Pith. sign in

REVIEW 3 major objections 5 minor 46 references

YOPOv2-Tracker: An End-to-End Agile Tracking and Navigation Framework from Perception to Action

T0 review · 3 major / 5 minor · reviewed 2026-08-15 · deepseek-v4-flash

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

Pith's one-line read The paper claims that a single fully convolutional network can turn RGB-D images and onboard state directly into attitude and thrust commands, enabling a small quadrotor to track an unpredictable human at up to 6 m/s through dense forests…

desk verdict Real end-to-end tracking with credible flight demos, but the headline smoothness/safety metrics are the training loss itself and the feasibility gap is real. read the letter →

arxiv 2505.06923 v1 pith:LYZD3LRK submitted 2025-05-11 cs.RO

Junjie Lu , Yulin Hui , Xuewei Zhang , Wencan Feng , Hongming Shen , Zhiyu Li , Bailing Tian This is my paper · ORCID
classification cs.RO
keywords end-to-endlearningaerialtargettrackingquadrotornavigationmotionprimitivestrajectoryoptimizationdisturbanceobservervision-basedcontrolRGB-Dperception
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

The paper tries to show that 'less is more' works for agile aerial tracking. Instead of chaining detection, mapping, planning, and control, one lightweight fully convolutional network can map RGB-D images and flight state directly to attitude and thrust commands. If this holds, quadrotors with limited onboard sensing and computation could track fast, evasive, uncooperative targets in cluttered environments where traditional cascaded systems are too slow. The authors claim this is the first end-to-end tracker to achieve high-speed tracking of an unpredictable target in real-world clutter, demonstrated by chasing a human at up to 6 m/s through forests and buildings with a 155 mm quadrotor. The design keeps interpretability by making each image grid cell correspond to a motion primitive, so the network performs detection-like scoring and trajectory refinement rather than black-box action regression, and it is trained without expert demonstrations or reinforcement learning by back-propagating trajectory costs computed from a privileged map.

What carries the argument

The load-bearing object is the primitive-anchored fully convolutional predictor. The image is divided into a 5 by 3 grid, and each cell corresponds to one motion primitive, a candidate endpoint on a sphere in the camera frustum described by polar angle, azimuth angle, and radius. The network predicts offsets and end derivatives that refine the primitive into a fifth-order Hermite polynomial with fixed execution time, plus a predicted cost, an objectness score, and the target position. During training, smoothness, safety, and goal costs are differentiated through the Hermite map to the network weights, so trajectory optimization and learning share one gradient path. At run time, the chosen trajectory's second derivative, corrected by a high-gain disturbance observer, is converted through differential flatness into desired attitude and thrust, removing the position controller and the reference-state mismatch that earlier end-to-end planners face.

What would settle it

During a 6 to 8 m/s forest or building run, record the commanded acceleration from the network's trajectory alongside the physical acceleration available from the thrust-to-weight ratio and the attitude controller's tracking error; if commanded accelerations regularly exceed the vehicle's capability, or if the disturbance observer's compensation diverges during sharp target turns, the core tracking claim fails.

Watch

Extended reading notes

Core claim

The central claim is that target tracking does not need the usual serial pipeline of detection, occupancy mapping, path search, trajectory optimization, and position control run one after another. A single fully convolutional network, called YOPOv2-Tracker, takes an RGB-D image and the current velocity and acceleration and outputs, for each of a grid of motion primitives spanning the camera frustum, refined offsets, endpoint derivatives, a trajectory cost, an objectness score, and a target position. The best primitive is turned into a fifth-order Hermite trajectory, and a disturbance observer converts the desired acceleration into thrust and attitude. Because mapping, planning, and detection are folded into one forward pass, the system responds in about 3 ms in simulation and 8.2 ms onboard, and in experiments it tracks an uncooperative human at up to 6 m/s through dense forest and buildings. The training signal is itself part of the claim: gradients of smoothness, safety, and goal costs from a privileged distance-to-obstacle map are back-propagated through the trajectory to the network weights, so no expert demonstrations or reinforcement-learning interactions are required.

Load-bearing premise

The trajectory the network generates is assumed to be physically flyable by the actual quadrotor, even though the training cost never checks whether the commanded acceleration stays within the motors' limits; if the planned path demands more acceleration than the attitude controller can provide, the safety and tracking guarantees degrade.

