Pith. sign in

REVIEW 4 major objections 5 minor 46 references

HEPP: Hyper-efficient Perception and Planning for High-speed Obstacle Avoidance of UAVs

T0 review · 4 major / 5 minor · reviewed 2026-08-07 · deepseek-v4-flash

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

Pith's one-line read HEPP claims a complete perception-and-planning pipeline that lets a drone avoid obstacles at 15 m/s in dense, unknown environments with planning loops of a few milliseconds, taking 79.24% less time than existing methods.

desk verdict A credible, well-engineered high-speed planner whose empirical results are promising, but whose safety and near-optimality claims outrun the evidence; referee it, but require caveats and more rigorous evaluation. read the letter →

arxiv 2505.17438 v1 pith:TCFJ65DQ submitted 2025-05-23 cs.RO

Minghao Lu , Xiyu Fan , Bowen Xu , Zexuan Yan , Rui Peng , Han Chen , Lixian Zhang , Peng Lu This is my paper · ORCID
classification cs.RO
keywords high-speedUAVnavigationobstacleavoidanceincrementalrobocentricmappingEuclideansigneddistancefieldtopologicalpathsearchtrajectoryoptimizationtimeallocationlidar-basedperception
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

HEPP sets out to prove that a drone can avoid obstacles at 15 m/s in dense, unknown spaces by making every computation local: a moving point-cloud map replaces global occupancy grids and distance fields, a topological search produces several genuinely different paths, and a gradient-based optimizer with adaptive time allocation turns each path into a trajectory. The paper reports that this pipeline runs in a few milliseconds per iteration, taking 79.24% less time than existing methods at high speed, and that picking the lowest-cost trajectory among the candidates brings the flight close to the global optimum in both duration and distance. A reader should care because high-speed autonomy in clutter has previously been limited to sparse fields or learning-based systems with narrow training envelopes; this work claims a fully classical pipeline that is both fast enough and robust enough for real forest flights.

What carries the argument

The load-bearing object is RESDF, the Robocentric Euclidean Signed Distance Field: for any query position, the distance to the nearest obstacle is the Euclidean distance to the single nearest map point found by an octree search, and the gradient is the finite difference of that distance along each axis. Because the field is generated on demand rather than maintained over the whole map, the system never performs obstacle inflation or a BFS-based ESDF update. The other mechanism is the adaptive high-speed trajectory initialization: waypoints are sampled uniformly along each topological path, and the per-segment durations are set so the drone accelerates to a desired speed $v_d$, flies at that speed, then decelerates, with $v_d$ iteratively scaled down until the path length can physically support the maneuver.

What would settle it

Run the pipeline in a cluttered forest while computing a ground-truth signed distance field from dense registered scans; if any trajectory that the planner accepts has true clearance below the safety threshold $d_s$, or if the mean error between RESDF and the ground-truth distance at sampled trajectory points is comparable to $d_s$, the nearest-neighbor distance assumption is falsified.

Watch

Extended reading notes

Core claim

The central claim is that high-speed obstacle avoidance does not require a global Euclidean signed distance field or a single trajectory optimization; it can be achieved by maintaining a robocentric point cloud that returns distance and gradient on demand (RESDF), enumerating multiple topologically distinct collision-free paths with a visibility-plane search, optimizing each path with an adaptive initialization, and selecting the trajectory with the smallest cost. Under this scheme the paper reports total per-iteration latencies of 4.5–7.2 ms in simulation and 90% success at a 15 m/s speed limit in an extremely dense forest, with the planned trajectory close to the global optimum in both the temporal and spatial domains.

Load-bearing premise

The paper treats the distance to the single nearest point in a sparse, sliding point cloud as the true distance to the nearest obstacle surface, and every collision check in the optimizer uses that number.

Editorial extensions

If this is right

  • The planning loop completes in 4.53–7.22 ms across the reported simulation scenarios, so the same onboard computer can react to newly observed obstacles within one lidar frame.
  • At a 15 m/s speed limit in a simulated forest with roughly triple obstacle density, the system keeps a 90% success rate (17 of 20 flights) while reaching an average maximum speed of 13.82 m/s.
  • Removing the global ESDF update means the mapping cost stays low as map resolution increases; the reported total mapping time is about 10.5% of the second-best method in the benchmark.
  • Real-world flights in a cluttered indoor room and a forest reach 8.0 m/s and 11.0 m/s respectively, indicating that the system transfers from simulation to hardware.

