arxiv: 2604.05499 · v1 · submitted 2026-04-07 · 💻 cs.RO · cs.SY· eess.SY

Recognition: 2 theorem links

· Lean Theorem

MARS-Dragonfly: Agile and Robust Flight Control of Modular Aerial Robot Systems

Rui Huang , Zhiqian Cai , Siyu Tang , Pengxuan Wei , Lidong Li , Xin Chen , Wenhan Cao , Zhenyu Zhang

show 1 more author

Lin Zhao

Authors on Pith no claims yet

Pith reviewed 2026-05-10 20:05 UTC · model grok-4.3

classification 💻 cs.RO cs.SYeess.SY

keywords modular aerial robotsdrone reconfigurationflight controlpredictive allocationvirtual quadrotoragile maneuveringpolytope constraints

0 comments

The pith

A force-torque-equivalent virtual quadrotor with polytope constraints enables agile control and smooth reconfiguration of modular drone systems.

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

The paper establishes that modular aerial robot systems composed of multiple connected drones can maintain stable, agile flight even while changing shape in the air. Existing approaches rely on simplified models that produce jerky commands and growing errors as more units join, leading to oscillations during docking or separation. The new method replaces those models with a single virtual quadrotor whose feasible forces and torques are bounded by a polytope, then feeds the virtual commands through a two-stage predictive allocator that maps them to balanced, continuous motor signals. If this holds, modular drone teams become practical for missions that require in-flight adaptation without sacrificing precision or speed. Real experiments confirm the result with 40-degree peak pitch angles and an average position error of 0.0896 meters across multiple configurations.

Core claim

The authors present a compact passive docking mechanism and a force-torque-equivalent virtual quadrotor whose polytope-constrained wrench set captures the full coupled dynamics of any connected formation. This abstraction lets any standard quadrotor controller operate unchanged on the whole system. A constrained predictive tracker first computes feasible virtual inputs; a dynamic allocator then distributes them to individual modules under balanced objectives, yielding smooth, trackable motor commands. Simulations and real flights across more than ten configurations demonstrate stable docking, locking, separation, and agile waypoint tracking.

What carries the argument

The force-torque-equivalent and polytope-constraint virtual quadrotor, which encodes the combined feasible wrench set of any MARS formation so that existing single-drone controllers can be applied directly.

If this is right

Stable docking, locking, and separation become possible without attitude error buildup.
Standard quadrotor controllers can be reused across any number of connected modules.
Agility improves through yaw optimization that maximizes available control authority.
Smooth motor commands are generated even as the formation changes shape.

Where Pith is reading between the lines

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

The same virtual-abstraction pattern could be tested on modular ground robots or underwater vehicles where connected units must share control authority.
Real-time reconfiguration might be explored in outdoor wind or with changing payloads to check whether the polytope bounds still prevent saturation.
The allocator's balanced objectives suggest possible extensions that minimize total energy use while preserving tracking accuracy.

Load-bearing premise

The virtual quadrotor and its polytope constraints accurately represent the full coupled dynamics of arbitrary MARS configurations without significant unmodeled effects during docking or separation.

What would settle it

Run a real-world experiment in which a multi-unit MARS docks or separates while tracking a trajectory that requires a 40-degree peak pitch change; if average position error stays near 0.0896 m and oscillations remain absent, the claim holds; otherwise it does not.

Figures

Figures reproduced from arXiv: 2604.05499 by Lidong Li, Lin Zhao, Pengxuan Wei, Rui Huang, Siyu Tang, Wenhan Cao, Xin Chen, Zhenyu Zhang, Zhiqian Cai.

**Figure 1.** Figure 1: Overview of the proposed MARS-Dragonfly framework. (a) Dragonfly-inspired MARS design. (b) A virtual quadrotor abstraction for arbitrary MARS [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗

**Figure 2.** Figure 2: Docking, locking and separation sequence inspired by dragonfly mat [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗

**Figure 3.** Figure 3: Microstructural connections in dragonfly mating that inspire the design of macroscopic mechanical structures. (a) The connection of dragonflies [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗

**Figure 4.** Figure 4: Docking and separation mechanism detail. (a) Exploded view of the mechanism. (b) Front and transparent views of the docking structure. Bottom [PITH_FULL_IMAGE:figures/full_fig_p006_4.png] view at source ↗

