Pith. sign in

REVIEW 3 major objections 5 minor 30 references

AeroThrow: An Autonomous Aerial Throwing System for Precise Payload Delivery

T0 review · 3 major / 5 minor · reviewed 2026-08-06 · deepseek-v4-flash

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

Pith's one-line read A quadrotor carrying a delta arm can throw payloads to a target with roughly ten-centimeter accuracy even in aggressive flight, by planning a release-time window and re-timing the drop online from predictive states.

desk verdict Solid aerial-throwing system paper with real experiments; the ballistic-model bias and missing error bars are fixable, but the core release-window idea is sound. read the letter →

arxiv 2507.13903 v1 pith:AGGDOLH4 submitted 2025-07-18 cs.RO

Ziliang Li , Hongming Chen , Yiyang Lin , Biyu Ye , Ximin Lyu This is my paper · ORCID
classification cs.RO
keywords aerialthrowingairdropmanipulatortrajectoryoptimizationnonlinearmodelpredictivecontrolreleasetimingreassessmentdisturbanceobserverpayloaddelivery
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

Autonomous airdrops from agile drones are hard because the exact instant of payload release strongly determines where the payload lands, and drones cannot track fast trajectories perfectly. This paper proposes a throwing system that makes the planned trajectory insensitive to release timing: instead of enforcing a landing constraint at a single release instant, the planner applies a smooth landing penalty over an entire time window around the release, so any release within that window still lands close to the target. The reference trajectory is tracked by a nonlinear model predictive controller with hierarchical disturbance compensation, and an online algorithm continuously re-evaluates the best release time from the controller's predicted states. In real flight tests the most aggressive trajectory landed within 10.1 cm mean error with the proposed method, versus 76.7 cm when releasing at the nominal time. If the approach scales beyond the tested platform, it would make fast, accurate aerial delivery practical without requiring the drone to hover or stop.

What carries the argument

The central object is a smooth relaxation function $L_\mu[x]$ that activates a landing-error penalty $G_l(t) = L_\mu[E_t+\tau]\|P_l - p_t^W\|^2$ over a time interval $\tau$ around the nominal release time, rather than at a single instant. The interval is inserted into the MINCO-based trajectory optimization, so the output trajectory contains a continuous set of feasible release times $\mathcal{T}_r = \{t \mid t \in [t_r-\tau, t_r+\tau]\}$. This is what converts a throw into a timing-tolerant task. The control stack then carries the argument: an NMPC over the quadrotor model tracks the reference, a nonlinear disturbance observer estimates the force change when the unknown-mass payload is dropped, an INDI inner loop rejects the configuration-dependent moment from the arm, and an online release-timing reassessment (Algorithm 1) queries the NMPC's predicted trajectory, evaluates the projectile landing point for every horizon state via Equations (19a)-(19b), and chooses the release instant with minimum predicted landing error.

What would settle it

Measure the actual trajectory of the 2 cm iron cube with a high-speed motion-capture system immediately after release and compare it with the ballistic prediction from Equations (4)-(6); if the observed flight path deviates systematically with release speed or payload orientation, or if wind-tunnel tests show landing bias that grows with crosswind speed, the point-mass assumption is falsified. A simpler check is to run the same planned trajectory with a payload of equal mass but larger cross-section and see whether the landing error grows, which it would if drag, rather than timing, dominates the residual error.

Watch

Extended reading notes

Core claim

The core claim is that continuous, interval-based constraints on the parabolic landing point are the key to accurate aerial throwing. Point-wise landing constraints, as used in earlier work, create a trajectory that only works if release happens at one precise instant; any timing error translates directly into landing error. By smoothing the landing-penalty activation over a time window $\tau$ around the optimal release time, the planner produces a feasible release interval $\mathcal{T}_r$, and the optimizer exploits the spatial redundancy of the arm to slow or reposition the drone so release uncertainty has little effect. On the control side, a hierarchical disturbance compensation scheme, combining a nonlinear disturbance observer for linear forces with incremental nonlinear dynamic inversion for the payload-induced moment, is embedded in an NMPC framework. The NMPC's predictions are then used online to reassess the release timing: for every predicted state in the horizon the algorithm simulates the ballistic landing point and picks the instant that minimizes landing error. The paper reports that this combined system reduces landing error on the most aggressive tested trajectory from 76.7/83.8 cm (MEAN/MAX) with a nominal trigger to 10.1/12.3 cm with the proposed reassessment.