Reading between the lines

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

  • Beyond the paper: the 'near-global optimum' claim depends on the $K=4$ angular sampling in the topological search, so in narrow-gap environments a gap thinner than the sampling resolution could be invisible to the optimizer; counting path diversity against a denser sampling or an obstacle skeleton would test this.
  • Beyond the paper: the single-neighbor distance estimate could be made conservative by tracking local point density or neighbor radii, which would let the same architecture claim a formal safety margin rather than relying on the tuned threshold $d_s$.
  • Beyond the paper: the cost-based trajectory selection ignores trajectory length, a limitation the paper itself acknowledges; adding length or energy to the selection criterion would likely shorten the reported 54.11 m average path length at 15 m/s.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 5 minor

Summary. The paper presents HEPP, a complete UAV navigation stack for high-speed obstacle avoidance consisting of three modules: an incremental robocentric point-cloud map with direct nearest-neighbor distance and gradient queries (IROP-Map/RESDF), an obstacle-aware topological path search that expands multiple homotopy candidates on a visual plane, and a MINCO-based trajectory optimizer with adaptive time allocation. The authors report sub-10 ms end-to-end latency, an 89.5% reduction in mapping time, a 79.24% reduction in total time relative to existing methods at 15 m/s, and successful autonomous flights in simulation and real-world cluttered environments at speeds up to 13 m/s speed limits. The abstract and contributions also claim that the planned trajectories are close to the global optimum in both temporal and spatial domains.

Significance. The paper is a solid systems contribution: the incremental mapping data structure, the topological path search, and the integrated full-stack validation are clear strengths, and the real-world flights at 8-11 m/s demonstrate that the system works in practice. The reported mapping and path-search efficiency gains are plausible and well motivated, and the paper is generally careful in describing the algorithmic pipeline. However, the headline claims -- safe 15 m/s obstacle avoidance, a 79.24% latency reduction over existing methods, and closeness to global optimality -- are not fully supported by the evidence as presented, because the collision model is not conservative, the latency comparison is not like-for-like, and the optimality claim is not operationalized. These issues are load-bearing for the paper's central claims, so the contribution is significant but conditional on revision.

major comments (4)
  1. [Section III-D, Eqs. (9)-(10), (24)] The RESDF collision model is not conservative. Eq. (9) defines D(wxr) as the distance to the single nearest map point, and Eq. (8) labels any position with no point within radius r as Free. Because points are quantized to voxel centers by Eq. (1), D can overestimate true clearance by up to sqrt(3)r/2 at observed surfaces, and the error is unbounded wherever a surface is undersampled or outside the sensor field of view. The collision cost in Eq. (24) is zero whenever D > ds, so unobserved or gappy space produces no repulsive signal at all. The paper provides no completeness bound, no conservative inflation, and no unknown-space penalty, so the 15 m/s safety claim rests on the unstated assumption that the point cloud densely samples every obstacle surface in the planning horizon. Please add a conservative treatment or explicitly qualify the safety claim as empirical.
  2. [Table III and Abstract] The headline latency comparison is not like-for-like. The 79.24% figure in the abstract is (26.30 - 5.46)/26.30, which uses only Ours and Agile-Auto from the vlim = 15 m/s block of Table III. Agile-Auto is a learning-based end-to-end policy that consumes depth images and has no mapping, path-search, or trajectory-optimization stages (those entries are shown as '/'), so comparing total algorithm latency conflates different system architectures. Moreover, Ego-PlannerV2 and Fast-Dodge, which the text says failed at 10 and 15 m/s, are omitted from the table; showing them with failed or zero success rates would make the comparison transparent. Please revise the claim to name the specific baseline and report all baselines in the table.
  3. [Abstract and Contributions] The global-optimality claim is not operationalized. The abstract and Section I state that the planned trajectory is close to the global optimum in temporal and spatial domains, but the paper never defines an optimality metric or computes an optimality gap. Selecting the trajectory with the smallest cost J in Eq. (20) among K = 4 angular samples and a finite set of topologies only guarantees best-of-finite-set with respect to the same objective being optimized; it does not establish closeness to the true global optimum of the nonconvex problem. Please either remove the global-optimality wording or support it with a concrete optimality bound or benchmark on tractable instances.
  4. [Section V-B, Eq. (18)] Eq. (18) appears to be an incorrect feasibility condition when v0 > 0. For a constant-acceleration phase from v0 to vd followed by deceleration to zero, the required distance is (vd^2 - v0^2)/(2ad) + vd^2/(2ad) = (2vd^2 - v0^2)/(2ad), not ((vd - v0)^2 + vd^2)/(2ad). The published expression is smaller when v0 > 0, so the check in Algorithm 2 can accept an infeasible initial time allocation during in-flight replanning, where v0 is the current velocity. Please correct Eq. (18) or state explicitly that v0 is assumed to be zero.
minor comments (5)
  1. [Section VI-D] There is a typo: 'Meanwile' should be 'Meanwhile'. Also, the sentence 'The red font indicates the case of error' does not match Table III, where no red font is visible; either add the highlighting or remove the sentence.
  2. [Algorithm 1, Lines 10-16] The condition 'if ι == Default' is confusing. Please clarify what Default represents and distinguish explicitly between 'no occlusion', 'first encounter with a new obstacle', and 're-encounter with a previously labeled obstacle'.
  3. [Section III-B, Eqs. (5)-(6)] The deletion rule is described as removing points when IΔ < 0, but Eq. (6) keeps points with IΔ > 0; the sign convention and the treatment of IΔ = 0 should be stated unambiguously.
  4. [Table III] Please specify whether the reported latencies are means over successful runs only, and include standard deviations or per-run variability; this is important for the high-speed claim given the modest number of runs (20 per group).
  5. [Section VI-D and Conclusion] The limitations paragraph discusses trajectory selection but does not mention the RESDF unknown-space assumption or the lack of a formal safety guarantee; adding this limitation would make the scope of the safety claim clearer.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: HEPP's contributions are benchmarked against external systems, and the global-optimality wording is a heuristic, explicitly limited by the authors, not a result that reduces to its own inputs.

