (Sec. III-B) enabling roll and turn maneuvers. The model's turning is introduced in Sec. III-A through a rolling-cone analogy that yields the yaw rate ψ̇ = -Sφ θ̇/(R(Cφ-Sφ)) (Eq. 6). However, the yaw coordinate ψ is not a generalized coordinate: the Lagrangian (45) and the explicit M, C, G entries in Appendix A depend only on θ, α, φ, δ and their derivatives, and the shell/pendulum/slider angular velocities in Eqs. (8), (29), (36) contain no ψ̇ term. The world trajectory is then obtained by integrating Eq. (6) into Eq. (12), so the simulated turning behavior is imposed kinematically rather than produced by the dynamics. No equation of motion governs ψ, and Γ and Φ cannot influence yaw except through the assumed algebraic relation. For a sphere rolling without slip on a horizontal plane, the no-slip condition constrains only the horizontal velocity at the contact point; the vertical component of angular velocity is unconstrained by rolling. Nothing in the derived equations guarantees that a physical Norma follows Eq. (6). The dynamics and the trajectory integration are therefore two separate, unconnected models. Even if Eq. (6) were a correct kinematic rule for a tilted sphere (which is not established), the 'complete' dynamic model would still be incomplete for turning maneuvers because the yaw rotational kinetic energy and the corresponding generalized force are absent.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper presents the design, dynamics modeling, and control of Norma, a spherical robot actuated by a pendulum and a slider on a diagonal shaft. The authors use the Euler-Lagrange method to derive an explicit 4-DOF model M(q)q̈ + C(q,q̇)q̇ + G(q) = τ with generalized coordinates q = [θ, α, φ, δ], then propose a pure-pursuit kinematic controller together with PID dynamic controllers for the pendulum torque and slider force. The reported contribution is that this 2-actuator mechanism can perform rolling and turning maneuvers, and that MATLAB simulations demonstrate accurate path tracking. The Appendix gives element-wise expressions for M, C, and G, and the Coriolis matrix is constructed via Christoffel symbols so that Ṁ − 2C is skew-symmetric.","tokens_in":22067,"tokens_out":6338,"duration_ms":71740,"significance":"If the derived model and kinematic relations were correct, the paper would be a useful reference for simulation-based control design of pendulum-and-slider spherical robots. The first-principles derivation is detailed, the inertia matrix is symmetric by construction, and the explicit Appendix A is a valuable checkable artifact. However, the turning kinematics rests on an unproven cone analogy, the yaw coordinate is absent from the dynamics, and the 'verification' is a self-simulation of the same model with no independent experimental or model-based benchmark. These issues directly affect the paper's central claim of a complete model that enables turning control, so the current manuscript does not establish that claim.","major_comments":[{"comment":"Equation (6) does not follow algebraically from Eqs. (3)-(5). Substituting c_r = R C_φ and ρ = R C_φ − R S_φ into c_r θ̇ = ρ ψ̇ gives ψ̇ = C_φ/(C_φ − S_φ) θ̇, not the printed ψ̇ = −S_φ/(R(C_φ − S_φ)) θ̇. Beyond the algebra, the imaginary cone model is not justified for a sphere: for a sphere rolling without slip on a horizontal plane, the no-slip condition constrains only the horizontal velocity at the contact point, and the yaw rate should be the vertical component of the sphere's angular velocity, which from Eq. (8) is ψ̇ = −S_φ θ̇ with no dependence on R. Because Eq. (12) and all of the Section V trajectories inherit this relation, the simulated turning behavior is not grounded in the dynamics of a rolling sphere.","section":"Section III-A, Eqs. (3)-(6)"},{"comment":"The yaw angle ψ is not a generalized coordinate. The Lagrangian in Eq. (45) and the explicit M, C, and G elements in Appendix A depend only on θ, α, φ, δ and their derivatives; the angular velocities in Eqs. (8), (29), and (36) contain no ψ̇ term; and the generalized force vector in Eq. (47) has zero component for yaw. Consequently there is no equation of motion governing ψ, and the turning motion in the simulations is imposed by the algebraic relation of Section III-A rather than produced by the dynamics. The paper's claim of a complete dynamic model that enables turning control is therefore not supported as written.","section":"Section III-B and Appendix A"},{"comment":"The claimed verification of the mathematical model is circular: the simulation plant is the same Euler-Lagrange model (46)-(47) whose accuracy is being claimed, and the paper provides no independent experimental