Load-bearing premise

The precision claim rests on the simplified projectile model: the payload is assumed to leave the end-effector with zero relative motion and then to fly under gravity alone, with no horizontal forces, so any drag, spin, or residual electromagnet effect would bias the predicted landing points and the release-time choice.

Editorial extensions

If this is right

  • Drones can deliver payloads while flying fast, because the release window makes landing accuracy robust to timing jitter and actuator delay.
  • The same trajectory can be adjusted from a throw to a near place by increasing $\tau$, giving mission planners a continuous trade-off between flight aggressiveness and precision.
  • The system tolerates unknown payload mass, since the disturbance observer and INDI loop absorb the model change at release without retuning.
  • Prediction-based release re-timing can correct for accumulated tracking errors in real time, which is the main reason aggressive trajectories improve from 76.7 cm to 10.1 cm error.
  • The approach narrows the gap between agile flight and delivery tasks, so airdrop no longer requires the drone to slow or hover over the target.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • The interval-relaxation trick is generic: the same smoothed activation over a time window could make catching, perching, or ball-striking tasks robust to actuation delay, which is an extension the paper does not develop.
  • The centimeter-level numbers are tied to the point-mass ballistic model; for larger or non-spherical payloads, or in wind, aerodynamic drag would bias the predicted landing points, so the method would likely need a learned drag or wind model to keep its accuracy outdoors.
  • A direct test of the method's headroom is to replace the projectile prediction in Equations (19a)-(19b) with a learned residual map from release state to landing offset; if accuracy improves further, the residual error is model bias rather than timing, and if it does not, timing uncertainty is already the dominant term.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

3 major / 5 minor

Summary. The paper proposes an autonomous aerial throwing system based on a quadrotor-plus-delta-arm manipulator. The core ideas are: (i) a trajectory planner that imposes a smooth, time-windowed penalty on the predicted parabolic landing point, generating a feasible release interval rather than a single release instant; (ii) an NMPC controller augmented with NDOB and INDI hierarchical disturbance compensation; and (iii) an online release-timing reassessment algorithm that uses the NMPC prediction horizon to re-select the release time based on predicted landing errors. The paper reports simulation and real-world experiments, including an ablation of the control modules and comparative precision trials. The headline result is that, for an aggressive trajectory, the proposed reassessment reduces mean/max landing error from 76.7/83.8 cm to 10.1/12.3 cm.

Significance. If the central claims hold, the paper makes a useful contribution to autonomous airdrop: the interval-valued landing constraint is a sensible extension of point-wise landing constraints and seems to reduce sensitivity to release timing, while the NMPC-based reassessment is a practical way to exploit model-predictive information for triggering. The real-world validation on a physical aerial manipulator, including an ablation of NDOB and INDI, is valuable and goes beyond simulation-only studies. However, the strength of the experimental evidence is limited by missing statistical reporting, and the online reassessment depends on an unquantified ballistic-model assumption. The conceptual contribution is clear, but the load-bearing experimental and modeling evidence needs strengthening before the precision claims can be accepted at face value.