full rationale

The derivation chain is self-contained engineering rather than a circular prediction. RESDF (Eq. 9) is a nearest-neighbor distance query, and Eq. 24 is a threshold penalty using that distance; no parameter is fitted to the performance it later 'predicts.' The topological search produces candidate paths from the same occupancy model, and the final trajectory is selected as the minimum of the optimization objective (Eq. 20); this is a standard minimization over an independently generated candidate set, not a tautology, though the 'close to global optimum' claim is heuristic. The authors themselves state in Section VII: 'the selection of the best trajectory is only based on the cost value. This strategy is not optimal in all conditions because some demands are not reflected in the cost functions, such as trajectory length.' That limitation is weighed here: it weakens the optimality claim but is not circular. Mapping and planning timings are compared against external baselines (ROG-Map, FIESTA, OctoMap, Raptor, Ego-PlannerV2, Fast-Dodge, Agile-Auto); self-citations [40] and [41] are related-work/baseline references, not load-bearing. The main risk noted by reviewers, that unobserved space is treated as free and the nearest-point distance can overestimate clearance, is a safety/completeness concern, not a circularity of the derivation. Therefore no circular step is established.

Assumptions & free parameters 7 free parameters · 5 assumptions · 0 invented entities

The system's empirical claims rest on a mix of standard robotics tools (MINCO, i-Octree, FAST-LIO2) and several unverified modeling assumptions. The main hand-set quantities are K=4, map resolution r, safety distance d_s, cost weights rho/lambda, and downsampling thresholds e_min/n_max; none are tuned by a formal procedure, and sensitivity is not reported. RESDF's nearest-neighbor distance is assumed to approximate the true obstacle distance without a conservative bound. No new physical entities are introduced, and no global optimality certificate is supplied.

