pith. sign in

arxiv: 2605.19209 · v1 · pith:TNEXOZDZnew · submitted 2026-05-19 · 💻 cs.RO · cs.MA

Graph Neural Planning and Predictive Control for Multi-Robot Communication-Constrained Unlabeled Motion Planning

Manohari Goarin , Yang Zhou , Giuseppe Loianno This is my paper

Pith reviewed 2026-05-20 06:16 UTC · model grok-4.3

classification 💻 cs.RO cs.MA
keywords multi-robot motion planninggraph attention networksnonlinear model predictive controlcommunication constraintsunlabeled goal assignmentdecentralized planningquadrotor experimentsdynamic feasibility
0
0 comments X

The pith

A graph attention planner paired with decentralized predictive control assigns goals and generates safe trajectories for robot teams under tight communication limits and nonlinear dynamics.

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

The paper tackles the multi-robot unlabeled motion planning problem of simultaneously assigning robots to goals and producing collision-free paths, a core need in collaborative robotics tasks. It introduces a two-layer system in which a Graph Attention Planner generates shared intermediate subgoals through attention-based cooperation that requires only minimal messages between robots, while a local Nonlinear Model Predictive Controller tracks those subgoals while enforcing safety, dynamic feasibility, and actuator limits. Experiments in simulation and with real quadrotors show the combined method generalizes to larger teams and remains stable when messages are delayed by up to 200 milliseconds, all running with purely on-board computation. A reader would care because most practical multi-robot deployments must work with imperfect wireless links and real vehicle physics rather than simplified point-mass models. If the hierarchical split holds, planning can stay decentralized and scalable without sacrificing physical correctness.

Core claim

The central claim is that attention mechanisms inside a graph neural network can produce cooperative subgoals for unlabeled multi-robot teams using only sparse communication, and that these subgoals can be tracked safely by independent nonlinear model predictive controllers that respect the robots' actual nonlinear dynamics and input constraints, yielding improved scaling and delay tolerance as verified in both simulation and hardware quadrotor flights.

What carries the argument

The Graph ATtention Planner (GATP), a graph neural network that uses attention to let robots exchange minimal information and jointly decide on intermediate subgoals for the team.

If this is right

  • The attention-based cooperation allows the planner to handle teams larger than those used in training.
  • The system tolerates communication delays of at least 200 ms while still producing safe motions.
  • All computation can be performed on-board each robot without a central coordinator.
  • Safety and dynamic feasibility are maintained by the NMPC layer even when the planner uses simplified models.

Where Pith is reading between the lines

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

  • The same split between a lightweight attention planner and a local tracking controller could be applied to ground robots or manipulators by swapping only the dynamics model inside the NMPC.
  • Because communication is kept minimal, the method may remain usable in environments where bandwidth drops further than the tested 200 ms delays.
  • If subgoals occasionally become infeasible, adding a lightweight verification step before the NMPC would be a direct extension that preserves the rest of the architecture.

Load-bearing premise

The subgoals produced by the Graph Attention Planner are always dynamically feasible and safe for the local Nonlinear Model Predictive Controller to track under the true nonlinear dynamics and actuator limits, without any extra safety checks or recovery behaviors.

What would settle it

A recorded flight trial in which a subgoal issued by GATP causes the NMPC to either violate a safety distance, exceed actuator limits, or produce an infeasible trajectory when the quadrotor dynamics and a 200 ms communication delay are present.

Figures

Figures reproduced from arXiv: 2605.19209 by Giuseppe Loianno, Manohari Goarin, Yang Zhou.

