Pith. sign in

REVIEW 4 major objections 5 minor 26 references

Tilt-Ropter, a fully actuated tilt-rotor quadcopter with passive wheels, proves that one NMPC can track trajectories in air, on the ground, and across transitions, with ground power at 7.2% of flight.

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

T0 review · deepseek-v4-flash

2026-08-03 05:34 UTC pith:WAIN6CK5

load-bearing objection A genuinely new hybrid platform with a coherent control stack; the main weaknesses are an unexposed allocation matrix and sloppy estimator equations, both fixable. the 4 major comments →

arxiv 2602.01700 v3 pith:WAIN6CK5 submitted 2026-02-02 cs.RO

Tilt-Ropter: A Fully Actuated Hybrid Aerial-Terrestrial Vehicle with Tilt Rotors and Passive Wheels

Ruoyu Wang , Xuchen Liu , Zongzhou Wu , Zixuan Guo , Wendi Ding , Ben M. Chen This is my paper
classification cs.RO
keywords hybrid aerial-terrestrial vehicletilt rotorfully actuatednonlinear model predictive controlpassive wheelsexternal wrench estimationtrajectory trackingenergy efficiency
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

This paper presents Tilt-Ropter, a hybrid aerial-terrestrial vehicle whose four tilt rotors and two passive wheels give it full six-degree-of-freedom actuation—decoupled force and torque control—rather than the underactuated coupling of a standard quadrotor. The authors argue that this full actuation lets the same nonlinear model predictive controller (NMPC) track reference trajectories in the air, on the ground, and through smooth air-ground transitions, while an external wrench estimator compensates for contact forces and disturbances. The central empirical evidence is low tracking errors—RMSE 0.052 m in flight, 0.145 m on the ground, and 0.125 m across transition—and ground locomotion consuming only 7.2% of flight-phase power, a 92.8% reduction. If the claims hold, the platform points toward long-endurance missions that combine efficient rolling traversal with aerial maneuverability. The load-bearing premise is that the control allocation matrix mapping wrenches to eight lateral/vertical thrust components is full-rank and that the resulting intermediate thrust vector is realizable by the four rotor-servo pairs.

Core claim

Tilt-Ropter is claimed to be a fully actuated HATV: by tilting each of four rotors independently via centrally mounted servos, the platform can generate arbitrary six-dimensional wrenches. The authors derive a static 6x8 allocation matrix A that maps a desired wrench W to an intermediate vector T of eight thrust components (lateral and vertical per rotor), computed via the Moore-Penrose pseudoinverse T = A†W. This decoupling underpins a unified NMPC that optimizes the wrench derivative rather than raw actuator commands, implicitly bounding servo rates and activating ground-specific non-holonomic constraints through an indicator function. An external wrench estimator, refined with a first-ord

What carries the argument

The central object is the static control allocation matrix A (6x8) that connects the wrench W to an eight-dimensional thrust vector T whose entries are the lateral and vertical components of each rotor's thrust. Because A is independent of tilt angle, the mapping from wrench to motor speeds and servo angles is T = A†W, followed by Ti = sqrt(T_i,l^2 + T_i,v^2) and alpha_i = atan2(T_i,v, T_i,l). This allocation carries the argument in two ways: it converts the decoupled force/torque capability into feasible actuator commands, and it yields an explicit relation between wrench-change rate and servo-angle rate, which the NMPC exploits by taking u = (F_dot, M_dot) as the control input and bounding

Load-bearing premise

The control allocation matrix A has full row rank and is well-conditioned, so every desired wrench maps to a realizable set of the eight lateral/vertical thrust components and hence to feasible rotor speeds and servo angles.

What would settle it

Compute the singular values of the allocation matrix A (or its rank) from the robot's stated geometry; if A is rank-deficient or extremely ill-conditioned, the decoupled force/torque claim collapses. Alternatively, command two conflicting wrench setpoints at hover, such as zero net force with a large yaw torque alongside a large lateral force with zero torque, and check whether the robot tracks both independently within actuator limits.