free parameters (7)
  • K (angle samples in topological search) = 4
    Chosen by hand to balance path diversity against search time (Section IV-B); no sensitivity analysis is provided.
  • r (minimum point spacing / resolution) = 0.1 m in deployment; 0.2 m in benchmark
    Controls map density, occupancy radius in Eq. (8), and memory; set by resolution choice rather than derived.
  • d_s (safety distance in collision cost) = not reported
    Defines how close to obstacles the optimized trajectory may pass (Eq. 24); a conservative value hides point-cloud gaps, a small value risks collision.
  • rho and lambda_star (cost weights) = not reported
    Time penalty and cost weights in Eqs. (20)-(22) are tunable; the paper does not state their values or sensitivity.
  • gamma (speed scaling factor) = not reported
    Iteratively reduces v_d in Algorithm 2 until Eq. (18) holds; its value affects initialization quality.
  • e_min, n_max (octree downsampling bounds) = not reported
    Control when points are merged in i-Octree (Section III-B3), influencing map fidelity and query time.
  • l_max (visual-plane search range) = not reported
    Upper bound on child node distance in Eq. (15); affects how far each topological detour can extend.
assumptions (5)
  • standard math MINCO spatial-temporal deformation c = M(q,T) is used without re-derivation
    Taken from [39]; the trajectory optimization rests on its linear-complexity gradient propagation (Eqs. 17, 23).
  • domain assumption Lidar-inertial odometry (FAST-LIO2 [46]) supplies accurate SE(3) robot poses
    The map sliding and raycasting in Section III-B assume w_bT_k is known and reliable.
  • domain assumption The quadrotor is differentially flat and piecewise 5-degree polynomials parameterize all feasible motions
    Section V-A; this is standard for quadrotor planning but restricts the dynamics model.
  • domain assumption A single nearest-neighbor distance in the sparse point cloud approximates the true Euclidean distance to the nearest obstacle surface
    RESDF in Eqs. (9)-(10) and the collision cost Eq. (24) rely on this; gaps in the point cloud can violate it.
  • ad hoc to paper Sampling K=4 angles on the visual plane captures sufficiently many distinct homotopy classes
    The topological search (Section IV-B) assumes J <= K new nodes cover the relevant free space; no completeness or coverage proof is given.

how reviews work

0 comments
Cite this review

Pith. "Pith review of HEPP: Hyper-efficient Perception and Planning for High-speed Obstacle Avoidance of UAVs." pith.science (2026). https://pith.science/paper/TCFJ65DQ

@misc{pith2026250517438,
  author       = {Pith},
  title        = {Pith review of: HEPP: Hyper-efficient Perception and Planning for High-speed Obstacle Avoidance of UAVs},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/TCFJ65DQ}},
  note         = {Machine review of arXiv:2505.17438}
}
read the original abstract

High-speed obstacle avoidance of uncrewed aerial vehicles (UAVs) in cluttered environments is a significant challenge. Existing UAV planning and obstacle avoidance systems can only fly at moderate speeds or at high speeds over empty or sparse fields. In this article, we propose a hyper-efficient perception and planning system for the high-speed obstacle avoidance of UAVs. The system mainly consists of three modules: 1) A novel incremental robocentric mapping method with distance and gradient information, which takes 89.5% less time compared to existing methods. 2) A novel obstacle-aware topological path search method that generates multiple distinct paths. 3) An adaptive gradient-based high-speed trajectory generation method with a novel time pre-allocation algorithm. With these innovations, the system has an excellent real-time performance with only milliseconds latency in each iteration, taking 79.24% less time than existing methods at high speeds (15 m/s in cluttered environments), allowing UAVs to fly swiftly and avoid obstacles in cluttered environments. The planned trajectory of the UAV is close to the global optimum in both temporal and spatial domains. Finally, extensive validations in both simulation and real-world experiments demonstrate the effectiveness of our proposed system for high-speed navigation in cluttered environments.

Figures

Figures reproduced from arXiv: 2505.17438 by the authors.