major comments (3)
  1. [Section V-C, Table II and Section V-B, Table I] The central precision claims rest on MEAN/MAX (or RMSE/MAX) values with no trial counts, standard deviations, or confidence intervals. Figure 8 indicates ten throws for one trajectory, but Table II does not state n for each row, and Table I is similarly unspecified. With small sample sizes, the large improvement on the aggressive trajectory (76.7 to 10.1 cm) cannot be statistically distinguished from run-to-run variation. Please report per-trial landing errors, n, mean±SD, and confidence intervals or a suitable test for every condition.
  2. [Section III-C and Section IV-C, Eqs. (19a)-(19b) and Algorithm 1] The online release-timing reassessment selects the release instant using predicted landing points computed from a drag-free point-mass projectile model. The released object is a 2 cm iron cube, and the platform flies in its own rotor downwash; spin, residual magnetization, and aerodynamic forces can bias every predicted landing point. Because the same biased model selects the release time, the error does not average out across the prediction horizon. The claimed landing errors (3–10 cm) are the same order as plausible aerodynamic biases, yet the paper provides no calibration of predicted versus observed landing points and no sensitivity analysis. Please add a validation dataset of predicted vs. actual landing positions, or analyze how much the selected release time and final landing error change under bounded model perturbations.
  3. [Section IV-A, Eqs. (8)-(10) and Section V-A] The claim that the planned trajectories are 'less sensitive to release timing' is not quantitatively demonstrated. Figure 6 shows durations below unspecified error thresholds, but the text does not report the landing error as a function of release-time offset for different values of τ, or the relationship between τ, μ, and the achieved landing tolerance. To support the central contribution, please provide quantitative sensitivity curves, e.g., landing error versus release timing error for the optimized trajectories with and without the interval constraint.
minor comments (5)
  1. [Notation, Eq. (9) and Algorithm 1] The symbol τ is used both for the release-window half-width in Eq. (9) and for the current time/loop variable in Algorithm 1, which is confusing; please use distinct symbols.
  2. [Eqs. (4)-(6)] The projectile equations mix vector and scalar notation: g is defined as a scalar absolute gravitational acceleration but appears in vector expressions. Please define the gravitational acceleration vector and the landing-plane coordinate explicitly.
  3. [Abstract and Introduction] There are several typographical issues, including 'UA V' with a stray space in the abstract and introduction; please proofread.
  4. [Section V-A, Fig. 6] The caption refers to 'different threshold values' without stating the actual thresholds; please report these values and how they were chosen.
  5. [Section IV-C, Algorithm 1] The stopping condition 'if ∆t ≤ dt' appears to break whenever k* = 1; please clarify the indexing of k* and explain how actuator delay is explicitly incorporated into the decision rule.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the landing-point planner and the online release-timing reassessment share the same ballistic model consistently, and the paper's self-citations are component reuse, not load-bearing circular support.

full rationale

The derivation chain is self-contained in the relevant sense. Trajectory planning uses the projectile model in Eqs. (4)-(6) to impose continuous landing-point penalties over a release window (Eqs. (9)-(10)), and the online release-timing reassessment uses the same model in Eqs. (19a)-(19b) and Eq. (20) to choose, within the NMPC prediction horizon, the release time that minimizes the predicted landing error. This is a consistent application of the same forward model at planning and execution time, not a case where a prediction is equivalent to its input by construction: no parameter is fitted to landing data and then renamed as a prediction, and the headline improvement in Table II is an empirical comparison against a nominal release trigger on real flight data. The self-citations [12] and [13] are used for delta-arm control design and for modeling corridor, kinematic, and dynamic constraints respectively; these are component reuse, while the central continuous-constraint idea is explicitly credited to external work [21]. None of the cited self-works is invoked to forbid alternatives, prove uniqueness, or supply the core precision claim. The unquantified drag-free point-mass simplification is a legitimate modeling risk, but it is not circularity: a shared approximate model used for both planning and online correction does not make the correction an input to itself. No circular step can be exhibited from the paper's own equations, so the appropriate finding is no significant circularity.