Editorial extensions

If this is right

  • Tracking success at target escape speeds of 3 to 7 m/s in simulation, and real-world human tracking at up to 6 m/s, becomes possible using only onboard RGB-D perception on a small quadrotor.
  • Perception-to-action latency drops to about 3 ms in simulation and 8.2 ms onboard, an order of magnitude faster than the compared cascaded trackers.
  • The same network, with the target-related outputs removed, acts as a high-speed navigation policy, so a single training recipe covers both tracking and obstacle avoidance.
  • Because the network outputs a chosen trajectory rather than a raw action, the system remains interpretable: each primitive corresponds to a feasible candidate within the camera's field of view.
  • Training by back-propagating privileged trajectory costs offers a third path between imitation learning and reinforcement learning for robot navigation tasks.

Reading between the lines

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

  • The anchor-primitive mechanism could transfer to multi-target tracking or to ground robots, since the objectness head could be extended to multiple detections per grid cell.
  • The fixed execution time is a bottleneck; learning or adapting the execution time as a network output would likely improve agility during fast escapes and sharp target turns.
  • The recipe of using a differentiable cost map as the critic could be applied to other optimization-based robot tasks beyond flight, such as manipulation or racing, wherever a privileged cost can be computed.
  • The claim that target motion prediction is unnecessary could be tested directly by adding a short-horizon predictive head and comparing success rates; if prediction helps substantially, the 'unpredictable target' argument would need qualification.
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 YOPOv2-Tracker, an end-to-end quadrotor tracking and navigation framework that directly maps RGB-D images and onboard state estimates to thrust and attitude commands. The architecture is a single fully convolutional network that predicts offsets and costs for a set of motion primitives, along with objectness scores and target positions, thereby integrating detection, search, and trajectory optimization into one forward pass. Training is performed by back-propagating differentiable trajectory costs (smoothness, safety, goal) and detection losses through the network without expert demonstrations or reinforcement learning. At deployment, a disturbance observer compensates for model uncertainty, and an EKF-based target estimator drives the yaw. The authors report simulation comparisons with Elastic Tracker and Visibility-aware Tracker, plus real-world flights in dense forests and buildings at tracking speeds up to 6 m/s and navigation speeds up to 8 m/s.

Significance. If the claimed performance holds, this is a notable contribution: the system demonstrates that a single lightweight network can replace the traditional detection-mapping-planning-control cascade while preserving interpretability through primitive anchors and explicit trajectory costs. The real-world flights with an uncooperative human target in cluttered environments are credible evidence of practical utility, and the low reported latency (3 ms in simulation, 8.2 ms onboard) is an important engineering result. The code-release commitment strengthens reproducibility. However, the quantitative validation of smoothness and safety is partially circular because the reported metrics are the same cost functions used in the training objective, and the training objective lacks a dynamic-feasibility term, leaving a gap between the planned polynomial trajectories and the physical actuator limits. These issues reduce confidence in the quantitative comparisons until addressed, but they do not invalidate the central design contribution.

major comments (3)
  1. [Sec. IV-B1, Fig. 10E-F; Eqs. (16), (17), (20), (25)] The smoothness and safety metrics reported in Fig. 10E-F are the very cost functions J_s and J_c that appear as training losses in Eq. (16), and the predicted cost y_c used for trajectory selection is supervised by the same quantities via Eq. (25). Therefore the finding that YOPOv2 achieves better smoothness and safety than Elastic Tracker and Visibility-aware Tracker is partly by construction: the network is explicitly trained to minimize these integrals, whereas the baselines optimize different objectives (e.g., hard visibility or corridor constraints). To substantiate the claimed superiority, please either report metrics that are not part of the training objective (for example, commanded-versus-actual acceleration, control effort, or the baseline methods' own cost definitions) or clearly acknowledge this circularity and soften the corresponding comparison.
  2. [Sec. III-C, III-D, III-E; Eqs. (7), (16); Sec. IV-C2] The training loss in Eq. (16) contains no actuator-feasibility term. Equation (7) bounds only the endpoint derivatives through the tanh nonlinearity, and the execution time is fixed to T = 2r/(alpha v_max), but the interior acceleration of the fifth-order Hermite polynomial (Eq. (2)) can substantially exceed the endpoint values, especially when the current state has large velocity or acceleration and the refined radius r' is small. The disturbance observer of Eq. (15) can compensate for lumped disturbances but cannot increase collective thrust beyond the physical limit; if the desired wrench F from Eq. (13) exceeds the platform's 4.7:1 thrust-to-weight ratio, the attitude loop saturates and the safety and tracking claims in Section IV are not guaranteed. Figure 19 shows one 8 m/s navigation trial with attitude accelerations mostly within 5 m/s², but that is not a certificate over the primitive library or over 6-7 m/s tracking with an evasive target. Please provide an analysis of the maximum feasible acceleration over the primitive and offset space, or add a feasibility penalty or constraint to the training objective, and report commanded-versus-actual acceleration statistics across the real-world trials.
  3. [Sec. IV-B1, Fig. 10B] The success-rate comparison in Fig. 10B is based on only 10 random forest realizations per target speed. With such a small sample and a stochastic evader whose goal is randomly switched, the reported differences at intermediate speeds may fall within sampling variability; no confidence intervals or statistical significance tests are provided. Since the central claim of superior success rates is load-bearing, please add more trials or report variance and confidence bounds to demonstrate that the differences are not due to chance.
minor comments (5)
  1. [Abstract] The abstract contains a typo: 'limitting' should be 'limiting'.
  2. [Author affiliation] The university name is misspelled as 'Egineering'; it should be 'Engineering'.
  3. [Sec. III-C1] The network input is described as '6-dimensional states (initial velocity and acceleration)', but the text later refers to 'state inputs concatenated with image features' without specifying whether the state includes other quantities such as the goal direction; please clarify the exact input vector.
  4. [Sec. IV-B3] The phrase 'over 2 times faster than existing state-of-the-art methods in cluttered real-world scenarios' is based on comparing the maximum tracking speed (6 m/s) with the 2.5 m/s reported in [2]; please state the comparison explicitly and note that different platforms and environment difficulties may limit the directness of this comparison.
  5. [Sec. III-E, Eq. (16)] The notation for the binary indicators 1_pos and 1_neg is clear, but the handling of ignored samples in the loss could be stated more explicitly; the current text notes they are not considered for L_obj but does not clearly specify which trajectory losses apply to them.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the end-to-end perception-to-action claim is supported by independent success-rate, latency, and real-world evidence, and the cost-metric alignment is an evaluation issue, not a derivation collapse.