**Figure 5.** Figure 5: Magnet arrangement optimization. (a) Original and Halbach array for a row of magnets. (b) The magnets are divided into [PITH_FULL_IMAGE:figures/full_fig_p007_5.png] view at source ↗

**Figure 6.** Figure 6: Coordinate system representation for a MARS with two units. [PITH_FULL_IMAGE:figures/full_fig_p008_6.png] view at source ↗

**Figure 7.** Figure 7: Model abstraction and APC framework for MARS. Localization provides odometry for high-level planning. Model abstraction converts a given MARS [PITH_FULL_IMAGE:figures/full_fig_p011_7.png] view at source ↗

**Figure 9.** Figure 9: Model abstraction for arbitrary-shaped transportation using MARS. [PITH_FULL_IMAGE:figures/full_fig_p012_9.png] view at source ↗

**Figure 10.** Figure 10: Simulation of trajectory tracking and object transport. (a)–(e) Trajectory tracking over ten consecutive laps along a circular path under different [PITH_FULL_IMAGE:figures/full_fig_p013_10.png] view at source ↗

**Figure 11.** Figure 11: Passive docking and lock sequences. The timer format is [PITH_FULL_IMAGE:figures/full_fig_p014_11.png] view at source ↗

**Figure 12.** Figure 12: Manually triggered active separation rotates the male component to repel the connected module. (a) Initial locking state. (b) At 00:00:00, the male [PITH_FULL_IMAGE:figures/full_fig_p014_12.png] view at source ↗

**Figure 13.** Figure 13: Mid-air docking demonstration. The timer format is [PITH_FULL_IMAGE:figures/full_fig_p015_13.png] view at source ↗

**Figure 14.** Figure 14: Mid-air separation demonstration. The timer format is [PITH_FULL_IMAGE:figures/full_fig_p016_14.png] view at source ↗

**Figure 15.** Figure 15: Separation comparison experiments. (a) One side enabled with repulsive force; the other disabled. (b) Initial position at 00:09:13. (c) The mechanism [PITH_FULL_IMAGE:figures/full_fig_p016_15.png] view at source ↗

**Figure 16.** Figure 16: Composite images of trajectory tracking. (a) Trajectory tracking at a velocity of 1 m/s. (b) Circular trajectory tracking with the docking mechanism [PITH_FULL_IMAGE:figures/full_fig_p017_16.png] view at source ↗

**Figure 17.** Figure 17: Agile trajectory tracking. (a) Pitch-angle responses along agile trajectories at [PITH_FULL_IMAGE:figures/full_fig_p017_17.png] view at source ↗

**Figure 19.** Figure 19: Object transportation with a sloshing load. (a) Composite images [PITH_FULL_IMAGE:figures/full_fig_p018_19.png] view at source ↗

**Figure 20.** Figure 20: Agile transport carrying 100 g, 200 g, and 600 g payloads. (a) Pitch angle responses along agile trajectories at [PITH_FULL_IMAGE:figures/full_fig_p019_20.png] view at source ↗

read the original abstract

Modular Aerial Robot Systems (MARS) comprise multiple drone units with reconfigurable connected formations, providing high adaptability to diverse mission scenarios, fault conditions, and payload capacities. However, existing control algorithms for MARS rely on simplified quasi-static models and rule-based allocation, which generate discontinuous and unbounded motor commands. This leads to attitude error accumulation as the number of drone units scales, ultimately causing severe oscillations during docking, separation, and waypoint tracking. To address these limitations, we first design a compact mechanical system that enables passive docking, detection-free passive locking, and magnetic-assisted separation using a single micro servo. Second, we introduce a force-torque-equivalent and polytope-constraint virtual quadrotor that explicitly models feasible wrench sets. Together, these abstractions capture the full MARS dynamics and enable existing quadrotor controllers to be applied across different configurations. We further optimize the yaw angle that maximizes control authority to enhance agility. Third, building on this abstraction, we design a two-stage predictive-allocation pipeline: a constrained predictive tracker computes virtual inputs while respecting force/torque bounds, and a dynamic allocator maps these inputs to individual modules with balanced objectives to produce smooth, trackable motor commands. Simulations across over 10 configurations and real-world experiments demonstrate stable docking, locking, and separation, as well as effective control performance. To our knowledge, this is the first real-world demonstration of MARS achieving agile flight and transport with 40 deg peak pitch while maintaining an average position error of 0.0896 m. The video is available at: https://youtu.be/yqjccrIpz5o