Assumptions & free parameters 4 free parameters · 4 assumptions · 0 invented entities

The paper introduces no new physical entities, forces, or conserved quantities. Its contribution is an integrated control and planning architecture built from standard components (MINCO, NMPC, NDOB, INDI). The free parameters listed are the central tuning knobs that the claimed performance depends on, none of which are reported with concrete values or systematic tuning procedures.

free parameters (4)
  • Release-window half-width tau = not reported
    Controls the size of the feasible release interval and directly trades throwing agility against drop-like behavior. It is central to the claim of reduced release-timing sensitivity, but no value or selection criterion is given in the text.
  • Relaxation smoothness mu = not reported
    Sets the smoothness of the landing penalty activation in Eq. (10). Sec. V-A shows that varying mu shifts the release action between throw and place, yet no fixed value or tuning rule is reported.
  • NDOB observer gain c = not reported
    In Eq. (13): 'the parameter c should be adjusted based on the actual operating conditions'. It determines disturbance estimation convergence and is hand-tuned, directly affecting the claimed 54% improvement from NDOB in Table I.
  • NMPC horizon h and weights Q, Qu, QN = not reported
    Used in Eq. (11) and in the assumption |t*_r - tr| <= h in Eq. (18). The prediction quality and the release timing reassessment depend on these values, which are not disclosed.
assumptions (4)
  • domain assumption Payload is a point mass with state equal to the end-effector state at release, and after release it experiences no horizontal forces and only gravity vertically.
    Sec. III-C, Eqs. (4)-(6), and used again in Eq. (19) for online reassessment. The drag-free ballistic model is load-bearing for every predicted landing point; if drag or spin is significant, the predicted landing locations are biased.
  • domain assumption The actual optimal release time t*_r lies within the NMPC prediction horizon of the reference release time: |t*_r - tr| <= h.
    Eq. (18), Sec. IV-C. If tracking errors or disturbances push the optimal release time beyond the horizon, the reassessment algorithm cannot find it and the method degrades to nominal timing.
  • domain assumption The quadrotor and delta-arm system is well captured by the standard differential-flat dynamics, with the delta arm's rotational effect approximated by the quadrotor's rotation.
    Eqs. (1)-(3), Sec. III-B. The planner and NMPC rely on this model; manipulator flexibility, joint compliance, and pendulum effects are treated as disturbances rather than explicitly modeled.
  • domain assumption The NDOB disturbance estimate f_ext computed from IMU acceleration and the thrust model is accurate enough for compensation.
    Eqs. (12)-(13). The estimate depends on filtering, the known thrust coefficients, and the gain c; IMU noise or inaccurate rotor models will bias the compensation and degrade landing precision.

how reviews work

0 comments
Cite this review

Pith. "Pith review of AeroThrow: An Autonomous Aerial Throwing System for Precise Payload Delivery." pith.science (2026). https://pith.science/paper/AGGDOLH4

@misc{pith2026250713903,
  author       = {Pith},
  title        = {Pith review of: AeroThrow: An Autonomous Aerial Throwing System for Precise Payload Delivery},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/AGGDOLH4}},
  note         = {Machine review of arXiv:2507.13903}
}
read the original abstract

Autonomous aerial systems play an increasingly vital role in a wide range of applications, particularly for transport and delivery tasks in complex environments. In airdrop missions, these platforms face the dual challenges of abrupt control mode switching and inherent system delays along with control errors. To address these issues, this paper presents an autonomous airdrop system based on an aerial manipulator (AM). The introduction of additional actuated degrees of freedom enables active compensation for UAV tracking errors. By imposing smooth and continuous constraints on the parabolic landing point, the proposed approach generates aerial throwing trajectories that are less sensitive to the timing of payload release. A hierarchical disturbance compensation strategy is incorporated into the Nonlinear Model Predictive Control (NMPC) framework to mitigate the effects of sudden changes in system parameters, while the predictive capabilities of NMPC are further exploited to improve the precision of aerial throwing. Both simulation and real-world experimental results demonstrate that the proposed system achieves greater agility and precision in airdrop missions.

Figures

Figures reproduced from arXiv: 2507.13903 by the authors.