data, no comparison against a separately derived model, and no benchmark from the spherical-robot literature. The abstract's phrase 'against experimental results' is not supported by any experimental section in the manuscript. The claims should either be reduced to 'closed-loop simulation of the derived model' or an independent validation should be added.","section":"Section V, Simulations"}],"minor_comments":[{"comment":"The equations are heavily garbled, with missing hats on unit vectors, missing dot notation, and corrupted trigonometric subscripts (for example in Eqs. (6), (8), and (45)); the manuscript needs a careful typesetting pass before it can be checked reliably.","section":"Throughout"},{"comment":"The text does not specify which PID block in Fig. 5 corresponds to PID1, PID2, and PID3 in Table 1, and the gains are said to be selected by trial-and-error without any stability or robustness analysis.","section":"Section IV-B and Table 1"},{"comment":"The expressions for the desired slider displacement and rolling velocity are ambiguous: the normalized error gain ||e||/(||e||+k3) appears in different forms, and the units of k1, k2, and k3 are not discussed.","section":"Eqs. (50)-(51)"},{"comment":"The desired trajectory specification is corrupted ('0.01 0.02, 2 ,1.5 S S') and should be written explicitly so that the simulation can be reproduced.","section":"Section V, Eq. (52)"},{"comment":"The paper repeatedly cites the authors' own preprint [4] but does not compare the proposed model with existing pendulum-driven spherical-robot models from references [12]-[16], [21]-[27]; such a comparison would help position the contribution.","section":"References"}],"recommendation":"major_revision","confidential_remarks":"The yaw issue is not a presentation artifact: if the turning relation in Section III-A is wrong or not dynamically coupled, the simulations in Section V do not demonstrate turning control at all. I would ask the authors to rederive the nonholonomic kinematics from the sphere's no-slip condition, either include ψ as a generalized coordinate or justify its elimination as a quasi-coordinate, and obtain at least one independent validation (experimental or against a separately derived model). The abstract and Section V also overstate experimental verification."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Colleague, here's my take on Moazami et al. The Norma mechanism is a genuine new combination — a slider and a pendulum on the same diagonal shaft — and the Euler-Lagrange derivation is careful and complete as far as it goes. The inertia matrix is symmetric, the Coriolis construction uses the standard Christoffel trick, and the Appendix gives the full expansion. That is real work, and someone wanting to build a slider-pendulum SR could start from these equations.\n\nThe soft spot is load-bearing. The turning yaw rate ψ̇ is introduced through a rolling-cone analogy in Section III-A, but ψ never appears as a generalized coordinate. The dynamics in (46)–(47) involve only θ, α, φ, δ; there is no ψ̇ in the shell angular velocity (8), in the kinetic energies, or in M, C, G. The world trajectory is then recovered by integrating the assumed algebraic relation (6). So the 'complete' model does not predict turning; it imposes turning kinematically. For a sphere rolling without slip on a plane, the vertical spin component is unconstrained by the no-slip condition, so Eq. (6) is not a general kinematic law. It may be a legitimate planning constraint if enforced by control, but it is not a dynamic consequence. As written, the paper overstates the completeness of its model.\n\nSecond, the validation is circular. The model is verified by simulating the same model; there is no experiment, no independent data, no comparison against a simpler benchmark, and no released code. The abstract I was given claims validation 'against experimental results,' but the full text has no experiments. That needs to be corrected either way.\n\nMinor issues: the pure-pursuit plus PID control is standard and fine; the gains are hand-tuned, which is normal. The citation pattern looks honest, and the self-citation in the conclusions is to a companion kinematics paper, which is fine.