Figure 1
Figure 1. A demonstration of high-speed navigation, the UAV flies at over 11.0 [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. An overview of our system framework. II. RELATED WORKS A. Local Mapping and ESDF In most robotic applications, the occupancy grid map is used for navigation and obstacle collision checks due to its intuitive principle and simple implementation [3]. The process usually consists of raycasting, updating the grid state, and so on. Based on this, the Euclidean Signed Distance Field (ESDF) [4] is widely used in gradient-b… view at source ↗
Figure 3
Figure 3. The illustration of the map sliding process. The points in the old map [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗
Figures from the paper (14 more)
Figure 5
Figure 5. Figure 5: The comparison of traditional ESDF (a) and our RESDF (b). ESDF [PITH_FULL_IMAGE:figures/full_fig_p005_5.png]
Figure 6
Figure 6. Figure 6: The illustration of the node extension process. While performing the [PITH_FULL_IMAGE:figures/full_fig_p006_6.png]
Figure 8
Figure 8. Figure 8: The illustration of the adaptive trajectory initialization. The allocation [PITH_FULL_IMAGE:figures/full_fig_p007_8.png]
Figure 9
Figure 9. Figure 9: The illustration of the parallel gradient-based trajectory optimization. [PITH_FULL_IMAGE:figures/full_fig_p008_9.png]
Figure 10
Figure 10. Figure 10: Hardware platform of our system. is a Micoair-NXT board that runs the PX4 flight stack. The overall system weighs 1.31 kg. The overview of our hardware platform is shown in [PITH_FULL_IMAGE:figures/full_fig_p008_10.png]
Figure 11
Figure 11. Figure 11: The presentation of our mapping method in indoor and outdoor scenarios. (a) is the outdoor Kitti [PITH_FULL_IMAGE:figures/full_fig_p009_11.png]
Figure 12
Figure 12. Figure 12: Comparison of the topological path search. Our proposed method [PITH_FULL_IMAGE:figures/full_fig_p010_12.png]
Figure 14
Figure 14. Figure 14: More challenging high-speed navigation in an extremely dense [PITH_FULL_IMAGE:figures/full_fig_p012_14.png]
Figure 13
Figure 13. Figure 13: Trajectory benchmark comparison with the speed limitation as 5 [PITH_FULL_IMAGE:figures/full_fig_p012_13.png]
Figure 17
Figure 17. Figure 17: Then, the UAV is required to patrol in the room and [PITH_FULL_IMAGE:figures/full_fig_p012_17.png]
Figure 16
Figure 16. Figure 16: More challenging flight in a dynamic environment with red color [PITH_FULL_IMAGE:figures/full_fig_p013_16.png]
Figure 15
Figure 15. Figure 15: More challenging high-speed patrol in an extremely dense environ [PITH_FULL_IMAGE:figures/full_fig_p013_15.png]
Figure 17
Figure 17. Figure 17: The results of an indoor high-speed navigation flight. (a) is the [PITH_FULL_IMAGE:figures/full_fig_p014_17.png]
Figure 19
Figure 19. Figure 19: The results of an outdoor high-speed flight test. (a)-(d) are snapshots [PITH_FULL_IMAGE:figures/full_fig_p015_19.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

46 extracted references · 38 canonical work pages

  1. [1]

    To- wards the unmanned aerial vehicles (uavs): A comprehensive review,

    S. A. H. Mohsan, M. A. Khan, F. Noor, I. Ullah, and M. H. Alsharif, “To- wards the unmanned aerial vehicles (uavs): A comprehensive review,” Drones, vol. 6, no. 6, p. 147, 2022

    work page 2022

  2. [2]

    A survey of guidance, navi- gation, and control systems for autonomous multi-rotor small unmanned aerial systems,

    J. A. Marshall, W. Sun, and A. L’Afflitto, “A survey of guidance, navi- gation, and control systems for autonomous multi-rotor small unmanned aerial systems,” Annual Reviews in control , vol. 52, pp. 390–427, 2021

    work page 2021

  3. [3]

    Using occupancy grids for mobile robot perception and navigation,

    A. Elfes, “Using occupancy grids for mobile robot perception and navigation,” Computer, vol. 22, no. 6, pp. 46–57, 1989

    1989

  4. [4]

    Signed distance fields: A natural representation for both mapping and planning,

    H. Oleynikova, A. Millane, Z. Taylor, E. Galceran, J. Nieto, and R. Siegwart, “Signed distance fields: A natural representation for both mapping and planning,” in RSS 2016 workshop: geometry and beyond- representations, physics, and scene understanding for robotics , Univer- sity of Michigan, 2016

    work page 2016

  5. [5]

    Raptor: Robust and perception- aware trajectory replanning for quadrotor fast flight,

    B. Zhou, J. Pan, F. Gao, and S. Shen, “Raptor: Robust and perception- aware trajectory replanning for quadrotor fast flight,” IEEE Transactions on Robotics, vol. 37, no. 6, pp. 1992–2009, 2021. 13 15 t(s) 7.5 (a) (b) (c) 17.9 v(m/s) 0 0.1 Fig. 15. More challenging high-speed patrol in an extremely dense environ- ment. When the speed limitation is set as 1...

    work page 1992

  6. [6]

    Seer: Safe efficient exploration for aerial robots using learning to predict information gain,

    Y . Tao, Y . Wu, B. Li, F. Cladera, A. Zhou, D. Thakur, and V . Kumar, “Seer: Safe efficient exploration for aerial robots using learning to predict information gain,” in 2023 IEEE International Conference on Robotics and Automation (ICRA) , pp. 1235–1241, IEEE, 2023

    work page 2023

  7. [7]

    Perceptive model predictive control for con- tinuous mobile manipulation,

    J. Pankert and M. Hutter, “Perceptive model predictive control for con- tinuous mobile manipulation,” IEEE Robotics and Automation Letters , vol. 5, no. 4, pp. 6177–6184, 2020

    work page 2020

  8. [8]