Figure 1
Figure 1. (a): Agile and precise autonomous aerial throwing. The aerial manipulator carries a payload of unknown mass (200 g in experiments) during flight, [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Coordinate definitions, relevant physical parameters, and notations. [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. Overview of our autonomous aerial throwing system. (a) The Aerial Throwing Planning module generates reference trajectories for both the UAV [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗
Figures from the paper (4 more)
Figure 4
Figure 4. Figure 4: Visualization of the relaxation function. (a): The transition of the [PITH_FULL_IMAGE:figures/full_fig_p005_4.png]
Figure 5
Figure 5. Figure 5: Real-world aerial manipulator platform used for experiment. [PITH_FULL_IMAGE:figures/full_fig_p006_5.png]
Figure 7
Figure 7. Figure 7: NDOB estimation of linear external disturbance forces during the [PITH_FULL_IMAGE:figures/full_fig_p007_7.png]
Figure 8
Figure 8. Figure 8: (a): Landing points from ten throws for each of the nominal release [PITH_FULL_IMAGE:figures/full_fig_p008_8.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

30 extracted references · 28 canonical work pages

  1. [1]

    Drone applications for emergency and urgent care: a systematic review,

    S. Sanz-Martos, M. D. Lopez-Franco, C. ´Alvarez-Garc´ıa, N. Granero- Moya, J. M. Lopez-Hens, S. Camara-Anguita, P. L. Pancorbo-Hidalgo, and I. M. Comino-Sanz, “Drone applications for emergency and urgent care: a systematic review,” Prehospital and disaster medicine , vol. 37, no. 4, pp. 502–508, 2022

    work page 2022

  2. [2]

    Wildfire fighting by unmanned aerial system exploiting its time-varying mass,

    D. A. Saikin, T. Baca, M. Gurtner, and M. Saska, “Wildfire fighting by unmanned aerial system exploiting its time-varying mass,” IEEE Robotics and Automation Letters , vol. 5, no. 2, pp. 2674–2681, 2020

    work page 2020

  3. [3]

    Ring- rotor: A novel retractable ring-shaped quadrotor with aerial grasping and transportation capability,

    Y . Wu, F. Yang, Z. Wang, K. Wang, Y . Cao, C. Xu, and F. Gao, “Ring- rotor: A novel retractable ring-shaped quadrotor with aerial grasping and transportation capability,” IEEE Robotics and Automation Letters , vol. 8, no. 4, pp. 2126–2133, 2023

    work page 2023

  4. [4]

    Differential flatness of quadrotor dynamics subject to rotor drag for accurate tracking of high- speed trajectories,

    M. Faessler, A. Franchi, and D. Scaramuzza, “Differential flatness of quadrotor dynamics subject to rotor drag for accurate tracking of high- speed trajectories,” IEEE Robotics and Automation Letters , vol. 3, no. 2, pp. 620–626, 2017

    work page 2017

  5. [5]

    Dynamic throwing with robotic material handling machines,

    L. Werner, F. Nan, P. Eyschen, F. A. Spinelli, H. Yang, and M. Hutter, “Dynamic throwing with robotic material handling machines,” in 2024 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS). IEEE, 2024, pp. 98–104

    work page 2024

  6. [6]

    Fast trajectory optimization for agile quadrotor maneuvers with a cable- suspended payload,

    P. Foehn, D. Falanga, N. Kuppuswamy, R. Tedrake, and D. Scaramuzza, “Fast trajectory optimization for agile quadrotor maneuvers with a cable- suspended payload,” 2017

    work page 2017

  7. [7]

    Disturbance observer based hovering control of quadrotor tail-sitter vtol uavs using H∞ synthesis,

    X. Lyu, J. Zhou, H. Gu, Z. Li, S. Shen, and F. Zhang, “Disturbance observer based hovering control of quadrotor tail-sitter vtol uavs using H∞ synthesis,” IEEE Robotics and Automation Letters , vol. 3, no. 4, pp. 2910–2917, 2018

    work page 2018

  8. [8]