full rationale

The paper's derivation chain is self-contained: the network outputs primitive offsets and derivatives, trajectories are closed-form Hermite polynomials (Eqs. 2-7), desired attitude is obtained from the flatness mapping and disturbance observer (Eqs. 11-15), and training back-propagates analytic trajectory-cost gradients (Eqs. 16-24). The claimed success rate, latency, and real-world tracking are measured by independent criteria (collision/success, wall-clock perception-to-action time, realized flight speed in forest and building experiments) rather than by the training losses. The fact that J_s and J_c appear both in the training objective and in Fig. 10E-F is an evaluation-alignment issue, not a circular reduction: the executed trajectories are not defined as the minimizers of these costs; they are generated from learned offsets and selected by a separately supervised cost head, and the comparison against external trackers on randomized held-out forests provides non-tautological evidence. No load-bearing self-citation or imported uniqueness theorem is used; prior work [6] is cited only as a baseline and point of departure.

Assumptions & free parameters 6 free parameters · 7 assumptions · 0 invented entities

The central claim rests primarily on standard quadrotor assumptions (differential flatness, point-mass model) and on design choices (horizon r, weights, normalization bounds) that are not fully specified in the preprint. The method introduces no new physical entities. The most consequential unstated premise is that the low-dimensional fixed-time polynomial family remains dynamically feasible, since feasibility is not enforced in the training cost.