    Fiesta: Fast incremental euclidean distance fields for online motion planning of aerial robots,

    L. Han, F. Gao, B. Zhou, and S. Shen, “Fiesta: Fast incremental euclidean distance fields for online motion planning of aerial robots,” in 2019 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), pp. 4423–4430, 2019

    work page 2019

Show all 46 references
  1. [9]

    V oxblox: Incremental 3d euclidean signed distance fields for on-board mav planning,

    H. Oleynikova, Z. Taylor, M. Fehr, R. Siegwart, and J. Nieto, “V oxblox: Incremental 3d euclidean signed distance fields for on-board mav planning,” in 2017 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) , pp. 1366–1373, 2017

    2017

  2. [10]

    V oxfield: Non-projective signed distance fields for online plan- ning and 3d reconstruction,

    Y . Pan, Y . Kompis, L. Bartolomei, R. Mascaro, C. Stachniss, and M. Chli, “V oxfield: Non-projective signed distance fields for online plan- ning and 3d reconstruction,” in 2022 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) , pp. 5331–5338, 2022

    2022

  3. [11]

    Fiimap: Fast incremental inflate mapping for autonomous mav navigation,

    Y . Li, L. Wang, Y . Ren, F. Chen, and W. Zhu, “Fiimap: Fast incremental inflate mapping for autonomous mav navigation,” Electronics, vol. 12, replan moving objects (a) (b) 6.0 t(s) 3.0 14.5 v(m/s) 0 (c) Fig. 16. More challenging flight in a dynamic environment with red color ...

    2023

  4. [12]

    Rog-map: An efficient robocentric occupancy grid map for large-scene and high- resolution lidar-based motion planning,

    Y . Ren, Y . Cai, F. Zhu, S. Liang, and F. Zhang, “Rog-map: An efficient robocentric occupancy grid map for large-scene and high- resolution lidar-based motion planning,” in 2024 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) , pp. 8119–8125, 2024

    2024

  5. [13]

    A universal grid map library: Implementa- tion and use case for rough terrain navigation,

    P. Fankhauser and M. Hutter, “A universal grid map library: Implementa- tion and use case for rough terrain navigation,” Robot Operating System (ROS) The Complete Reference (Volume 1) , pp. 99–120, 2016

    2016

  6. [14]

    Octomap: An efficient probabilistic 3d mapping framework based on octrees,

    A. Hornung, K. M. Wurm, M. Bennewitz, C. Stachniss, and W. Burgard, “Octomap: An efficient probabilistic 3d mapping framework based on octrees,” Autonomous robots, vol. 34, pp. 189–206, 2013

    2013

  7. [15]

    Ufomap: An efficient probabilistic 3d mapping framework that embraces the unknown,

    D. Duberg and P. Jensfelt, “Ufomap: An efficient probabilistic 3d mapping framework that embraces the unknown,” IEEE Robotics and Automation Letters, vol. 5, no. 4, pp. 6411–6418, 2020

    2020

  8. [16]

    Multi-resolution 3d mapping with explicit free space representation for fast and accurate mobile robot motion planning,

    N. Funk, J. Tarrio, S. Papatheodorou, M. Popovi ´c, P. F. Alcantarilla, and S. Leutenegger, “Multi-resolution 3d mapping with explicit free space representation for fast and accurate mobile robot motion planning,”IEEE Robotics and Automation Letters , vol. 6, no. 2, pp. 3553–3...

    2021

  9. [17]

    i-octree: A fast, lightweight, and dynamic octree for proximity search,

    J. Zhu, H. Li, Z. Wang, S. Wang, and T. Zhang, “i-octree: A fast, lightweight, and dynamic octree for proximity search,” in 2024 IEEE In- ternational Conference on Robotics and Automation (ICRA), pp. 12290– 12296, 2024

    2024

  10. [18]

    Path deformation roadmaps: Compact graphs with useful cycles for motion planning,

    L. Jaillet and T. Sim ´eon, “Path deformation roadmaps: Compact graphs with useful cycles for motion planning,” The International Journal of Robotics Research, vol. 27, no. 11-12, pp. 1175–1188, 2008

    2008

  11. [19]

    Search-based path planning with homotopy class con- straints,

    S. Bhattacharya, “Search-based path planning with homotopy class con- straints,” in Proceedings of the AAAI conference on artificial intelligence, vol. 24, pp. 1230–1237, 2010