Watch this falsifier — get emailed when new claim-graph text bears on it.

If this is right

  • Ground locomotion at 7.2% of flight power means long-endurance missions can be planned as mostly rolling with short flights over obstacles.
  • A single NMPC spanning flight, ground, and transitions removes the need for mode-specific controllers or heuristic switching logic.
  • Wrench-rate NMPC with explicit servo-rate bounds should prevent the oscillations that arise when servos lag fast wrench changes.
  • External wrench estimation from actuator measurements makes the platform robust to uneven terrain and disturbances without adding force sensors.
  • Decoupled force/torque control opens the way to physical interaction tasks such as pushing, perching, or manipulation while in contact.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The 92.8% power reduction is specific to the tested trajectory and hardware; a fairer comparison would match speeds and terrain to quantify the trade-off between added mass and ground efficiency.
  • If A is indeed full-rank, the same wrench-rate NMPC formulation could generalize to other overactuated tilt-rotor designs simply by substituting their allocation matrices.
  • Using the tilt arms as wheel axles is a clever mass-saving choice, but it couples wheel-ground contact forces into the tilt mechanism; the external wrench estimator may be essential for compensating that coupling.
  • A direct test of full actuation—commanding independent force and torque setpoints at hover—would make the decoupling claim measurable and falsifiable beyond trajectory tracking.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit.

Referee Report

4 major / 5 minor

Summary. The paper presents Tilt-Ropter, a tilt-rotor quadrotor with two passive wheels, and claims that this design is fully actuated, i.e., capable of generating arbitrary six-dimensional wrenches. The authors develop a unified NMPC that outputs wrench derivatives, a control allocation scheme based on an 8-dimensional thrust vector, and an external wrench estimator intended to handle wheel-ground contact. Validation is provided through Gazebo SIM and real-world experiments, including aerial/ground trajectory tracking, a seamless air-ground transition, and a power-consumption comparison reporting a 92.8% reduction on the ground. The central claims are full actuation, low tracking error, and high energy efficiency.

Significance. If the technical claims are substantiated, Tilt-Ropter would be a useful contribution to the HATV literature: a lightweight fully actuated platform with passive wheels and a unified NMPC could enable efficient multimodal locomotion with independent force and torque control. The measured power reduction is striking and the design choice of using tilt arms as wheel axles is elegant. However, the paper's primary theoretical claim rests on an unverified control allocation matrix, and the external wrench estimator is presented inconsistently. The experimental results, while suggestive, are limited to single trajectories and do not explicitly demonstrate decoupled wrench generation.