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Desk Editor's Note private letter to a colleague

The paper gives a working virtual-quadrotor-plus-polytope method for controlling reconfigurable drones and backs it with real agile flights, but the transition phases are not checked in detail.

read the letter

The main takeaway is that this group built a passive docking mechanism and then abstracted the whole modular system as a single virtual quadrotor whose wrench limits are captured by a polytope. That lets them run standard quadrotor planners and add a yaw optimizer plus a two-stage allocator that turns virtual commands into smooth motor signals. They show the result in hardware: stable docking, locking, separation, and waypoint tracking across more than ten configurations, with 40-degree peak pitch and 0.09 m average position error. That is the first reported real-world agile MARS flight with those numbers, and it moves past the quasi-static or rule-based methods in earlier work on the same platform family. The experiments and simulations are the strongest part; they actually flew the hardware and measured what matters for agility and accuracy. The mechanical design is compact and uses only one servo, which is a practical plus. The math stays grounded in standard quadrotor dynamics and feasible-set constraints rather than fitted black-box models. The soft spot is exactly where the stress-test note points: during docking and separation the system experiences contact forces, compliance, and magnetic torques that are not folded into the wrench polytope or virtual mapping. The paper gives overall position and attitude numbers but does not break out residuals, oscillation spectra, or prediction errors specifically in those intervals. Without that isolation or an ablation on the polytope bounds, it is hard to know how much the abstraction itself is carrying versus the low-level controllers. Minor gaps like missing error bars and no detailed model-validation plots also make the central claim a bit harder to assess at full strength. This is for aerial-robotics groups that need control that works across changing numbers of modules without rewriting the planner each time. A reader who wants concrete hardware results and a reusable abstraction will get value; someone looking for exhaustive dynamic modeling of every contact phase will want more data. It deserves a serious referee. The real-flight evidence and the clear pipeline are enough to justify review time, even if the transition modeling needs tightening.

Referee Report

2 major / 2 minor

Summary. The manuscript presents MARS-Dragonfly, a modular aerial robot system featuring a compact mechanical design for passive docking, detection-free locking, and magnetic-assisted separation via a single micro servo. It introduces a force-torque-equivalent virtual quadrotor abstraction constrained by wrench polytopes to model feasible sets across configurations, enabling reuse of standard quadrotor controllers. A two-stage predictive allocation pipeline is proposed: a constrained tracker for virtual inputs and a dynamic allocator for smooth motor commands, with yaw optimization to maximize authority. Simulations over 10+ configurations and real-world experiments claim stable agile flight, docking, and transport, with a reported peak pitch of 40 deg and average position error of 0.0896 m, asserted as the first such real-world demonstration.

Significance. If the virtual quadrotor and polytope abstractions hold under dynamic transitions, the work offers a practical unification of MARS control that reduces configuration-specific modeling to standard quadrotor methods with explicit wrench bounds. The real-world experiments across multiple configurations, combined with the mechanically simple docking hardware, provide concrete empirical grounding and could facilitate scalable applications in payload transport and fault-tolerant flight. The predictive allocator's balanced objectives and yaw optimization are constructive extensions of existing allocation techniques.

major comments (2)

[Abstract and real-world experiments section] Abstract and real-world experiments section: The central claim that the force-torque-equivalent virtual quadrotor and polytope constraints 'capture the full MARS dynamics' is load-bearing for applying standard controllers, yet experiments report only aggregate metrics (0.0896 m average position error, 40 deg peak pitch) without isolating model-prediction residuals, wrench-set mismatch, or oscillation spectra specifically during passive docking, locking, and magnetic separation intervals. These phases involve unmodeled contact compliance and transient torques not folded into the polytope, so the two-stage allocator operates on an incomplete feasible set precisely when discontinuities are most likely.
[Methods description of the virtual quadrotor and polytope] Methods description of the virtual quadrotor and polytope: The wrench polytope is constructed from individual module force/torque sets, but no quantitative bound or sensitivity analysis is provided on how docking-induced mechanical coupling alters the effective feasible set relative to the virtual model. This directly affects the constrained predictive tracker's guarantees and the claim of configuration-independent stability.

minor comments (2)

[Abstract] Abstract: The reported position error of 0.0896 m lacks accompanying standard deviation, error bars, or per-phase breakdown, making it difficult to assess consistency across the claimed agile maneuvers.
[Simulation results] Simulation results: While over 10 configurations are mentioned, the manuscript would benefit from an explicit table or figure summarizing per-configuration peak errors and constraint violation rates to strengthen the cross-configuration claim.