free parameters (6)
  • Trajectory cost weights λ_s, λ_c, λ_g = not given (λ1=0.2, λ2=0.5 in Eq. 16)
    Hand-selected weights balancing smoothness, safety, and goal terms in the training loss and in the selection cost; they directly shape the behavior of the policy.
  • Planning horizon radius r and execution time T = 2r/(α vmax) = not given
    The primitive endpoints lie on a sphere of radius r and T is fixed by r and the speed scaling α; these choices determine trajectory length, speed, and safety margin.
  • Training speed bounds vmax, amax = not given
    Used to normalize states and outputs (Eqs. 5 and 7); they define the feasible output range of the network and the speed envelope of the policy.
  • Safety potential function c(·) = unspecified, described only as e.g. exponential of negative distance
    J_c in Eq. 20 requires an explicit obstacle potential; its shape and scale determine the safety gradients and thus the trained avoidance behavior.
  • Disturbance observer gains α1, α2, ζ = not given
    Eq. 15 defines a high-gain observer whose convergence and noise sensitivity depend on these constants, which are not specified in the paper.
  • EKF process noise and inconsistency threshold = not given ('large' process noise, threshold unspecified)
    These parameters control target state estimation and false-positive rejection in Section III-C.2 and are not quantified.
assumptions (7)
  • domain assumption Quadrotor dynamics are differentially flat; desired attitude can be computed from acceleration, yaw, and gravity (Eqs. 11-13).
    The control strategy in Section III-D assumes a rigid-body quadrotor model with four rotors and uses differential flatness to map acceleration to attitude and thrust.
  • domain assumption A fixed-time fifth-order polynomial (Hermite) trajectory suffices to represent agile, collision-free maneuvers for tracking.
    The trajectory representation in Section III-B restricts motion to a single polynomial per axis with fixed T; if the required maneuver cannot be captured by this low-dimensional family, the policy cannot execute it.
  • domain assumption The privileged ESDF map and its gradients used in training are accurate and representative of deployment environments.
    Training back-propagates J_c from ESDF gradients (Eq. 21); if the map or its gradients differ from real depth observations, the learned safety behavior may not transfer.
  • domain assumption A point-mass model with a lumped disturbance d(t) captures all unmodeled aerodynamic and external forces.
    Section III-D replaces full dynamics with Eq. 11 and assumes the high-gain observer (Eq. 15) can estimate d(t) online.
  • domain assumption The target is visible at mission start and objectness and position can be learned from RGB-D imagery.
    The tracking task assumes a known initial target in view and relies on the network's detection predictions plus EKF for subsequent frames.
  • standard math Time scaling f(αt) with normalized inputs preserves trajectory geometry and safety across different flight speeds.
    Eq. 5 relies on the polynomial scaling property; this is mathematically valid for the trajectory shape but assumes perception and latency do not break the invariance.
  • domain assumption Every predicted Hermite trajectory between the current state and the bounded endpoint is dynamically feasible for the quadrotor, even though no feasibility term is in the cost.
    The training cost (Eq. 16) contains no actuator or acceleration limit; the paper relies on the tanh-bounded outputs and the fixed T to keep trajectories within the flight envelope without explicitly verifying it.

how reviews work

0 comments
Cite this review

Pith. "Pith review of YOPOv2-Tracker: An End-to-End Agile Tracking and Navigation Framework from Perception to Action." pith.science (2026). https://pith.science/paper/LYZD3LRK

@misc{pith2026250506923,
  author       = {Pith},
  title        = {Pith review of: YOPOv2-Tracker: An End-to-End Agile Tracking and Navigation Framework from Perception to Action},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/LYZD3LRK}},
  note         = {Machine review of arXiv:2505.06923}
}
read the original abstract

Traditional target tracking pipelines including detection, mapping, navigation, and control are comprehensive but introduce high latency, limitting the agility of quadrotors. On the contrary, we follow the design principle of "less is more", striving to simplify the process while maintaining effectiveness. In this work, we propose an end-to-end agile tracking and navigation framework for quadrotors that directly maps the sensory observations to control commands. Importantly, leveraging the multimodal nature of navigation and detection tasks, our network maintains interpretability by explicitly integrating the independent modules of the traditional pipeline, rather than a crude action regression. In detail, we adopt a set of motion primitives as anchors to cover the searching space regarding the feasible region and potential target. Then we reformulate the trajectory optimization as regression of primitive offsets and associated costs considering the safety, smoothness, and other metrics. For tracking task, the trajectories are expected to approach the target and additional objectness scores are predicted. Subsequently, the predictions, after compensation for the estimated lumped disturbance, are transformed into thrust and attitude as control commands for swift response. During training, we seamlessly integrate traditional motion planning with deep learning by directly back-propagating the gradients of trajectory costs to the network, eliminating the need for expert demonstration in imitation learning and providing more direct guidance than reinforcement learning. Finally, we deploy the algorithm on a compact quadrotor and conduct real-world validations in both forest and building environments to demonstrate the efficiency of the proposed method.

Figures

Figures reproduced from arXiv: 2505.06923 by the authors.