major comments (4)
  1. [Sec. IV-C, Eqs. (2)-(3)] The matrix A is the load-bearing element for the 'fully actuated' claim, but its entries are never given and its rank is never checked. The use of T=A†W is only valid if A has full row rank and reasonable conditioning. Without specifying A, its relation to the arm geometry in Sec. III-A, or at least a singular-value/condition-number check, the statement that Tilt-Ropter 'is a fully actuated aerial vehicle, capable of producing arbitrary six-dimensional wrenches' is an unverified algebraic assumption. Please provide A explicitly, verify full row rank, and show that the wrench bounds in Table II remain feasible after allocation.
  2. [Sec. VI, Eqs. (11)-(12)] These two equations are not equivalent. Differentiating the first line of Eq. (11) yields dot F_hat_e = K_f(m a - F - F_hat_e), not K_f(F - F_hat_e). The second line similarly produces a term in J dot omega and a sign inconsistency, so the low-pass filter form in Eq. (12) does not follow from Eq. (11). The external wrench estimator is a stated contribution and is used in the dynamics of Eq. (1c)-(1d); the inconsistency is therefore load-bearing. Please correct the derivation, state precisely which estimator is implemented, and clarify how the servo dynamics of Eq. (13) enter.
  3. [Sec. V-A, Eqs. (7)-(10)] The state vector x includes F and M, with control input u = [dot F, dot M], but the dynamics constraint in Eq. (8b) says f is the discrete-time formulation of Eqs. (1a)-(1d), which do not contain dynamics for F or M. The integrator equations dot F = u_F and dot M = u_M must be included explicitly for the NMPC to be well posed. In addition, the tracking error in Eq. (10) includes F_k and M_k, but no reference wrench F_r, M_r is ever defined; the paper must state whether these are penalized to zero, to a trim wrench, or to some other reference.
  4. [Sec. VII-B and Table II] The reported experiments do not actually exercise the fully actuated capability. The body-frame force constraints in Table II are F_min = (-0.2, -0.2, 0.0) N and F_max = (0.2, 0.2, 20) N, so the horizontal force is essentially clamped to near zero; the aerial figure-eight trajectory can be tracked by a conventional quadrotor using body tilt and a single thrust direction. No experiment demonstrates independent wrench generation, e.g., tracking a nonzero lateral force at zero roll/pitch or applying a pure yaw torque step. Without such an experiment, the paper's primary contribution—decoupled force/torque control—is not validated experimentally.
minor comments (5)
  1. [Sec. IV-C, Eq. (6)] The expression for dot alpha_i appears to be the negative of the standard derivative of atan2(y,x) = (x dot y - y dot x)/(x^2+y^2). If the sign is used in a signed servo-rate constraint, this is an error; if only the magnitude is used, please state that explicitly.
  2. [Sec. IV-B / Sec. VI] The definition of measured acceleration a = q^{-1} ⊙ (dot v + g) in Eq. (11) is inconsistent with the dynamics in Eq. (1c), where gravity appears outside the quaternion rotation: dot v = m^{-1} q ⊙ (F + F_e) + g. Check the sign and frame convention for gravity in the estimator.
  3. [Sec. V-A, Eq. (8b)] The indicator function δ in Eq. (9) uses exact equality p_z = r to decide ground contact. In practice this will cause chattering; use a tolerance band or smooth activation function.
  4. [Table II] The NMPC weight matrices Q, R, and Q_N are not reported. Since the paper emphasizes the controller's performance, at least the relative scaling should be given for reproducibility.
  5. [Sec. VII-B] The experimental results are presented as single trials without error bars or statistical repetition. Given the strong quantitative claims (e.g., 0.052 m RMSE, 92.8% power reduction), the authors should report repeated trials and variance, or temper the claims accordingly.

Circularity Check

0 steps flagged

No significant circularity: experimental tracking and power results are measured, and the fitted servo time constant is identified independently and is not used to fit the reported outcomes.

full rationale

The paper's derivation chain is self-contained in the sense required for a circularity finding: the headline quantities (tracking RMSEs of 0.052 m, 0.145 m, and 0.125 m; power 653.64 W vs 47.38 W; the 92.8% reduction) are measured from motion capture and battery voltage/current, not recovered from fitted model constants. The only fitted parameter of consequence, the servo time constant tau=0.08 s, is identified by least-squares fitting to an independent step response (Sec. VI, Fig. 6) and used in the external wrench estimator; it is not adjusted to match the trajectory or power results. The external wrench estimator is adopted from an external prior work [26], not from a same-author citation chain. The central 'fully actuated' claim is formulated via W=AT and T=A†W (Eqs. 2-3), but the allocation matrix A is never written out and its full row rank is never verified; that is an under-specification/verification gap, not circularity, because the claim is not equivalent to any fitted output. No prediction in the paper reduces by construction to an input, and no load-bearing argument rests on a self-citation chain.