    2010

  12. [20]

    Topological constraints in search-based robot path planning,

    S. Bhattacharya, M. Likhachev, and V . Kumar, “Topological constraints in search-based robot path planning,” Autonomous Robots , vol. 33, pp. 273–290, 2012

    2012

  13. [21]

    Prob- abilistic roadmaps for path planning in high-dimensional configuration spaces,

    L. E. Kavraki, P. Svestka, J.-C. Latombe, and M. H. Overmars, “Prob- abilistic roadmaps for path planning in high-dimensional configuration spaces,” IEEE transactions on Robotics and Automation , vol. 12, no. 4, pp. 566–580, 1996

    1996

  14. [22]

    Ctopprm: Clustering topo- logical prm for planning multiple distinct paths in 3d environments,

    M. Novosad, R. Penicka, and V . V onasek, “Ctopprm: Clustering topo- logical prm for planning multiple distinct paths in 3d environments,” 14 8.0 4.0 0 0 (c) (a) Start End (b) (d) t(s) 5.3 v(m/s) (c) Fig. 17. The results of an indoor high-speed navigation flight. (a) is the co...

    2023

  15. [23]

    An online motion planning approach of mobile robots in distinctive homotopic classes by a sparse roadmap,

    X. Zhang, B. Zhang, C. Qi, Z. Li, and H. Li, “An online motion planning approach of mobile robots in distinctive homotopic classes by a sparse roadmap,” in Intelligent Robotics and Applications: 12th International Conference, ICIRA 2019, Shenyang, China, August 8–11, 2019, Pro...

    2019

  16. [24]

    An improved probabilistic roadmap planning method for safe indoor flights of unmanned aerial vehicles,

    Q. Jin, Q. Hu, P. Zhao, S. Wang, and M. Ai, “An improved probabilistic roadmap planning method for safe indoor flights of unmanned aerial vehicles,” Drones, vol. 7, no. 2, p. 92, 2023

    2023

  17. [25]

    Visibility-based probabilis- tic roadmaps for motion planning,

    T. Sim ´eon, J.-P. Laumond, and C. Nissoux, “Visibility-based probabilis- tic roadmaps for motion planning,” Advanced Robotics, vol. 14, no. 6, pp. 477–493, 2000

    2000

  18. [26]

    Far planner: Fast, attemptable route planner using dynamic visibility update,

    F. Yang, C. Cao, H. Zhu, J. Oh, and J. Zhang, “Far planner: Fast, attemptable route planner using dynamic visibility update,” in 2022 Start End Waypoint 1 Waypoint 2 8.0 t(s) 4.0 (d) 14.5 v(m/s) 0 0 (c) (b) (a) Fig. 18. The results of an indoor high-speed patrol flight. (a) is...

    2022

  19. [27]

    Obstacle-aware topologi- cal planning over polyhedral representation for quadrotors,

    J. Gao, F. He, W. Zhang, and Y . Yao, “Obstacle-aware topologi- cal planning over polyhedral representation for quadrotors,” in 2023 IEEE International Conference on Robotics and Automation (ICRA) , pp. 10097–10103, 2023

    2023

  20. [28]

    Fht-map: Feature- based hybrid topological map for relocalization and path planning,

    K. Song, W. Liu, G. Chen, X. Xu, and Z. Xiong, “Fht-map: Feature- based hybrid topological map for relocalization and path planning,” IEEE Robotics and Automation Letters , 2024

    2024

  21. [29]

    Ego-swarm: A fully autonomous and decentralized quadrotor swarm system in cluttered environments,

    X. Zhou, J. Zhu, H. Zhou, C. Xu, and F. Gao, “Ego-swarm: A fully autonomous and decentralized quadrotor swarm system in cluttered environments,” in 2021 IEEE International Conference on Robotics and Automation (ICRA), pp. 4101–4107, 2021

    2021

  22. [30]

    Topological semantic graph memory for image-goal navigation,

    N. Kim, O. Kwon, H. Yoo, Y . Choi, J. Park, and S. Oh, “Topological semantic graph memory for image-goal navigation,” in Conference on 15 (f) UA V (a) (b) (d) (c) 12.0 t(s) 6.0 (g) 9.0 v(m/s) 0 0 (e) Fig. 19. The results of an outdoor high-speed flight test. (a)-(d) are snapsh...