Simulated Author's Rebuttal

2 responses · 0 unresolved

Thank you for the opportunity to respond to the referee's constructive report. We address each major comment point by point below, providing clarifications on the scope of our abstractions and indicating revisions to strengthen the validation of transient phases and polytope assumptions.

read point-by-point responses

Referee: [Abstract and real-world experiments section] Abstract and real-world experiments section: The central claim that the force-torque-equivalent virtual quadrotor and polytope constraints 'capture the full MARS dynamics' is load-bearing for applying standard controllers, yet experiments report only aggregate metrics (0.0896 m average position error, 40 deg peak pitch) without isolating model-prediction residuals, wrench-set mismatch, or oscillation spectra specifically during passive docking, locking, and magnetic separation intervals. These phases involve unmodeled contact compliance and transient torques not folded into the polytope, so the two-stage allocator operates on an incomplete feasible set precisely when discontinuities are most likely.

Authors: We thank the referee for this observation. The force-torque virtual quadrotor and polytope are formulated for the rigidly connected configuration that constitutes the primary mode of agile flight and transport; the mechanical design (passive docking with detection-free locking) is intended to keep contact transients brief and mechanically constrained. Aggregate metrics and the video evidence show that the two-stage allocator maintains stability without large oscillations through these intervals. We acknowledge that isolating residuals and spectra specifically during docking/separation would provide stronger evidence for the claim of capturing full dynamics. In the revision we will add time-series error plots focused on those phases, a brief discussion of unmodeled compliance, and clarification that the polytope applies to the locked state while the allocator handles short transitions. revision: yes
Referee: [Methods description of the virtual quadrotor and polytope] Methods description of the virtual quadrotor and polytope: The wrench polytope is constructed from individual module force/torque sets, but no quantitative bound or sensitivity analysis is provided on how docking-induced mechanical coupling alters the effective feasible set relative to the virtual model. This directly affects the constrained predictive tracker's guarantees and the claim of configuration-independent stability.

Authors: The polytope is derived from per-module wrench sets under the assumption of rigid post-docking connection, consistent with the passive-locking hardware that minimizes compliance. We did not include a dedicated sensitivity study because the primary contribution centers on control performance across configurations. We agree that quantitative bounds on coupling effects would better support the configuration-independent guarantees. In the revised manuscript we will add a short analysis (simulation-based variation of coupling stiffness) quantifying the deviation of the effective feasible set from the nominal polytope and its influence on the predictive tracker. revision: yes

Circularity Check

0 steps flagged

No significant circularity in derivation chain

full rationale

The paper's central modeling step defines a force-torque-equivalent virtual quadrotor and wrench polytope directly from physical actuator limits and configuration geometry using standard rigid-body wrench summation; this equivalence is constructed from first-principles dynamics rather than from the target control performance or any fitted data. The subsequent two-stage allocator and yaw optimization are then derived as standard constrained optimization over that explicitly computed feasible set, without renaming known results or invoking self-cited uniqueness theorems as load-bearing premises. Experiments serve only as external validation and do not retroactively define the model or its predictions. No step reduces by construction to its own inputs.

Axiom & Free-Parameter Ledger

2 free parameters · 2 axioms · 0 invented entities

The approach relies on standard rigid-body dynamics and optimization assumptions rather than new fitted constants or invented physical entities. The virtual quadrotor is an abstraction, not a new postulated object.

free parameters (2)

yaw optimization objective weights
Chosen to maximize control authority; specific values not stated in abstract.
predictive horizon and constraint bounds
Tuned for the two-stage pipeline; not enumerated.

axioms (2)

domain assumption Quadrotor wrench sets can be represented as polytopes that remain valid under module reconfiguration.
Invoked to justify the virtual quadrotor abstraction.
standard math Standard quadrotor dynamics apply to the aggregated MARS system.
Background assumption for applying existing controllers.