Axiom & Free-Parameter Ledger

5 free parameters · 6 axioms · 0 invented entities

The central claims rest on standard rigid-body dynamics, a lumped rotor-thrust model, an unproven full-rank allocation matrix, a simplified contact model, and an estimated (not measured) servo state. The only numerically fitted quantity of consequence is the servo time constant, though several controller gains and geometric constants are left as tuning or calibration values.

free parameters (5)
  • Thrust coefficient c_t = not reported
    Appears in Eq. (2) and (5) as the constant mapping rotor speed squared to thrust; it must be identified or calibrated for the platform, but no value or measurement is given.
  • Servo time constant tau = 0.08 s
    Identified by least-squares fitting to a step response (Fig. 6), used to estimate actual servo angles because the servos lack angle feedback. The fit uncertainty is not reported.
  • External wrench estimator gains K_f, K_m = not reported
    Gains in Eq. (11) control convergence and noise filtering; no values or tuning procedure are given.
  • NMPC cost weights Q, R, Q_N = not reported
    These weights determine tracking performance and are tuned; only constraint bounds are shown in Table II, so the reported RMSE cannot be reproduced or compared.
  • Wheel radius r = not reported
    Used in the ground-contact height constraint p_z = r in Eq. (8b)-(9); the physical value is never stated, although it is a design input.
axioms (6)
  • domain assumption Rigid-body model with lumped mass m and inertia J (Eq. 1).
    The vehicle is treated as a single rigid body with point CoM; structural flexibilities and wheel dynamics other than contact are neglected.
  • domain assumption Thrust is proportional to rotor speed squared and decomposes via tilt angle into lateral and vertical components.
    Standard first-order rotor model; rotor angular acceleration and gyroscopic effects are explicitly neglected in Section IV-C.
  • ad hoc to paper A (6x8) has full row rank, enabling arbitrary 6-D wrench generation.
    The pseudoinverse T=A^†W is used in Eq. (3) without presenting A or checking its rank/conditioning; this underlies the 'fully actuated' claim.
  • domain assumption Ground locomotion can be captured by binary non-holonomic constraints v_B,y=0, omega_x=0, p_z=r, switched by delta in Eq. (9).
    Assumes no lateral wheel slip, no roll, and constant ground height; no friction or terrain model is included.
  • domain assumption Servo dynamics are first-order with known tau; estimated servo angles can be simulated from commanded angles.
    Hobby servos lack angle feedback; Eq. (13) and tau from step-response identification are used to reconstruct actual angles.
  • domain assumption External wrench estimator converges under a first-order low-pass filter.
    Eq. (12) is asserted as a low-pass filter; convergence and stability properties are not analyzed.

pith-pipeline@v1.3.0-alltime-deepseek · 10549 in / 17340 out tokens · 164976 ms · 2026-08-03T05:34:12.582332+00:00 · methodology

0 comments
read the original abstract

In this work, we present Tilt-Ropter, a fully actuated hybrid aerial-terrestrial vehicle (HATV) that integrates tilt rotors with passive wheels to enable efficient multi-modal locomotion. Unlike conventional underactuated HATVs, the fully actuated design of Tilt-Ropter allows decoupled force and torque control, improving maneuverability and ground locomotion efficiency. A unified nonlinear model predictive controller (NMPC) is developed to track reference trajectories, enforce non-holonomic constraints, and accommodate contact effects across locomotion modes, while ensuring actuator feasibility through dedicated control allocation. To address complex wheel-ground dynamics, an external wrench estimator is incorporated to provide real-time interaction wrench estimates. The system is validated through simulation and real-world experiments, including seamless air-ground transitions and trajectory tracking tasks. Experimental results demonstrate low tracking errors in both modes and reveal a 92.8% reduction in power consumption during ground locomotion compared with flight, highlighting the platform's potential for long-duration missions where energy efficiency is critical.

Figures

Figures reproduced from arXiv: 2602.01700 by Ben M. Chen, Ruoyu Wang, Wendi Ding, Xuchen Liu, Zixuan Guo, Zongzhou Wu.