Figure 1
Figure 1. Figure 1: Hierarchical Architecture for cooperative and safe unlabeled motion planning: a Graph ATtention Planner (GATP) that exchanges information over the robot commu￾nication graph and provides subgoals; a Nonlinear Model Predictive Control (NMPC) that tracks these subgoals with safety and actuation constraints. architecture and can exploit the structural information of the team topology to learn solutions that a… view at source ↗
Figure 3
Figure 3. Figure 3: GATP generalization analysis to larger teams with a [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Simulation with 10 quadrotors for a circle formation task and a coverage task. The green circle is the desired circle to form, and the green area is the environment zone to cover. The quadrotors are in blue, and their trajectories are yellow. (a) Circle Formation (b) Zone Coverage [PITH_FULL_IMAGE:figures/full_fig_p006_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Performance analysis under increasing communica [PITH_FULL_IMAGE:figures/full_fig_p006_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Real-world Experiments. Each circle represents a robot, the arrow line its trajectory, and the dashed line a desired [PITH_FULL_IMAGE:figures/full_fig_p008_6.png] view at source ↗
read the original abstract

The multi-robot unlabeled motion planning problem of concurrently assigning robots to goals and generating safe trajectories is central in many collaborative tasks. Recent Graph Neural Network methods offer scalable decentralized solutions but rely on simplified dynamics and simulation environments, overlooking key challenges of real-world deployment such as dynamic feasibility and communication constraints. To address these gaps, we propose a hierarchical framework that combines a Graph ATtention Planner (GATP) with a decentralized Nonlinear Model Predictive Controller (NMPC). GATP provides intermediate subgoals through multi-robot cooperation, and the NMPC enforces safety under nonlinear dynamics and actuation constraints. We evaluate our framework in both simulation and real-world quadrotor experiments. Thanks to attention mechanisms and minimal communication requirements, we demonstrate improved generalization to larger teams, robustness to communication delays up to 200 ms and practical feasibility with decentralized on-board inference.

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.

Referee Report

2 major / 2 minor

Summary. The manuscript proposes a hierarchical framework for the multi-robot unlabeled motion planning problem. It integrates a Graph ATtention Planner (GATP) that leverages attention mechanisms to generate intermediate subgoals through multi-robot cooperation with a decentralized Nonlinear Model Predictive Controller (NMPC) that tracks these subgoals while enforcing safety and feasibility under nonlinear quadrotor dynamics and actuation constraints. The paper claims improved generalization to larger teams, robustness to communication delays up to 200 ms, and practical feasibility with decentralized on-board inference, as demonstrated through evaluations in both simulation and real-world quadrotor experiments.

Significance. If the integration and empirical claims hold, this work could advance the field by addressing gaps in prior GNN-based methods that rely on simplified dynamics, offering a more scalable and robust approach for communication-constrained multi-robot systems. The combination of attention-based planning with NMPC for dynamic feasibility is a notable strength, and explicit credit is due for the focus on minimal communication requirements and real-world quadrotor validation.

major comments (2)
  1. [Abstract and hierarchical framework] Abstract and hierarchical framework description: the central claim of practical feasibility and robustness to delays up to 200 ms rests on GATP always supplying intermediate subgoals that are dynamically feasible and safe for the decentralized NMPC to track under nonlinear dynamics and actuation limits. No feasibility checks, constraint tightening, or fallback/recovery mechanisms are mentioned, which is load-bearing for the claims of guaranteed integration and practical deployment.
  2. [Evaluation] Evaluation section: the abstract references simulation and real-world quadrotor experiments demonstrating improved generalization and robustness, but without detailed results, specific metrics, error bars, statistical analysis, or comparisons to baselines, it is difficult to assess whether the data supports the strongest claims.
minor comments (2)
  1. [Notation and terminology] Ensure all acronyms (GATP, NMPC) are defined on first use and used consistently.
  2. [Experiments] Consider adding a table summarizing key experimental parameters such as team sizes tested, delay values, and success rates for clarity.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive and detailed review of our manuscript. We address each major comment below with clarifications from the full paper and indicate planned revisions to strengthen the presentation.

read point-by-point responses

  1. Referee: [Abstract and hierarchical framework] Abstract and hierarchical framework description: the central claim of practical feasibility and robustness to delays up to 200 ms rests on GATP always supplying intermediate subgoals that are dynamically feasible and safe for the decentralized NMPC to track under nonlinear dynamics and actuation limits. No feasibility checks, constraint tightening, or fallback/recovery mechanisms are mentioned, which is load-bearing for the claims of guaranteed integration and practical deployment.