    Mixed integer quadratic program trajectory generation for a quadrotor with a cable-suspended payload,

    S. Tang and V . Kumar, “Mixed integer quadratic program trajectory generation for a quadrotor with a cable-suspended payload,” in 2015 IEEE international conference on robotics and automation (ICRA) . IEEE, 2015, pp. 2216–2222

    work page 2015

Show all 30 references
  1. [9]

    Differential flatness based path planning with direct collocation on hybrid modes for a quadrotor with a cable-suspended payload,

    J. Zeng, P. Kotaru, M. W. Mueller, and K. Sreenath, “Differential flatness based path planning with direct collocation on hybrid modes for a quadrotor with a cable-suspended payload,” IEEE Robotics and Automation Letters, vol. 5, no. 2, pp. 3074–3081, 2020

    2020

  2. [10]

    Autotrans: A complete planning and control framework for autonomous uav payload transportation,

    H. Li, H. Wang, C. Feng, F. Gao, B. Zhou, and S. Shen, “Autotrans: A complete planning and control framework for autonomous uav payload transportation,” IEEE Robotics and Automation Letters , vol. 8, no. 10, pp. 6859–6866, 2023

    2023

  3. [11]

    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,” in 2024 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) . IEEE, 2024, pp. 10 356–10 361

    2024

  4. [12]

    Ndob-based control of a uav with delta-arm considering manipulator dynamics,

    H. Chen, B. Ye, X. Liang, W. Deng, and X. Lyu, “Ndob-based control of a uav with delta-arm considering manipulator dynamics,” arXiv preprint arXiv:2501.06122, 2025

    2025

  5. [13]

    Whole-body integrated motion planning for aerial manipulators,

    W. Deng, H. Chen, B. Ye, H. Chen, and X. Lyu, “Whole-body integrated motion planning for aerial manipulators,” arXiv preprint arXiv:2501.06493, 2025

    2025

  6. [14]

    Approach methods for autonomous precision aerial drop from a small unmanned aerial vehicle,

    S. H. Mathisen, V . Grindheim, and T. A. Johansen, “Approach methods for autonomous precision aerial drop from a small unmanned aerial vehicle,” IF AC-PapersOnLine, vol. 50, no. 1, pp. 3566–3573, 2017

    2017

  7. [15]

    Autonomous ballistic airdrop of objects from a small fixed- wing unmanned aerial vehicle,

    S. G. Mathisen, F. S. Leira, H. H. Helgesen, K. Gryte, and T. A. Johansen, “Autonomous ballistic airdrop of objects from a small fixed- wing unmanned aerial vehicle,” Autonomous Robots , vol. 44, pp. 859– 875, 2020

    2020

  8. [16]

    Time-optimal trajectory planning with clearly defined initial guess for aerial suspended payload throwing,

    R. Cao, Y . Fang, and X. Liang, “Time-optimal trajectory planning with clearly defined initial guess for aerial suspended payload throwing,” IEEE Transactions on Automation Science and Engineering , 2025

    2025

  9. [17]

    An nmpc framework for tracking and releasing a cable-suspended load to a ground target using a multirotor uav,

    F. Panetsos, G. C. Karras, and K. J. Kyriakopoulos, “An nmpc framework for tracking and releasing a cable-suspended load to a ground target using a multirotor uav,” in 2024 IEEE International Conference on Robotics and Automation (ICRA) . IEEE, 2024, pp. 10 057–10 063

    2024

  10. [18]

    Robust trajectory design for object throwing based on sensitivity for model uncertainties,

    M. Okada, A. Pekarovskiy, and M. Buss, “Robust trajectory design for object throwing based on sensitivity for model uncertainties,” in 2015 IEEE International Conference on Robotics and Automation (ICRA) . IEEE, 2015, pp. 3089–3094

    2015

  11. [19]