Figure 1
Figure 1. Figure 1: Tilt-Ropter: a novel hybrid aerial-terrestrial vehicle. [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Exploded view of Tilt-Ropter key components. [PITH_FULL_IMAGE:figures/full_fig_p002_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Illustration of the design of the tilt arm. [PITH_FULL_IMAGE:figures/full_fig_p003_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Definition of coordinate frames. B. Rigid Body Dynamics Assuming that Tilt-Ropter can be modeled as a rigid body with mass m and inertia J, which is driven by the total force F and torque M of actuators acting on the CoM, and the gravity g = [0, 0, −9.81] m · s −2 . p˙ = v (1a) q˙ = 1 2 q ⊗  0 ω  (1b) v˙ = m−1 q ⊙ (F + Fe) + g (1c) ω˙ = J −1 (M + Me − ω × (Jω)) (1d) [PITH_FULL_IMAGE:figures/full_fig_p00… view at source ↗
Figure 5
Figure 5. Figure 5: Control architecture of Tilt-Ropter. pm and qm are the measured position and quaternion of the robot from motion capture system. Ωm is the measured rotor speeds from ESC. A. Nonlinear Model Predictive Control The control problem is formulated as a finite-time optimal control problem (OCP), where the objective is to minimize the tracking error of the robot state and input to the ref￾erence trajectory while … view at source ↗
Figure 6
Figure 6. Figure 6: System identification of the servo with attached motor and propeller. [PITH_FULL_IMAGE:figures/full_fig_p006_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: The structure of hardware-in-the-loop simulation. [PITH_FULL_IMAGE:figures/full_fig_p006_7.png] view at source ↗
Figure 8
Figure 8. Figure 8: Trajectories of the Tilt-Ropter in aerial and ground modes. (a) [PITH_FULL_IMAGE:figures/full_fig_p007_8.png] view at source ↗
Figure 9
Figure 9. Figure 9: Hybrid trajectory of Tilt-Ropter showcasing seamless air-ground [PITH_FULL_IMAGE:figures/full_fig_p007_9.png] view at source ↗
Figure 10
Figure 10. Figure 10: Comparison of power consumption between aerial flight and [PITH_FULL_IMAGE:figures/full_fig_p008_10.png] view at source ↗

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

26 extracted references · 5 linked inside Pith

  1. [1]

    Modeling and performance assessment of the hytaq, a hybrid terrestrial/aerial quadrotor,

    A. Kalantari and M. Spenko, “Modeling and performance assessment of the hytaq, a hybrid terrestrial/aerial quadrotor,”IEEE Transactions on Robotics, vol. 30, no. 5, pp. 1278–1285, 2014

    2014

  2. [2]

    Autonomous and adaptive navigation for terrestrial-aerial bimodal vehicles,

    R. Zhang, Y . Wu, L. Zhang, C. Xu, and F. Gao, “Autonomous and adaptive navigation for terrestrial-aerial bimodal vehicles,”IEEE Robotics and Automation Letters, vol. 7, no. 2, pp. 3008–3015, 2022

    2022

  3. [3]

    Autonomous hybrid ground/aerial mo- bility in unknown environments,

    D. D. Fan, R. Thakker, T. Bartlett, M. B. Miled, L. Kim, E. Theodorou, and A.-a. Agha-mohammadi, “Autonomous hybrid ground/aerial mo- bility in unknown environments,” in2019 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS). IEEE, 2019, pp. 3070–3077

    2019

  4. [4]

    Model-based planning and control for terrestrial-aerial bimodal vehi- cles with passive wheels,

    R. Zhang, J. Lin, Y . Wu, Y . Gao, C. Wang, C. Xu, Y . Cao, and F. Gao, “Model-based planning and control for terrestrial-aerial bimodal vehi- cles with passive wheels,” in2023 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS). IEEE, 2023, pp. 1070– 1077

    2023

  5. [5]

    Unified terres- trial/aerial motion planning for hytaqs via nmpc,

    T. Wu, Y . Zhu, L. Zhang, J. Yang, and Y . Ding, “Unified terres- trial/aerial motion planning for hytaqs via nmpc,”IEEE Robotics and Automation Letters, vol. 8, no. 2, pp. 1085–1092, 2023

    2023

  6. [6]