\n\nBottom line: the mechanism is worth knowing about, and the derivation is a useful starting point, but the turning model is a serious gap. The authors could fix it by adding ψ (or an equivalent spin coordinate) to the generalized coordinates and re-deriving the kinetic energy, or by clearly reframing the paper as modeling the robot under a prescribed nonholonomic rolling constraint. As submitted, I would not trust the simulated turning trajectories as predictions.","headline":"A novel slider-pendulum spherical robot with a careful EL derivation, but the turning model is imposed kinematically and the validation is self-referential.","tokens_in":22633,"tokens_out":4390,"would_cite":false,"duration_ms":42915,"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":"This paper derives the full dynamics of a two-actuator spherical robot and shows, in simulation, that it can roll, turn, and track a smooth path.","keywords":["Spherical robot","Euler-Lagrange dynamics","Pendulum-driven robot","Slider actuation","Nonholonomic mobile robot","Path tracking control","PID control","Rolling cone kinematics"],"falsifier":"Lock the slider at a known off-center position so the ball tilts by $\\phi$, roll it forward at a roughly constant roll rate $\\dot\\theta$ on a flat hard floor, and measure the heading $\\psi$ over several full revolutions. Compare $\\dot\\psi$ with $-\\dot\\theta\\,\\sin\\phi/(R\\cos\\phi-R\\sin\\phi)$; a disagreement larger than the measurement error would refute the rolling-cone model and, with it, the derived dynamics and the simulated tracking results.","tokens_in":21559,"feed_emoji":"🤖","tokens_out":8271,"duration_ms":89524,"temperature":0.7,"pith_summary":"Norma is a spherical robot built around a diagonal shaft fixed inside a ball. A pendulum rotating about the shaft rolls the ball forward and backward, while a slider translating along the shaft shifts the center of gravity sideways, tilting the ball and making it turn. The paper derives the robot's full equations of motion, $M(q)\\ddot q + C(q,\\dot q)\\dot q + G(q)=\\tau$, for the four generalized coordinates $q=[\\theta,\\alpha,\\phi,\\delta]^T$, using the Euler-Lagrange method under minimal simplifying assumptions. It then couples a pure-pursuit kinematic controller with PID loops that convert desired roll speed and slider position into pendulum torque $\\Gamma$ and slider force $\\Phi$. In MATLAB simulation the closed-loop system tracks a smooth curved path, which is the evidence offered that the model and controller are sound. If correct, this is a mechanical design that needs only two actuators for both forward motion and steering, with a dynamics model detailed enough to guide control design before hardware is built.","feed_headline":"Two actuators steer a spherical robot in simulation","feed_subtitle":"A full Euler-Lagrange model plus pure-pursuit and PID control lets Norma roll and turn to follow a smooth path.","key_machinery":"The load-bearing object is the four-coordinate Euler-Lagrange dynamics written in the standard second-order form, together with the rolling-cone kinematic constraint used for turning. The generalized coordinates are $\\theta$ (sphere roll about the transverse shaft), $\\alpha$ (pendulum rotation about the shaft), $\\phi$ (tilt about the longitudinal axis), and $\\delta$ (slider displacement along the shaft). The rolling-cone relation supplies yaw rate from roll rate and tilt angle, while the Lagrangian built from the kinetic and potential energies of the sphere, pendulum, and slider supplies the matrices $M$, $C$, and $G$. Christoffel symbols are used to make the Coriolis/centripetal matrix skew-symmetric with $\\dot M-2C$. This structure is what lets the authors separate kinematics, handled by pure pursuit, from dynamics, handled by PID tracking of the desired roll rate and slider position.","core_discovery":"The central claim is that a spherical robot can obtain both rolling and steering from two internal actuators, and that this behavior is captured by a closed Euler-Lagrange model rather than by a simplified or decoupled approximation. The tilt angle $\\phi$ is coupled to yaw rate through a rolling-cone relation $\\dot\\psi=-\\dot\\theta\\,\\sin\\phi/(R\\cos\\phi-R\\sin\\phi)$, so slider motion that changes $\\phi$ produces turning, whereas pendulum torque $\\Gamma$ drives $\\dot\\theta$ and hence forward motion. The derived model has the standard robot form $M(q)\\ddot q+C(q,\\dot q)\\dot q+G(q)=\\tau$, and the paper constructs the Coriolis/centripetal matrix using Christoffel symbols so that $\\dot M-2C$ is skew-symmetric. A pure-pursuit outer loop specifies desired values for $\\dot\\theta$ and $\\delta$, and PID loops generate $\\Gamma$ and $\\Phi$; simulations show the robot converging to the desired trajectory with tracking error near zero.","pith_inferences":["A direct hardware test of the cone kinematic relation is the fastest way to decide whether the model transfers to a real robot: lock the slider at fixed $\\phi$, roll the ball at known $\\dot\\theta$, and compare measured heading change with $\\dot\\psi=-\\dot\\theta\\,\\sin\\phi/(R\\cos\\phi-R\\sin\\phi)$.","The