pith-pipeline@v0.9.0 · 5626 in / 1385 out tokens · 63241 ms · 2026-05-10T20:05:27.895903+00:00 · methodology

discussion (0)

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

IndisputableMonolith/Cost/FunctionalEquation.lean washburn_uniqueness_aczel unclear

?

unclear
Relation between the paper passage and the cited Recognition theorem.

force–torque–equivalent and polytope-constraint virtual quadrotor that explicitly models feasible wrench sets... two-stage predictive–allocation pipeline
IndisputableMonolith/Foundation/BranchSelection.lean branch_selection unclear

?

unclear
Relation between the paper passage and the cited Recognition theorem.

optimize the yaw angle that maximizes control authority

What do these tags mean?

matches: The paper's claim is directly supported by a theorem in the formal canon.
supports: The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends: The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses: The paper appears to rely on the theorem as machinery.
contradicts: The paper's claim conflicts with a theorem or certificate in the canon.
unclear: Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

Reference graph

Works this paper leans on

45 extracted references · 45 canonical work pages

[1]

What de- termines mate choices? heterospecific mating in sympetrum dragonflies,

V . Kornov´a, E. B ´ılkov´a, P. Pyszko, A. Doln `y, and S. O ˇzana, “What de- termines mate choices? heterospecific mating in sympetrum dragonflies,” Freshwater Biology, vol. 69, no. 4, pp. 527–537, 2024

work page 2024
[2]

Egg-laying traits reflect shifts in dragonfly assemblages in response to different amount of tropical forest cover,

M. E. Rodrigues, F. D. O. Roque, R. Guillermo-Ferreira, V . S. Saito, and M. J. Samways, “Egg-laying traits reflect shifts in dragonfly assemblages in response to different amount of tropical forest cover,” Insect Conservation and Diversity, vol. 12, no. 3, pp. 231–240, 2019

work page 2019
[3]

The dodecacopter: A versatile multirotor system of dodecahedron-shaped modules,

K. Garanger, T. Khamvilai, J. Epps, and E. Feron, “The dodecacopter: A versatile multirotor system of dodecahedron-shaped modules,”IEEE Transactions on Robotics, 2025

work page 2025
[4]

The distributed flight array: Design, implementation, and analysis of a modular vertical take-off and landing vehicle,

R. Oung and R. D’Andrea, “The distributed flight array: Design, implementation, and analysis of a modular vertical take-off and landing vehicle,”The International Journal of Robotics Research, vol. 33, no. 3, pp. 375–400, 2014

work page 2014
[6]

Self- reconfiguration in response to faults in modular aerial systems,

N. Gandhi, D. Saldana, V . Kumar, and L. T. X. Phan, “Self- reconfiguration in response to faults in modular aerial systems,”IEEE Robotics and Automation Letters, vol. 5, no. 2, pp. 2522–2529, 2020

work page 2020
[7]

A multi-vtol modular aspect ratio reconfigurable aerial robot,

S. J. Carlson, P. Arora, and C. Papachristos, “A multi-vtol modular aspect ratio reconfigurable aerial robot,” in2022 International conference on robotics and automation (ICRA). IEEE, 2022, pp. 8–15

work page 2022
[8]

Fault-tolerant control of an overactuated uav platform built on quadcopters and passive hinges,

Y . Su, P. Yu, M. J. Gerber, L. Ruan, and T.-C. Tsao, “Fault-tolerant control of an overactuated uav platform built on quadcopters and passive hinges,”IEEE/ASME Transactions on Mechatronics, vol. 29, no. 1, pp. 602–613, 2023

work page 2023
[9]

Adaptive configuration control of combined uavs based on leader-wingman mode,

R. Huang, H. Shenget al., “Adaptive configuration control of combined uavs based on leader-wingman mode,”Chinese Journal of Aeronautics, vol. 37, no. 12, pp. 416–433, 2024

work page 2024
[10]

Modular multirotors: From quadrotors to fully-actuated aerial vehicles,

J. Xu, D. S. D’antonio, and D. Salda ˜na, “Modular multirotors: From quadrotors to fully-actuated aerial vehicles,”IEEE Transactions on Automation Science and Engineering, 2025

work page 2025
[11]