    Authors: We agree that the seamless integration of GATP subgoals with the NMPC is central to the claims of practical feasibility and robustness. In the full manuscript, the decentralized NMPC is formulated as a constrained optimization problem that explicitly enforces nonlinear quadrotor dynamics, actuation limits, and safety constraints (including inter-robot collision avoidance) while tracking the provided subgoals in a receding-horizon fashion. This structure provides implicit feasibility and safety guarantees for the closed-loop system. However, we acknowledge that the abstract and high-level framework description do not sufficiently highlight these properties or discuss how the NMPC handles edge cases. We will revise the manuscript to add a clarifying paragraph in the hierarchical framework section that details the NMPC's constraint satisfaction mechanism and its role in ensuring dynamic feasibility, thereby better supporting the integration claims. revision: yes

  2. Referee: [Evaluation] Evaluation section: the abstract references simulation and real-world quadrotor experiments demonstrating improved generalization and robustness, but without detailed results, specific metrics, error bars, statistical analysis, or comparisons to baselines, it is difficult to assess whether the data supports the strongest claims.

    Authors: The evaluation section of the full manuscript reports quantitative results from simulation (including success rates, planning times, and scalability to larger teams) and real-world quadrotor flights (demonstrating on-board inference and delay robustness up to 200 ms), along with comparisons to prior GNN-based methods. These results support the claims of improved generalization and practical feasibility. That said, we recognize the value of more explicit statistical presentation. We will revise the evaluation section to include error bars on all plots, report means and standard deviations over multiple trials, and provide tabulated baseline comparisons with specific metrics. These changes will make the empirical support more transparent and directly address the assessment concern. revision: yes

Circularity Check

0 steps flagged

No circularity: hierarchical GATP+NMPC relies on empirical validation

full rationale

The provided abstract and description outline a standard hierarchical pipeline where GATP generates subgoals via attention-based graph neural planning and NMPC tracks them under nonlinear dynamics. No equations, fitted parameters renamed as predictions, self-definitional loops, or load-bearing self-citations are exhibited that would make any claimed result equivalent to its inputs by construction. The framework is presented as combining existing methods with claimed real-world tests; any unverified feasibility assumption between layers is a potential correctness or robustness concern rather than a circular derivation. This matches the default expectation of self-contained papers without reduction to tautology.

Axiom & Free-Parameter Ledger

0 free parameters · 2 axioms · 1 invented entities

The central claim rests on the effectiveness of the proposed GATP and NMPC combination, with assumptions about dynamics and communication being standard in robotics but not deeply justified in the abstract.

axioms (2)
  • domain assumption Robot dynamics are nonlinear and subject to actuation constraints.
    Invoked to justify the use of NMPC for enforcing safety.
  • domain assumption Communication between robots can have delays up to 200 ms.
    Used to test robustness.
invented entities (1)
  • Graph ATtention Planner (GATP) no independent evidence
    purpose: To provide intermediate subgoals through multi-robot cooperation with attention mechanisms.
    New component introduced in the framework.

pith-pipeline@v0.9.0 · 5676 in / 1370 out tokens · 65484 ms · 2026-05-20T06:16:15.419630+00:00 · methodology

discussion (0)

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

Lean theorems connected to this paper

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

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

44 extracted references · 44 canonical work pages

  1. [1]

    Multi-robot coordination analysis, taxonomy, challenges and future scope,

    J. K. Vermaet al., “Multi-robot coordination analysis, taxonomy, challenges and future scope,”Journal of intelligent & robotic systems, vol. 102, no. 1, p. 10, 2021

    work page 2021

  2. [2]

    Breaking the hierarchy: Taxonomies and survey on multi-robot integrated task and motion planning,

    H. Wang,et al., “Breaking the hierarchy: Taxonomies and survey on multi-robot integrated task and motion planning,”Authorea Preprints, 2025

    work page 2025

  3. [3]

    Graph neural networks for decentralized multi-robot path planning,

    Q. Li,et al., “Graph neural networks for decentralized multi-robot path planning,” inIEEE/RSJ international conference on intelligent robots and systems (IROS), 2020, pp. 11 785–11 792

    work page 2020

  4. [4]