point-mass assumptions for the slider and pendulum bob, plus a massless rod, are convenient for the Lagrangian but will be violated in a physical build; adding those inertias or showing that they are negligible is a natural extension the paper does not address.","If the ideal no-slip cone model were replaced by a contact model with a finite contact patch and slip, the same control law could be re-evaluated to show how much of the simulated tracking performance depends on the no-slip assumption.","The control demonstration is limited to smooth trajectories; extending the scheme to paths with corners or to 3D terrain would likely require a modified kinematic law for $\\dot\\psi$."],"forward_implications":["A physical Norma robot would need only the pendulum torque and the slider force as inputs to execute both forward motion and turns on flat ground.","Because the model is in $M(q)\\ddot q+C(q,\\dot q)\\dot q+G(q)=\\tau$ form, standard robot-control techniques such as computed torque, gain scheduling, or adaptive parameter estimation can be applied directly to it.","The pure-pursuit plus PID architecture gives a decoupled design procedure: choose desired roll speed and slider position from path error, then tune two independent PID loops for the actuators.","The simulations imply that, with the stated masses, lengths, and PID gains, tracking error converges to a small neighborhood of zero on a smooth path.","A working two-input design would make spherical robots simpler to build than multi-actuator gimbal or flywheel designs, since steering and propulsion come from the same internal mass shifts."],"supporting_citations":[{"why":"Supplies the equal-arc-length no-slip rule that converts cone rolling into the yaw-rate relation $\\dot\\psi=-\\dot\\theta\\,\\sin\\phi/(R\\cos\\phi-R\\sin\\phi)$.","marker":"[32]"},{"why":"Supplies the pure-pursuit method used to generate desired roll rate and slider displacement from path error.","marker":"[33]"},{"why":"Supplies the Christoffel-symbol construction used to build the Coriolis/centripetal matrix $C$ so that $\\dot M-2C$ is skew-symmetric.","marker":"[34]"}],"fun_headline_variants":["Slider and pendulum give a ball robot steering","A ball-bot that rolls and turns via two internal actuators","Norma's slider and pendulum enable nonholonomic motion","Two actuators drive both rolling and turning of a sphere"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is that a tilted sphere rolling on flat ground behaves exactly like a solid cone, without slipping, so each roll changes the heading by an amount fixed by the tilt angle; if that geometric rule is wrong, the yaw rate used to build the dynamics and the controller is wrong too.","fun_headline_variants_meta":{"raw":{"variants":["Slider and pendulum give a ball robot steering","A ball-bot that rolls and turns via two internal actuators","Norma's slider and pendulum enable nonholonomic motion","Two actuators drive both rolling and turning of a sphere"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000199,"raw_usage":{"total_tokens":1360,"prompt_tokens":922,"completion_tokens":438,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":538,"completion_tokens_details":{"reasoning_tokens":374}},"tokens_in":538,"tokens_out":438,"duration_ms":5324,"temperature":1.0,"reasoning_tokens":374,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T14:49:18.393303+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Lock the slider at a known off-center position so the ball tilts by $\\phi$, roll it forward at a roughly constant roll rate $\\dot\\theta$ on a flat hard floor, and measure the heading $\\psi$ over several full revolutions. Compare $\\dot\\psi$ with $-\\dot\\theta\\,\\sin\\phi/(R\\cos\\phi-R\\sin\\phi)$; a disagreement larger than the measurement error would refute the rolling-cone model and, with it, the derived dynamics and the simulated tracking results.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the equal-arc-length no-slip rule that converts cone rolling into the yaw-rate relation $\\dot\\psi=-\\dot\\theta\\,\\sin\\phi/(R\\cos\\phi-R\\sin\\phi)$."},{"cited_title":"Corke, Robotics, Vision and Control: Fundamental Algorithms In MATLAB® Second, Completely Revised","cited_arxiv_id":null,"evidence_quote":"Supplies the pure-pursuit method used to generate desired roll rate and slider displacement from path error."},{"cited_title":"Modern Robotics-Mechanics, Planning, and Control: Video supplements and software,","cited_arxiv_id":null,"evidence_quote":"Supplies the Christoffel-symbol construction used to build the Coriolis/centripetal matrix $C$ so that $\\dot M-2C$ is skew-symmetric."}],"review_version":1}