    Skywalker: A compact and agile air-ground omnidirectional vehicle,

    N. Pan, J. Jiang, R. Zhang, C. Xu, and F. Gao, “Skywalker: A compact and agile air-ground omnidirectional vehicle,”IEEE Robotics and Automation Letters, vol. 8, no. 5, pp. 2534–2541, 2023

    2023

  7. [7]

    Aircrab: A hybrid aerial-ground manipulator with an active wheel,

    M. Cao, J. Zhao, X. Xu, and L. Xie, “Aircrab: A hybrid aerial-ground manipulator with an active wheel,”arXiv preprint arXiv:2403.15805, 2024

    Pith/arXiv arXiv 2024

  8. [8]

    A hybrid quadrotor with a passively reconfigurable wheeled leg capable of robust terrestrial maneuvers,

    S. Yu, B. Pu, K. Dong, S. Bai, and P. Chirarattananon, “A hybrid quadrotor with a passively reconfigurable wheeled leg capable of robust terrestrial maneuvers,”IEEE Robotics and Automation Letters, 2025

    2025

  9. [9]

    Sytab: A class of smooth-transition hybrid terrestrial/aerial bicopters,

    J. Yang, Y . Zhu, L. Zhang, Y . Dong, and Y . Ding, “Sytab: A class of smooth-transition hybrid terrestrial/aerial bicopters,”IEEE Robotics and Automation Letters, vol. 7, no. 4, pp. 9199–9206, 2022

    2022

  10. [10]

    Hybrid aerial-ground locomo- tion with a single passive wheel,

    Y . Qin, Y . Li, X. Wei, and F. Zhang, “Hybrid aerial-ground locomo- tion with a single passive wheel,” in2020 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS). IEEE, 2020, pp. 1371–1376

    2020

  11. [11]

    Skater: A novel bi-modal bi-copter robot for adaptive locomotion in air and diverse terrain,

    J. Lin, R. Zhang, N. Pan, C. Xu, and F. Gao, “Skater: A novel bi-modal bi-copter robot for adaptive locomotion in air and diverse terrain,” IEEE Robotics and Automation Letters, 2024

    2024

  12. [12]

    Design of multirotor aerial vehicles: A taxonomy based on input allocation,

    M. Hamandi, F. Usai, Q. Sabl ´e, N. Staub, M. Tognon, and A. Franchi, “Design of multirotor aerial vehicles: A taxonomy based on input allocation,”The International Journal of Robotics Research, vol. 40, no. 8-9, pp. 1015–1044, 2021

    2021

  13. [13]

    Active interaction force control for contact- based inspection with a fully actuated aerial vehicle,

    K. Bodie, M. Brunner, M. Pantic, S. Walser, P. Pf ¨andler, U. Angst, R. Siegwart, and J. Nieto, “Active interaction force control for contact- based inspection with a fully actuated aerial vehicle,”IEEE Transac- tions on Robotics, vol. 37, no. 3, pp. 709–722, 2020

    2020

  14. [14]

    Towards efficient full pose omnidirectionality with overactuated mavs,

    K. Bodie, Z. Taylor, M. Kamel, and R. Siegwart, “Towards efficient full pose omnidirectionality with overactuated mavs,” inProceedings of the 2018 International Symposium on Experimental Robotics. Springer, 2020, pp. 85–95

    2018

  15. [15]

    Multi-modal mobility morphobot (m4) with appendage repurpos- ing for locomotion plasticity enhancement,

    E. Sihite, A. Kalantari, R. Nemovi, A. Ramezani, and M. Gharib, “Multi-modal mobility morphobot (m4) with appendage repurpos- ing for locomotion plasticity enhancement,”Nature communications, vol. 14, no. 1, p. 3323, 2023

    2023

  16. [16]

    Atmo: An aerially transforming morphobot for dynamic ground-aerial transition,

    I. Mandralis, R. Nemovi, A. Ramezani, R. M. Murray, and M. Gharib, “Atmo: An aerially transforming morphobot for dynamic ground-aerial transition,”arXiv preprint arXiv:2503.00609, 2025