Figure 1
Figure 1. Overview of the proposed aerial tracking and navigation system. (A-B) show the agile tracking in dense forests and cluttered buildings, while (C-D) [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Comparison with existing tracking pipeline. [PITH_FULL_IMAGE:figures/full_fig_p002_2.png] view at source ↗
Figure 3
Figure 3. System Overview. The proposed framework takes RGB-D images and state observations as inputs, and outputs attitude and thrust as control commands. [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗
Figures from the paper (14 more)
Figure 5
Figure 5. Figure 5: We predict offsets and derivatives to improve the primitives (red) and [PITH_FULL_IMAGE:figures/full_fig_p006_5.png]
Figure 7
Figure 7. Figure 7: Planning in differential flatness space and tracking by position [PITH_FULL_IMAGE:figures/full_fig_p007_7.png]
Figure 8
Figure 8. Figure 8: Due to the multimodal nature of navigation, the distance to limited [PITH_FULL_IMAGE:figures/full_fig_p009_8.png]
Figure 9
Figure 9. Figure 9: Probability density function of state sampling. [PITH_FULL_IMAGE:figures/full_fig_p010_9.png]
Figure 10
Figure 10. Figure 10: Tracking comparison of real-time performance, success rate, visibility, smoothness, and safety metrics. (A) Real-time performance comparison with [PITH_FULL_IMAGE:figures/full_fig_p011_10.png]
Figure 11
Figure 11. Figure 11: Visualization of large-scale tracking in simulation, where the target’s path is drawn in white and the tracker’s path is color-coded by speed. In [PITH_FULL_IMAGE:figures/full_fig_p013_11.png]
Figure 12
Figure 12. Figure 12: Distribution of target estimation error throughout tracking process. [PITH_FULL_IMAGE:figures/full_fig_p013_12.png]
Figure 13
Figure 13. Figure 13: Agile tracking in real-world dense forests. (A) Experimental scenario with tree spacing of 3-4 m. (B) The tracking path of quadrotor in large-scale [PITH_FULL_IMAGE:figures/full_fig_p014_13.png]
Figure 14
Figure 14. Figure 14: Agile tracking in real-world buildings. (A) Large-scale architectural scene spanning an area of 120 [PITH_FULL_IMAGE:figures/full_fig_p015_14.png]
Figure 15
Figure 15. Figure 15: Visualization of high-speed navigation at 10 m/s in simulation, where the executed path is depicted by odometry axes. In snapshots A-F, all predicted [PITH_FULL_IMAGE:figures/full_fig_p016_15.png]
Figure 16
Figure 16. Figure 16: Success rate comparison at various speeds and clearances. [PITH_FULL_IMAGE:figures/full_fig_p016_16.png]
Figure 17
Figure 17. Figure 17: Real-world Experiment 1. (A) Experimental scenario with tree spacing of 3-4 m. (B) Visualization of flight path during high-speed navigation, with [PITH_FULL_IMAGE:figures/full_fig_p017_17.png]
Figure 18
Figure 18. Figure 18: Real-world Experiment 2. (A) Visualization of flight path with the [PITH_FULL_IMAGE:figures/full_fig_p017_18.png]
Figure 19
Figure 19. Figure 19: Measurements of the disturbance observer, desired acceleration, and [PITH_FULL_IMAGE:figures/full_fig_p018_19.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

46 extracted references · 32 canonical work pages

  1. [43]

    D-vat: End- to-end visual active tracking for micro aerial vehicles,

    A. Dionigi, S. Felicioni, M. Leomanni, and G. Costante, “D-vat: End- to-end visual active tracking for micro aerial vehicles,”IEEE Robotics and Automation Letters, 2024

    work page 2024

  2. [1]

    Elastic tracker: A spatio-temporal trajectory planner for flexible aerial tracking,

    J. Ji, N. Pan, C. Xu, and F. Gao, “Elastic tracker: A spatio-temporal trajectory planner for flexible aerial tracking,” in2022 International Conference on Robotics and Automation (ICRA), pp. 47–53. IEEE, 2022

    work page 2022

  3. [2]

    Visibility-aware trajectory optimization with application to aerial tracking,

    Q. Wang, Y . Gao, J. Ji, C. Xu, and F. Gao, “Visibility-aware trajectory optimization with application to aerial tracking,” in2021 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), pp. 5249–5256. IEEE, 2021

    work page 2021

  4. [3]

    Adaptive tracking and perching for quadrotor in dynamic scenarios,

    Y . Gao, J. Ji, Q. Wang, R. Jin, Y . Lin, Z. Shang, Y . Cao, S. Shen, C. Xu, and F. Gao, “Adaptive tracking and perching for quadrotor in dynamic scenarios,”IEEE Transactions on Robotics, 2023

    work page 2023

  5. [4]

    Learning high-speed flight in the wild,

    A. Loquercio, E. Kaufmann, R. Ranftl, M. M ¨uller, V . Koltun, and D. Scaramuzza, “Learning high-speed flight in the wild,”Science Robotics, vol. 6, no. 59, p. eabg5810, 2021

    2021

  6. [5]

    Back to newton’s laws: Learning vision-based agile flight via differentiable physics,

    Y . Zhang, Y . Hu, Y . Song, D. Zou, and W. Lin, “Back to newton’s laws: Learning vision-based agile flight via differentiable physics,”arXiv preprint arXiv:2407.10648, 2024

    arXiv 2024

  7. [6]