    2023

  23. [31]

    Multi-layer path planning control for the simulation of manipulation tasks: Involving semantics and topology,

    S. Cailhol, P. Fillatreau, Y . Zhao, and J.-Y . Fourquet, “Multi-layer path planning control for the simulation of manipulation tasks: Involving semantics and topology,” Robotics and Computer-Integrated Manufac- turing, vol. 57, pp. 17–28, 2019

    2019

  24. [32]

    Etpnav: Evolving topological planning for vision-language navigation in continuous environments,

    D. An, H. Wang, W. Wang, Z. Wang, Y . Huang, K. He, and L. Wang, “Etpnav: Evolving topological planning for vision-language navigation in continuous environments,” IEEE Transactions on Pattern Analysis and Machine Intelligence , 2024

    2024

  25. [33]

    Minimum snap trajectory generation and control for quadrotors,

    D. Mellinger and V . Kumar, “Minimum snap trajectory generation and control for quadrotors,” in 2011 IEEE International Conference on Robotics and Automation , pp. 2520–2525, 2011

    2011

  26. [34]

    Online generation of collision-free trajectories for quadrotor flight in unknown cluttered environments,

    J. Chen, T. Liu, and S. Shen, “Online generation of collision-free trajectories for quadrotor flight in unknown cluttered environments,” in 2016 IEEE International Conference on Robotics and Automation (ICRA), pp. 1476–1483, 2016

    2016

  27. [35]

    Real-time planning of minimum-time trajectories for agile uav flight,

    K. Teissing, M. Novosad, R. Penicka, and M. Saska, “Real-time planning of minimum-time trajectories for agile uav flight,” IEEE Robotics and Automation Letters, vol. 9, no. 11, pp. 10351–10358, 2024

    2024

  28. [36]

    Real- time trajectory replanning for mavs using uniform b-splines and a 3d circular buffer,

    V . Usenko, L. V on Stumberg, A. Pangercic, and D. Cremers, “Real- time trajectory replanning for mavs using uniform b-splines and a 3d circular buffer,” in 2017 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) , pp. 215–222, IEEE, 2017

    2017

  29. [37]

    Faster: Fast and safe trajectory planner for navigation in unknown environments,

    J. Tordesillas, B. T. Lopez, M. Everett, and J. P. How, “Faster: Fast and safe trajectory planner for navigation in unknown environments,” IEEE Transactions on Robotics , vol. 38, no. 2, pp. 922–938, 2022

    2022

  30. [38]

    Generation of bezier curve-based flyable trajectories for multi-uav systems with parallel genetic algorithm,

    O. K. Sahingoz, “Generation of bezier curve-based flyable trajectories for multi-uav systems with parallel genetic algorithm,” Journal of Intelligent & Robotic Systems , vol. 74, pp. 499–511, 2014

    2014

  31. [39]

    Geometrically constrained tra- jectory optimization for multicopters,

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

    2022

  32. [40]

    Perception and avoidance of multiple small fast moving objects for quadrotors with only low-cost rgbd camera,

    M. Lu, H. Chen, and P. Lu, “Perception and avoidance of multiple small fast moving objects for quadrotors with only low-cost rgbd camera,” IEEE Robotics and Automation Letters , vol. 7, no. 4, pp. 11657–11664, 2022

    2022

  33. [41]

    Fapp: Fast and adaptive percep- tion and planning for uavs in dynamic cluttered environments,

    M. Lu, X. Fan, H. Chen, and P. Lu, “Fapp: Fast and adaptive percep- tion and planning for uavs in dynamic cluttered environments,” IEEE Transactions on Robotics , vol. 41, pp. 871–886, 2025

    2025

  34. [42]

    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,” in 2022 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) , pp. 6332–6339, 2022

    2022

  35. [43]

    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

  36. [44]

    Learning perception- aware agile flight in cluttered environments,

    Y . Song, K. Shi, R. Penicka, and D. Scaramuzza, “Learning perception- aware agile flight in cluttered environments,” in 2023 IEEE International Conference on Robotics and Automation (ICRA) , pp. 1989–1995, 2023

    2023

  37. [45]

    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

    2024 arXiv

  38. [46]

    Fast-lio2: Fast direct lidar-inertial odometry,

    W. Xu, Y . Cai, D. He, J. Lin, and F. Zhang, “Fast-lio2: Fast direct lidar-inertial odometry,” IEEE Transactions on Robotics , vol. 38, no. 4, pp. 2053–2073, 2022

    2022

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.