    Tossing- bot: Learning to throw arbitrary objects with residual physics,

    A. Zeng, S. Song, J. Lee, A. Rodriguez, and T. Funkhouser, “Tossing- bot: Learning to throw arbitrary objects with residual physics,” IEEE Transactions on Robotics , vol. 36, no. 4, pp. 1307–1319, 2020

    2020

  12. [20]

    State space analysis of timing: exploiting task redundancy to reduce sensitivity to timing,

    R. G. Cohen and D. Sternad, “State space analysis of timing: exploiting task redundancy to reduce sensitivity to timing,” Journal of neurophys- iology, vol. 107, no. 2, pp. 618–627, 2012

    2012

  13. [21]

    Tube acceleration: robust dexterous throwing against release uncertainty,

    Y . Liu and A. Billard, “Tube acceleration: robust dexterous throwing against release uncertainty,” IEEE Transactions on Robotics , 2024

    2024

  14. [22]

    A solution to adaptive mobile manipulator throwing,

    Y . Liu, A. Nayak, and A. Billard, “A solution to adaptive mobile manipulator throwing,” in 2022 Ieee/Rsj International Conference On Intelligent Robots And Systems (Iros) . IEEE, 2022, pp. 1625–1632

    2022

  15. [23]

    Learning accurate whole- body throwing with high-frequency residual policy and pullback tube acceleration,

    Y . Ma, Y . Liu, K. Qu, and M. Hutter, “Learning accurate whole- body throwing with high-frequency residual policy and pullback tube acceleration,” arXiv preprint arXiv:2506.16986 , 2025

    2025 arXiv

  16. [24]

    A comparative study of nonlinear mpc and differential-flatness-based control for quadrotor agile flight,

    S. Sun, A. Romero, P. Foehn, E. Kaufmann, and D. Scaramuzza, “A comparative study of nonlinear mpc and differential-flatness-based control for quadrotor agile flight,” IEEE Transactions on Robotics , vol. 38, no. 6, pp. 3357–3373, 2022

    2022

  17. [25]

    Geometrically constrained tra- jectory optimization for multicopters,

    Z. Wang, X. Zhou, C. Xu, and F. Gao, “Geometrically constrained tra- jectory optimization for multicopters,” IEEE Transactions on Robotics , vol. 38, no. 5, pp. 3259–3278, 2022

    2022

  18. [26]

    Real-time trajectory planning for aerial perching,

    J. Ji, T. Yang, C. Xu, and F. Gao, “Real-time trajectory planning for aerial perching,” in 2022 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) . IEEE, 2022, pp. 10 516–10 522

    2022

  19. [27]

    Towards a modular software package for embedded optimization,

    R. Verschueren, G. Frison, D. Kouzoupis, N. van Duijkeren, A. Zanelli, R. Quirynen, and M. Diehl, “Towards a modular software package for embedded optimization,” IF AC-PapersOnLine, vol. 51, no. 20, pp. 374– 380, 2018

    2018

  20. [28]

    qpoases: A parametric active-set algorithm for quadratic programming,

    H. J. Ferreau, C. Kirches, A. Potschka, H. G. Bock, and M. Diehl, “qpoases: A parametric active-set algorithm for quadratic programming,” Mathematical Programming Computation , vol. 6, pp. 327–363, 2014

    2014

  21. [29]

    Dob-based wind estimation of a uav using its onboard sensor,

    H. Yu, X. Liang, and X. Lyu, “Dob-based wind estimation of a uav using its onboard sensor,” in 2024 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) . IEEE, 2024, pp. 8126–8133

    2024

  22. [30]

    Accurate tracking of aggressive quadrotor tra- jectories using incremental nonlinear dynamic inversion and differential flatness,

    E. Tal and S. Karaman, “Accurate tracking of aggressive quadrotor tra- jectories using incremental nonlinear dynamic inversion and differential flatness,” IEEE Transactions on Control Systems Technology , vol. 29, no. 3, pp. 1203–1218, 2020

    2020

Pith tools

Reviewed August 6, 2026 · model on record in the stance chip above.