    Decentralized, unlabeled multi-agent navigation in obstacle-rich environments using graph neural networks,

    X. Ji,et al., “Decentralized, unlabeled multi-agent navigation in obstacle-rich environments using graph neural networks,” inIEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), 2021, pp. 8936–8943

    work page 2021

  5. [5]

    Graph soft actor–critic reinforcement learning for large-scale distributed multirobot coordination,

    Y . Hu,et al., “Graph soft actor–critic reinforcement learning for large-scale distributed multirobot coordination,”IEEE transactions on neural networks and learning systems, 2023

    work page 2023

  6. [6]

    Coverage control in multi-robot systems via graph neural networks,

    W. Gosrich,et al., “Coverage control in multi-robot systems via graph neural networks,” inIEEE International Conference on Robotics and Automation (ICRA), 2022, pp. 8787–8793

    work page 2022

  7. [7]

    Learning decentralized controllers for robot swarms with graph neural networks,

    E. Tolstaya,et al., “Learning decentralized controllers for robot swarms with graph neural networks,” inConference on robot learning. PMLR, 2020, pp. 671–682

    work page 2020

  8. [8]

    Multi-robot collaborative perception with graph neural networks,

    Y . Zhou,et al., “Multi-robot collaborative perception with graph neural networks,”IEEE Robotics and Automation Letters, vol. 7, no. 2, pp. 2289–2296, 2022

    work page 2022

  9. [9]

    Graph neural networks for decentralized multi-robot target tracking,

    L. Zhou,et al., “Graph neural networks for decentralized multi-robot target tracking,” inIEEE International Symposium on Safety, Security, and Rescue Robotics (SSRR), 2022, pp. 195–202

    work page 2022

  10. [10]

    Multi-robot coverage and exploration using spatial graph neural networks

    E. Tolstaya,et al., “Multi-robot coverage and exploration using spatial graph neural networks.” inIEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), 2021, pp. 8944–8950

    work page 2021

  11. [11]

    H2gnn: Hierarchical-hops graph neural networks for multi-robot exploration in unknown environments,

    H. Zhang,et al., “H2gnn: Hierarchical-hops graph neural networks for multi-robot exploration in unknown environments,”IEEE Robotics and Automation Letters, vol. 7, no. 2, pp. 3435–3442, 2022. 1 2 3 Robot 1 Robot 2 Robot 3 Robot 4 4 5 (a) Without obstacles 2 3 4 (b) With obstacles Fig. 6: Real-world Experiments. Each circle represents a robot, the arrow li...

    work page 2022

  12. [12]

    Large scale distributed collaborative unlabeled motion planning with graph policy gradients,

    A. Khan,et al., “Large scale distributed collaborative unlabeled motion planning with graph policy gradients,”IEEE Robotics and Automation Letters, vol. 6, no. 3, pp. 5340–5347, 2021

    work page 2021

  13. [13]

    Generalizability of graph neural net- works for decentralized unlabeled motion planning,

    S. Muthusamy,et al., “Generalizability of graph neural net- works for decentralized unlabeled motion planning,”arXiv preprint arXiv:2409.19829, 2024

    work page arXiv 2024

  14. [14]

    Graph policy gradients for large scale unlabeled motion planning with constraints,

    A. Khan,et al., “Graph policy gradients for large scale unlabeled motion planning with constraints,”arXiv preprint arXiv:1909.10704, 2019

    work page arXiv 1909

  15. [15]

    Hierarchical relational graph learning for autonomous multirobot cooperative navigation in dynamic environments,

    T. Wang,et al., “Hierarchical relational graph learning for autonomous multirobot cooperative navigation in dynamic environments,”IEEE Transactions on Computer-Aided Design of Integrated Circuits and Systems, vol. 42, no. 11, pp. 3559–3570, 2023

    work page 2023

  16. [16]

    On the hardness of unlabeled multi-robot motion planning,

    K. Soloveyet al., “On the hardness of unlabeled multi-robot motion planning,”The International Journal of Robotics Research, vol. 35, no. 14, pp. 1750–1759, 2016

    work page 2016

  17. [17]