Versatile tasks on integrated aerial platforms using only onboard sensors: Control, estimation, and validation,

K. Wang, G. Lai, Y . Yu, J. Du, J. Sun, B. Xu, A. Franchi, and F. Sun, “Versatile tasks on integrated aerial platforms using only onboard sensors: Control, estimation, and validation,”IEEE Transactions on Robotics, 2025

work page 2025
[12]

Modquad: The flying modular structure that self-assembles in midair,

D. Saldana, B. Gabrich, G. Li, M. Yim, and V . Kumar, “Modquad: The flying modular structure that self-assembles in midair,” in2018 IEEE International Conference on Robotics and Automation (ICRA). IEEE, 2018, pp. 691–698

work page 2018
[13]

Modquad-vi: A vision-based self-assembling modular quadrotor,

G. Li, B. Gabrich, D. Saldana, J. Das, V . Kumar, and M. Yim, “Modquad-vi: A vision-based self-assembling modular quadrotor,” in 2019 International Conference on Robotics and Automation (ICRA). IEEE, 2019, pp. 346–352

work page 2019
[14]

Design and control of aerial modules for inflight self-disassembly,

D. Saldana, P. M. Gupta, and V . Kumar, “Design and control of aerial modules for inflight self-disassembly,”IEEE Robotics and Automation Letters, vol. 4, no. 4, pp. 3410–3417, 2019

work page 2019
[15]

A decentralized algorithm for assembling structures with modular robots,

D. Saldana, B. Gabrich, M. Whitzer, A. Prorok, M. F. Campos, M. Yim, and V . Kumar, “A decentralized algorithm for assembling structures with modular robots,” in2017 IEEE/RSJ international conference on intelligent robots and systems (IROS). IEEE, 2017, pp. 2736–2743

work page 2017
[16]

Vision-based self-assembly for modular multirotor structures,

Y . Litman, N. Gandhi, L. T. X. Phan, and D. Salda ˜na, “Vision-based self-assembly for modular multirotor structures,”IEEE Robotics and Automation Letters, vol. 6, no. 2, pp. 2202–2208, 2021

work page 2021
[17]

Robust self-reconfiguration for fault-tolerant control of modular aerial robot systems,

R. Huang, S. Tang, Z. Cai, and L. Zhao, “Robust self-reconfiguration for fault-tolerant control of modular aerial robot systems,” in2025 IEEE International Conference on Robotics and Automation (ICRA), 2025, pp. 12 614–12 620

work page 2025
[18]

Transformars: Fault-tolerant self-reconfiguration for arbitrarily shaped modular aerial robot systems,

R. Huang, Z. Gao, S. Tang, J. Zhang, L. He, Z. Zhang, and L. Zhao, “Transformars: Fault-tolerant self-reconfiguration for arbitrarily shaped modular aerial robot systems,”arXiv preprint arXiv:2509.14025, 2025

work page arXiv 2025
[19]

Finding structure configurations for flying modular robots,

B. Gabrich, D. Salda ˜na, and M. Yim, “Finding structure configurations for flying modular robots,” in2021 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS). IEEE, 2021, pp. 6970–6976

work page 2021
[20]

Flight structure optimization of modular reconfigurable uavs,

Y . Su, Z. Jiao, Z. Zhang, J. Zhang, H. Li, M. Wang, and H. Liu, “Flight structure optimization of modular reconfigurable uavs,” in2024 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS). IEEE, 2024, pp. 4556–4562

work page 2024
[21]

A modular aerial system based on homogeneous quadrotors with fault-tolerant control,

M. Li, K. Cui, and H. Koeppl, “A modular aerial system based on homogeneous quadrotors with fault-tolerant control,” in2024 IEEE International Conference on Robotics and Automation (ICRA). IEEE, 2024, pp. 8408–8414

work page 2024
[22]

Mars-ftcp: Robust fault-tolerant control and agile trajectory planning for modular aerial robot systems,

R. Huang, Z. Zhang, S. Tang, Z. Cai, and L. Zhao, “Mars-ftcp: Robust fault-tolerant control and agile trajectory planning for modular aerial robot systems,” in2025 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS). IEEE, 2025, pp. 12 086–12 092

work page 2025
[23]

Versatile articulated aerial robot dragon: Aerial manipulation and grasping by vectorable thrust control,