    You only plan once: A learning-based one-stage planner with guidance learning,

    J. Lu, X. Zhang, H. Shen, L. Xu, and B. Tian, “You only plan once: A learning-based one-stage planner with guidance learning,”IEEE Robotics and Automation Letters, 2024

    work page 2024

  8. [7]

    Faster r-cnn: Towards real-time object detection with region proposal networks,

    S. Ren, K. He, R. Girshick, and J. Sun, “Faster r-cnn: Towards real-time object detection with region proposal networks,”IEEE transactions on pattern analysis and machine intelligence, vol. 39, no. 6, pp. 1137–1149, 2016

    2016

Show all 46 references
  1. [8]

    Robust real-time uav replanning using guided gradient-based optimization and topological paths,

    B. Zhou, F. Gao, J. Pan, and S. Shen, “Robust real-time uav replanning using guided gradient-based optimization and topological paths,”2020 IEEE International Conference on Robotics and Automation (ICRA), pp. 1208–1214, 2019. [Online]. Available: https://api.semanticscholar.or...

    2020

  2. [9]

    Tgk-planner: An efficient topology guided kinodynamic planner for autonomous quadrotors,

    H. Ye, X. Zhou, Z. Wang, C. Xu, J. Chu, and F. Gao, “Tgk-planner: An efficient topology guided kinodynamic planner for autonomous quadrotors,”IEEE Robotics and Automation Letters, vol. 6, no. 2, pp. 494–501, 2020

    2020

  3. [10]

    You only look once: Unified, real-time object detection,

    J. Redmon, S. Divvala, R. Girshick, and A. Farhadi, “You only look once: Unified, real-time object detection,” inProceedings of the IEEE conference on computer vision and pattern recognition, pp. 779–788, 2016

    2016

  4. [11]

    Dtpp: Differentiable joint conditional prediction and cost evaluation for tree policy planning in autonomous driving,

    Z. Huang, P. Karkus, B. Ivanovic, Y . Chen, M. Pavone, and C. Lv, “Dtpp: Differentiable joint conditional prediction and cost evaluation for tree policy planning in autonomous driving,” in2024 IEEE International Conference on Robotics and Automation (ICRA), pp. 6806–6812. IEEE, 2024

    2024

  5. [12]

    Deep-panther: Learning-based perception- aware trajectory planner in dynamic environments,

    J. Tordesillas and J. P. How, “Deep-panther: Learning-based perception- aware trajectory planner in dynamic environments,”IEEE Robotics and Automation Letters, vol. 8, no. 3, pp. 1399–1406, 2023

    2023

  6. [13]

    Reach- ing the limit in autonomous racing: Optimal control versus reinforcement learning,

    Y . Song, A. Romero, M. M ¨uller, V . Koltun, and D. Scaramuzza, “Reach- ing the limit in autonomous racing: Optimal control versus reinforcement learning,”Science Robotics, vol. 8, no. 82, p. eadg1462, 2023

    2023

  7. [14]

    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,”arXiv preprint arXiv:2406.12505, 2024

    2024 arXiv

  8. [15]

    Vision transformers for end-to-end vision- based quadrotor obstacle avoidance,

    A. Bhattacharya, N. Rao, D. Parikh, P. Kunapuli, Y . Wu, Y . Tao, N. Matni, and V . Kumar, “Vision transformers for end-to-end vision- based quadrotor obstacle avoidance,”arXiv preprint arXiv:2405.10391, 2024

    2024 arXiv

  9. [16]

    Learning minimum-time flight in cluttered environments,

    R. Penicka, Y . Song, E. Kaufmann, and D. Scaramuzza, “Learning minimum-time flight in cluttered environments,”IEEE Robotics and Automation Letters, vol. 7, no. 3, pp. 7209–7216, 2022

    2022

  10. [18]

    Online quadrotor trajectory generation and autonomous navigation on point clouds,

    F. Gao and S. Shen, “Online quadrotor trajectory generation and autonomous navigation on point clouds,” in2016 IEEE International Symposium on Safety, Security, and Rescue Robotics (SSRR), pp. 139–

  11. [19]

    Real-time safe trajectory generation for quadrotor flight in cluttered environments,

    J. Chen, K. Su, and S. Shen, “Real-time safe trajectory generation for quadrotor flight in cluttered environments,” in2015 IEEE International Conference on Robotics and Biomimetics (ROBIO), pp. 1678–1685. IEEE, 2015

    2015

  12. [20]

    Gradient-based online safe trajectory generation for quadrotor flight in complex environments,