    Pith/arXiv arXiv 2025

  17. [17]

    Mtabot: An efficient morphable terrestrial-aerial robot with two transformable wheels,

    K. Shi, Z. Jiang, L. Ma, L. Qi, and M. Jin, “Mtabot: An efficient morphable terrestrial-aerial robot with two transformable wheels,” IEEE Robotics and Automation Letters, vol. 9, no. 2, pp. 1875–1882, 2024

    2024

  18. [18]

    Doublebee: A hybrid aerial-ground robot with two active wheels,

    M. Cao, X. Xu, S. Yuan, K. Cao, K. Liu, and L. Xie, “Doublebee: A hybrid aerial-ground robot with two active wheels,” in2023 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS). IEEE, 2023, pp. 6962–6969

    2023

  19. [19]

    A novel overactuated quadrotor unmanned aerial vehicle: Modeling, control, and experimen- tal validation,

    M. Ryll, H. H. B ¨ulthoff, and P. R. Giordano, “A novel overactuated quadrotor unmanned aerial vehicle: Modeling, control, and experimen- tal validation,”IEEE Transactions on Control Systems Technology, vol. 23, no. 2, pp. 540–556, 2014

    2014

  20. [20]

    V oliro: An omnidirectional hexa- copter with tiltable rotors,

    M. Kamel, S. Verling, O. Elkhatib, C. Sprecher, P. Wulkop, Z. Taylor, R. Siegwart, and I. Gilitschenski, “V oliro: An omnidirectional hexa- copter with tiltable rotors,”arXiv preprint arXiv:1801.04581, 2018

    Pith/arXiv arXiv 2018

  21. [21]

    Servo integrated nonlinear model predictive control for overactuated tiltable-quadrotors,

    J. Li, J. Sugihara, and M. Zhao, “Servo integrated nonlinear model predictive control for overactuated tiltable-quadrotors,”IEEE Robotics and Automation Letters, 2024

    2024

  22. [22]

    A multimodal agile land-air aircraft (alaa) that can fly, roll, and stand,

    Q. Guo, Z. Guo, Y . Shi, Z. Zhou, and D. Ma, “A multimodal agile land-air aircraft (alaa) that can fly, roll, and stand,”Journal of Field Robotics, 2024

    2024

  23. [23]

    Tactv: A class of hybrid terrestrial/aerial coaxial tilt-rotor vehicles,

    Y . Dong, Y . Zhu, L. Zhang, and Y . Ding, “Tactv: A class of hybrid terrestrial/aerial coaxial tilt-rotor vehicles,”arXiv preprint arXiv:2411.12359, 2024

    Pith/arXiv arXiv 2024

  24. [24]

    Tj-flyingfish: Design and implementation of an aerial-aquatic quadrotor with tiltable propulsion units,

    X. Liu, M. Dou, D. Huang, S. Gao, R. Yan, B. Wang, J. Cui, Q. Ren, L. Dou, Z. Gao,et al., “Tj-flyingfish: Design and implementation of an aerial-aquatic quadrotor with tiltable propulsion units,” in2023 IEEE International Conference on Robotics and Automation (ICRA). IEEE, 2023, pp. 7324–7330

    2023

  25. [25]

    Mpc with learned residual dynamics with application on omnidirectional mavs,

    M. Brunner, W. Zhang, A. Roumie, M. Tognon, and R. Siegwart, “Mpc with learned residual dynamics with application on omnidirectional mavs,”arXiv preprint arXiv:2207.01451, 2022

    Pith/arXiv arXiv 2022

  26. [26]

    External wrench estimation, collision detection, and reflex reaction for flying robots,

    T. Tomi ´c, C. Ott, and S. Haddadin, “External wrench estimation, collision detection, and reflex reaction for flying robots,”IEEE Trans- actions on Robotics, vol. 33, no. 6, pp. 1467–1482, 2017

    2017