M. Zhao, K. Okada, and M. Inaba, “Versatile articulated aerial robot dragon: Aerial manipulation and grasping by vectorable thrust control,” The International Journal of Robotics Research, vol. 42, no. 4-5, pp. 214–248, 2023

work page 2023
[24]

Hierarchical trajectory planning of floating-base multi-link robot for maneuvering in confined environments,

Y . Chen, J. Li, H. Liu, Z. Luo, K. Kaneko, and M. Zhao, “Hierarchical trajectory planning of floating-base multi-link robot for maneuvering in confined environments,”IEEE Transactions on Automation Science and Engineering, 2026

work page 2026
[25]

Design, control, and motion strategy of trady: Tilted-rotor-equipped aerial robot with autonomous in-flight assembly and disassembly ability,

J. Sugihara, T. Nishio, K. Nagato, M. Nakao, and M. Zhao, “Design, control, and motion strategy of trady: Tilted-rotor-equipped aerial robot with autonomous in-flight assembly and disassembly ability,”Advanced Intelligent Systems, vol. 5, no. 10, p. 2300191, 2023

work page 2023
[26]

Beatle—self- reconfigurable aerial robot: Design, control and experimental validation,

J. Sugihara, M. Zhao, T. Nishio, K. Okada, and M. Inaba, “Beatle—self- reconfigurable aerial robot: Design, control and experimental validation,” IEEE/ASME Transactions on Mechatronics, 2024

work page 2024
[27]

Modquad-dof: A novel yaw actuation for modular quadrotors,

B. Gabrich, G. Li, and M. Yim, “Modquad-dof: A novel yaw actuation for modular quadrotors,” in2020 IEEE international conference on robotics and automation (ICRA). IEEE, 2020, pp. 8267–8273

work page 2020
[28]

A flying gripper based on cuboid modular robots,

B. Gabrich, D. Saldana, V . Kumar, and M. Yim, “A flying gripper based on cuboid modular robots,” in2018 IEEE International Conference on Robotics and Automation (ICRA). IEEE, 2018, pp. 7024–7030

work page 2018
[29]

Design and control of rapid in-air reconfiguration for modular quadrotors with full controllable degrees of freedom,

J. Zhang, F. Li, X. Lu, C. Zhang, Y . Xin, R. Zhao, and S. Lyu, “Design and control of rapid in-air reconfiguration for modular quadrotors with full controllable degrees of freedom,”IEEE Robotics and Automation Letters, 2024

work page 2024
[30]

Functional morphology of the male caudal appendages of the damselfly ischnura elegans (zy- goptera: Coenagrionidae),

J. Willkommen, J. Michels, and S. N. Gorb, “Functional morphology of the male caudal appendages of the damselfly ischnura elegans (zy- goptera: Coenagrionidae),”Arthropod structure & development, vol. 44, no. 4, pp. 289–300, 2015

work page 2015
[31]

The discrete dipole approximation: an overview and recent developments,

M. A. Yurkin and A. G. Hoekstra, “The discrete dipole approximation: an overview and recent developments,”Journal of Quantitative Spec- troscopy and Radiative Transfer, vol. 106, no. 1-3, pp. 558–589, 2007

work page 2007
[32]

Casadi: A software framework for nonlinear optimization and optimal control,

J. A. E. Andersson, J. Gillis, G. Horn, J. B. Rawlings, and M. Diehl, “Casadi: A software framework for nonlinear optimization and optimal control,”Mathematical Programming Computation, vol. 11, no. 1, pp. 1–36, 2019

work page 2019
[33]

Decentralized aerial manipulation of a cable-suspended load using multi-agent reinforcement learning,

J. Zeng, A. M. Gimenez, E. Vinitsky, J. Alonso-Mora, and S. Sun, “Decentralized aerial manipulation of a cable-suspended load using multi-agent reinforcement learning,”arXiv preprint arXiv:2508.01522, 2025

work page arXiv 2025
[34]

Agile and cooperative aerial manipulation of a cable-suspended load,

S. Sun, X. Wang, D. Sanalitro, A. Franchi, M. Tognon, and J. Alonso- Mora, “Agile and cooperative aerial manipulation of a cable-suspended load,”Science Robotics, vol. 10, no. 107, p. eadu8015, 2025

work page 2025
[35]