    Capt: Concurrent assignment and planning of trajectories for multiple robots,

    M. Turpin,et al., “Capt: Concurrent assignment and planning of trajectories for multiple robots,”The International Journal of Robotics Research, vol. 33, no. 1, pp. 98–112, 2014

    work page 2014

  18. [18]

    The hungarian method for the assignment problem,

    H. W. Kuhn, “The hungarian method for the assignment problem,” Naval research logistics quarterly, vol. 2, no. 1-2, pp. 83–97, 1955

    work page 1955

  19. [19]

    Energy-optimal goal assignment of multi-agent system with goal trajectories in polynomials,

    H. Bang,et al., “Energy-optimal goal assignment of multi-agent system with goal trajectories in polynomials,” in29th Mediterranean Conference on Control and Automation (MED), 2021, pp. 1228–1233

    work page 2021

  20. [20]

    Decentralized unlabeled multi-agent navigation in continuous space,

    S. Dergachevet al., “Decentralized unlabeled multi-agent navigation in continuous space,” inInternational Conference on Interactive Collaborative Robotics. Springer, 2024, pp. 186–200

    work page 2024

  21. [21]

    Convergent multiagent formation control with collision avoidance,

    J. Hu,et al., “Convergent multiagent formation control with collision avoidance,”IEEE Transactions on Robotics, vol. 36, no. 6, pp. 1805– 1818, 2020

    work page 2020

  22. [22]

    Decentralized goal assignment and safe trajectory generation in multirobot networks via multiple lyapunov functions,

    D. Panagou,et al., “Decentralized goal assignment and safe trajectory generation in multirobot networks via multiple lyapunov functions,” IEEE Transactions on Automatic Control, vol. 65, no. 8, pp. 3365– 3380, 2019

    work page 2019

  23. [23]

    A distributed pipeline for scalable, deconflicted formation flying,

    P. C. Lusk,et al., “A distributed pipeline for scalable, deconflicted formation flying,”IEEE Robotics and Automation Letters, vol. 5, no. 4, pp. 5213–5220, 2020

    work page 2020

  24. [24]

    Distributed assignment with limited communication for multi-robot multi-target tracking,

    Y . Sung,et al., “Distributed assignment with limited communication for multi-robot multi-target tracking,”Autonomous robots, vol. 44, no. 1, pp. 57–73, 2020

    work page 2020

  25. [25]

    Multi-robot task allocation and path planning with maximum range constraints,

    G. Xu,et al., “Multi-robot task allocation and path planning with maximum range constraints,”arXiv preprint arXiv:2409.06531, 2024

    work page arXiv 2024

  26. [26]

    Swarm assignment and trajectory optimization using variable-swarm, distributed auction assignment and sequential convex programming,

    D. Morgan,et al., “Swarm assignment and trajectory optimization using variable-swarm, distributed auction assignment and sequential convex programming,”The International Journal of Robotics Re- search, vol. 35, no. 10, pp. 1261–1285, 2016

    work page 2016

  27. [27]

    Graph neural network for decentralized multi-robot goal assignment,

    M. Goarinet al., “Graph neural network for decentralized multi-robot goal assignment,”IEEE Robotics and Automation Letters, vol. 9, no. 5, pp. 4051–4058, 2024

    work page 2024

  28. [28]

    Learning safe unlabeled multi-robot planning with motion constraints,

    A. Khan,et al., “Learning safe unlabeled multi-robot planning with motion constraints,” inIEEE/RSJ International Conference on Intelli- gent Robots and Systems (IROS), 2019, pp. 7558–7565

    work page 2019

  29. [29]

    Cooperative multi-robot hierarchical reinforce- ment learning,

    G. E. Setyawan,et al., “Cooperative multi-robot hierarchical reinforce- ment learning,”International Journal of Advanced Computer Science and Applications, vol. 13, no. 9, 2022

    work page 2022

  30. [30]

    Joint optimization of multi-uav target assignment and path planning based on multi-agent reinforcement learning,

    H. Qie,et al., “Joint optimization of multi-uav target assignment and path planning based on multi-agent reinforcement learning,”IEEE access, vol. 7, pp. 146 264–146 272, 2019

    work page 2019

  31. [31]