    F. Gao, Y . Lin, and S. Shen, “Gradient-based online safe trajectory generation for quadrotor flight in complex environments,” in2017 IEEE/RSJ international conference on intelligent robots and systems (IROS), pp. 3681–3688. IEEE, 2017

    2017

  13. [21]

    Continuous-time trajectory optimization for online uav replan- ning,

    H. Oleynikova, M. Burri, Z. Taylor, J. Nieto, R. Siegwart, and E. Gal- ceran, “Continuous-time trajectory optimization for online uav replan- ning,” in2016 IEEE/RSJ international conference on intelligent robots and systems (IROS), pp. 5332–5339. IEEE, 2016

    2016

  14. [22]

    Robust and effi- cient quadrotor trajectory generation for fast autonomous flight,

    B. Zhou, F. Gao, L. Wang, C. Liu, and S. Shen, “Robust and effi- cient quadrotor trajectory generation for fast autonomous flight,”IEEE Robotics and Automation Letters, vol. 4, no. 4, pp. 3529–3536, 2019

    2019

  15. [23]

    Ego-planner: An esdf- free gradient-based local planner for quadrotors,

    X. Zhou, Z. Wang, H. Ye, C. Xu, and F. Gao, “Ego-planner: An esdf- free gradient-based local planner for quadrotors,”IEEE Robotics and Automation Letters, vol. 6, no. 2, pp. 478–485, 2020

    2020

  16. [24]

    Flight with limited field of view: A parallel and gradient-free strategy for micro aerial vehicle,

    H. Lu, Q. Zong, S. Lai, B. Tian, and L. Xie, “Flight with limited field of view: A parallel and gradient-free strategy for micro aerial vehicle,” IEEE Transactions on Industrial Electronics, vol. 69, no. 9, pp. 9258– 9267, 2021

    2021

  17. [25]

    Bubble planner: Planning high-speed smooth quadrotor trajectories using receding corridors,

    Y . Ren, F. Zhu, W. Liu, Z. Wang, Y . Lin, F. Gao, and F. Zhang, “Bubble planner: Planning high-speed smooth quadrotor trajectories using receding corridors,” in2022 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), pp. 6332–6339. IEEE, 2022

    2022

  18. [26]

    Safety-assured high-speed navigation for mavs,

    Y . Ren, F. Zhu, G. Lu, Y . Cai, L. Yin, F. Kong, J. Lin, N. Chen, and F. Zhang, “Safety-assured high-speed navigation for mavs,”Science Robotics, vol. 10, no. 98, p. eado6187, 2025

    2025

  19. [27]

    Learning monocular reactive uav control in cluttered natural environments,

    S. Ross, N. Melik-Barkhudarov, K. S. Shankar, A. Wendel, D. Dey, J. A. Bagnell, and M. Hebert, “Learning monocular reactive uav control in cluttered natural environments,” in2013 IEEE international conference on robotics and automation, pp. 1765–1772. IEEE, 2013

    2013

  20. [28]

    Lpnet: A reaction- based local planner for autonomous collision avoidance using imitation learning,

    J. Lu, B. Tian, H. Shen, X. Zhang, and Y . Hui, “Lpnet: A reaction- based local planner for autonomous collision avoidance using imitation learning,”IEEE Robotics and Automation Letters, vol. 8, no. 11, pp. 7058–7065, 2023

    2023

  21. [29]

    Motion primitives- based navigation planning using deep collision prediction,

    H. Nguyen, S. H. Fyhn, P. De Petris, and K. Alexis, “Motion primitives- based navigation planning using deep collision prediction,” in2022 International Conference on Robotics and Automation (ICRA), pp. 9660–9667. IEEE, 2022

    2022

  22. [30]

    Mavrl: Learn to fly in cluttered environments with varying speed,

    H. Yu, C. De Wagter, and G. C. E. de Croon, “Mavrl: Learn to fly in cluttered environments with varying speed,”IEEE Robotics and Automation Letters, 2024. 19

    2024

  23. [31]

    Learning agility adaptation for flight in clutter,

    G. Zhao, T. Wu, Y . Chen, and F. Gao, “Learning agility adaptation for flight in clutter,”arXiv preprint arXiv:2403.04586, 2024

    2024 arXiv

  24. [32]

    Deep learning for opti- mization of trajectories for quadrotors,

    Y . Wu, X. Sun, I. Spasojevic, and V . Kumar, “Deep learning for opti- mization of trajectories for quadrotors,”IEEE Robotics and Automation Letters, 2024

    2024

  25. [33]

    Autonomous flying cameraman with embedded person detection and tracking while applying cinematographic rules,