Decentralized real-time planning for multi-uav cooperative manipulation via imitation learning,

S. Agarwal, J. Alonso-Mora, and S. Sun, “Decentralized real-time planning for multi-uav cooperative manipulation via imitation learning,” arXiv preprint arXiv:2510.17143, 2025

work page arXiv 2025
[36]

Safe and agile transportation of cable-suspended payload via multiple aerial robots,

Y . Wang, J. Wang, X. Zhou, T. Yang, C. Xu, and F. Gao, “Safe and agile transportation of cable-suspended payload via multiple aerial robots,” arXiv preprint arXiv:2501.15272, 2025

work page arXiv 2025
[37]

Pampc: Perception- aware model predictive control for quadrotors,

D. Falanga, P. Foehn, P. Lu, and D. Scaramuzza, “Pampc: Perception- aware model predictive control for quadrotors,” in2018 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS). IEEE, 2018, pp. 1–8

work page 2018
[38]

Nonlinear mpc for quadrotors in close-proximity flight with neural network downwash prediction,

J. Li, L. Han, H. Yu, Y . Lin, Q. Li, and Z. Ren, “Nonlinear mpc for quadrotors in close-proximity flight with neural network downwash prediction,” in2023 62nd IEEE Conference on Decision and Control (CDC). IEEE, 2023, pp. 2122–2128

work page 2023
[39]

Demonstrating agile flight from pixels without state estimation,

I. Geles, L. Bauersfeld, A. Romero, J. Xing, and D. Scaramuzza, “Demonstrating agile flight from pixels without state estimation,” Robotics Science and Systems online Proceedings, no. 20, p. online, 2024

work page 2024
[40]

Actor–critic model predictive control: Differentiable optimization meets reinforce- ment learning for agile flight,

A. Romero, E. Aljalbout, Y . Song, and D. Scaramuzza, “Actor–critic model predictive control: Differentiable optimization meets reinforce- ment learning for agile flight,”IEEE Transactions on Robotics, vol. 42, pp. 673–692, 2025

work page 2025
[41]

Autonomous drone racing: A survey,

D. Hanover, A. Loquercio, L. Bauersfeld, A. Romero, R. Penicka, Y . Song, G. Cioffi, E. Kaufmann, and D. Scaramuzza, “Autonomous drone racing: A survey,”IEEE Transactions on Robotics, vol. 40, pp. 3044–3067, 2024

work page 2024
[42]

Polynomial trajectory planning for aggressive quadrotor flight in dense indoor environments,

C. Richter, A. Bry, and N. Roy, “Polynomial trajectory planning for aggressive quadrotor flight in dense indoor environments,” inRobotics Research. Springer, 2016, pp. 649–666. IEEE JOURNAL TEMPLATE 21

work page 2016
[43]

Design and implementation of epm based modular micro-uavs for autonomous midair docking,

X. Zhang, H. Lu, Z. Feng, and X. Xiao, “Design and implementation of epm based modular micro-uavs for autonomous midair docking,”IEEE Robotics and Automation Letters, vol. 11, no. 1, pp. 514–521, 2025

work page 2025
[44]

H-modquad: Modular multi- rotors with 4, 5, and 6 controllable dof,

J. Xu, D. S. D’antonio, and D. Salda ˜na, “H-modquad: Modular multi- rotors with 4, 5, and 6 controllable dof,” in2021 IEEE International Conference on Robotics and Automation (ICRA). IEEE, 2021, pp. 190–196

work page 2021
[45]

Champion-level drone racing using deep reinforcement learning,

E. Kaufmann, L. Bauersfeld, A. Loquercio, M. M ¨uller, V . Koltun, and D. Scaramuzza, “Champion-level drone racing using deep reinforcement learning,”Nature, vol. 620, no. 7976, pp. 982–987, 2023

work page 2023
[46]

Unlocking aerobatic potential of quadcopters: Autonomous freestyle flight generation and execution,

M. Wang, Q. Wang, Z. Wang, Y . Gao, J. Wang, C. Cui, Y . Li, Z. Ding, K. Wang, C. Xuet al., “Unlocking aerobatic potential of quadcopters: Autonomous freestyle flight generation and execution,” Science Robotics, vol. 10, no. 101, p. eadp9905, 2025

work page 2025