    Multirobot coordination with deep reinforcement learning in complex environments,

    D. Wanget al., “Multirobot coordination with deep reinforcement learning in complex environments,”Expert Systems with Applications, vol. 180, p. 115128, 2021

    work page 2021

  32. [32]

    End-to-end deep reinforcement learning for decentralized task allocation and navigation for a multi-robot system,

    A. Elfakharanyet al., “End-to-end deep reinforcement learning for decentralized task allocation and navigation for a multi-robot system,” Applied Sciences, vol. 11, no. 7, p. 2895, 2021

    work page 2021

  33. [33]

    Autonomous multi-robot allocation and formation control for remote sensing in environmental exploration,

    T. Sellers,et al., “Autonomous multi-robot allocation and formation control for remote sensing in environmental exploration,” inAu- tonomous Systems: Sensors, Processing, and Security for Ground, Air , Sea, and Space V ehicles and Infrastructure 2023, vol. 12540. SPIE, 2023, pp. 250–266

    work page 2023

  34. [34]

    A comprehensive survey on graph neural networks,

    Z. Wu,et al., “A comprehensive survey on graph neural networks,” IEEE transactions on neural networks and learning systems, vol. 32, no. 1, pp. 4–24, 2020

    work page 2020

  35. [35]

    A critical review of communications in multi-robot systems,

    J. Gielis,et al., “A critical review of communications in multi-robot systems,”Current robotics reports, vol. 3, no. 4, pp. 213–225, 2022

    work page 2022

  36. [36]

    Multi-robot obstacle-avoidance formation based on graph neural networks and imitation learning,

    Y . Wang,et al., “Multi-robot obstacle-avoidance formation based on graph neural networks and imitation learning,” inChina Automation Congress (CAC), 2024, pp. 5499–5504

    work page 2024

  37. [37]

    A framework for real-world multi-robot sys- tems running decentralized gnn-based policies,

    J. Blumenkamp,et al., “A framework for real-world multi-robot sys- tems running decentralized gnn-based policies,” inIEEE International Conference on Robotics and Automation (ICRA), 2022, pp. 8772– 8778

    work page 2022

  38. [38]

    Minimum snap trajectory generation and control for quadrotors,

    D. Mellingeret al., “Minimum snap trajectory generation and control for quadrotors,” in2011 IEEE international conference on robotics and automation. IEEE, 2011, pp. 2520–2525

    work page 2011

  39. [39]

    Graph attention networks,

    P. Veli ˇckovi´c,et al., “Graph attention networks,” inInternational Conference on Learning Representations, 2018

    work page 2018

  40. [40]

    Decentralized nonlinear model predictive control for safe collision avoidance in quadrotor teams with limited detection range,

    M. Goarin,et al., “Decentralized nonlinear model predictive control for safe collision avoidance in quadrotor teams with limited detection range,” inIEEE International Conference on Robotics and Automation (ICRA), 2025, pp. 5387–5393

    work page 2025

  41. [41]

    Deep graph library: Towards efficient and scalable deep learning on graphs,

    M. Y . Wang, “Deep graph library: Towards efficient and scalable deep learning on graphs,” inICLR workshop on representation learning on graphs and manifolds, 2019

    work page 2019

  42. [42]

    Learning quadrotor dynamics for precise, safe, and agile flight control,

    A. Savioloet al., “Learning quadrotor dynamics for precise, safe, and agile flight control,”Annual Reviews in Control, vol. 55, pp. 45–60, 2023

    work page 2023

  43. [43]

    acados – a modular open-source framework for fast embedded optimal control,

    R. Verschueren,et al., “acados – a modular open-source framework for fast embedded optimal control,”Mathematical Programming Com- putation, 2021

    work page 2021

  44. [44]

    Estimation, control, and planning for aggressive flight with a small quadrotor with a single camera and imu,

    G. Loianno,et al., “Estimation, control, and planning for aggressive flight with a small quadrotor with a single camera and imu,”IEEE Robotics and Automation Letters, vol. 2, no. 2, pp. 404–411, April 2017

    work page 2017