    D. Hulens and T. Goedem ´e, “Autonomous flying cameraman with embedded person detection and tracking while applying cinematographic rules,” in2017 14th Conference on Computer and Robot Vision (CRV), pp. 56–63. IEEE, 2017

    2017

  26. [34]

    Motion prediction and robust tracking of a dynamic and temporarily-occluded target by an unmanned aerial vehicle,

    J.-M. Li, C.-W. Chen, and T.-H. Cheng, “Motion prediction and robust tracking of a dynamic and temporarily-occluded target by an unmanned aerial vehicle,”IEEE Transactions on Control Systems Technology, vol. 29, no. 4, pp. 1623–1635, 2020

    2020

  27. [35]

    Fast-tracker: A robust aerial system for tracking agile target in cluttered environments,

    Z. Han, R. Zhang, N. Pan, C. Xu, and F. Gao, “Fast-tracker: A robust aerial system for tracking agile target in cluttered environments,” in2021 IEEE international conference on robotics and automation (ICRA), pp. 328–334. IEEE, 2021

    2021

  28. [36]

    Tracking a moving target in cluttered environments using a quadrotor,

    J. Chen, T. Liu, and S. Shen, “Tracking a moving target in cluttered environments using a quadrotor,” in2016 IEEE/RSJ International Con- ference on Intelligent Robots and Systems (IROS), pp. 446–453. IEEE, 2016

    2016

  29. [37]

    Integrated motion planner for real- time aerial videography with a drone in a dense environment,

    B. Jeon, Y . Lee, and H. J. Kim, “Integrated motion planner for real- time aerial videography with a drone in a dense environment,” in2020 IEEE International Conference on Robotics and Automation (ICRA), pp. 1243–1249. IEEE, 2020

    2020

  30. [38]

    Online trajectory generation of a mav for chasing a moving target in 3d dense environments,

    B. F. Jeon and H. J. Kim, “Online trajectory generation of a mav for chasing a moving target in 3d dense environments,” in2019 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), pp. 1115–1121. IEEE, 2019

    2019

  31. [39]

    Using a quadrotor to track a moving target with arbitrary relative motion patterns,

    J. Chen and S. Shen, “Using a quadrotor to track a moving target with arbitrary relative motion patterns,” in2017 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), pp. 5310–5317. IEEE, 2017

    2017

  32. [40]

    Intention-aware planner for robust and safe aerial tracking,

    Q. Ren, H. Yu, J. Dai, Z. Zheng, J. Meng, L. Xu, C. Xu, F. Gao, and Y . Cao, “Intention-aware planner for robust and safe aerial tracking,” in2024 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), pp. 8153–8160. IEEE, 2024

    2024

  33. [41]

    Coni-mpc: Cooperative non-inertial frame based model predictive con- trol,

    B. Zhang, X. Chen, Z. Li, G. Beltrame, C. Xu, F. Gao, and Y . Cao, “Coni-mpc: Cooperative non-inertial frame based model predictive con- trol,”IEEE Robotics and Automation Letters, 2023

    2023

  34. [42]

    Intercepting rogue robots: An algorithm for capturing multiple evaders with multiple pursuers,

    A. Pierson, Z. Wang, and M. Schwager, “Intercepting rogue robots: An algorithm for capturing multiple evaders with multiple pursuers,”IEEE Robotics and Automation Letters, vol. 2, no. 2, pp. 530–537, 2016

    2016

  35. [44]

    Aggressive 3-d collision avoidance for high-speed navigation

    B. T. Lopez and J. P. How, “Aggressive 3-d collision avoidance for high-speed navigation.” inICRA, pp. 5759–5765, 2017

    2017

  36. [45]

    Motion primitives-based path planning for fast and agile exploration using aerial robots,

    M. Dharmadhikari, T. Dang, L. Solanka, J. Loje, H. Nguyen, N. Khedekar, and K. Alexis, “Motion primitives-based path planning for fast and agile exploration using aerial robots,” in2020 IEEE International Conference on Robotics and Automation (ICRA), pp. 179–

  37. [46]

    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: The 16th International Symposium ISRR, pp. 649–666. Springer, 2016

    2016

  38. [47]

    Fuel: Fast uav exploration using incremental frontier structure and hierarchical planning,

    B. Zhou, Y . Zhang, X. Chen, and S. Shen, “Fuel: Fast uav exploration using incremental frontier structure and hierarchical planning,”IEEE Robotics and Automation Letters, vol. 6, no. 2, pp. 779–786, 2021

    2021

Pith tools

Reviewed August 15, 2026 · model on record in the stance chip above.