Pith. sign in

REVIEW 3 major objections 5 minor 1 cited by

UAV-based path planning for efficient localization of non-uniformly distributed weeds using prior knowledge: A reinforcement-learning approach

T0 review · 3 major / 5 minor · reviewed 2026-08-11 · deepseek-v4-flash

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

Pith's one-line read This paper claims that a simulation-trained DQN search policy can find non-uniformly distributed weeds with far shorter UAV flight paths than row-by-row coverage.

desk verdict Solid simulation study with a guarded real-data transfer; the headline 66% saving omits the cost of the prior-knowledge survey, so read the real-world claim as ~20-30% end-to-end. read the letter →

arxiv 2412.11717 v2 pith:ZEPIP327 submitted 2024-12-16 cs.RO

Rick van Essen , Eldert van Henten , Gert Kootstra This is my paper · ORCID
classification cs.RO
keywords deepreinforcementlearningpathplanningUAVweeddetectionobjectsearchpriorknowledgesim-to-realtransferQ-learningagriculturalrobotics
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

This paper claims that a drone can find weeds scattered in patchy clusters far faster by learning to search than by flying row-by-row over the whole field. The authors train a deep-Q-learning agent in simulation, feeding it a low-resolution prior map of likely weed locations plus a high-resolution in-flight detection map, and let it choose the next flight direction and when to land. In simulation the learned policy found most weeds in roughly a third of the steps the row-by-row path needed, and it stayed efficient under noisy detections and imprecise prior knowledge. Transferred to four real-world field datasets without retraining, the policy flew a 66% shorter path than row-by-row while finding 10% fewer weeds. The practical point is that for patchy objects like weeds, a search policy can trade a small miss rate for a large saving in flight time.

What carries the argument

The central object is the dual-map deep-Q-network state representation: a drone-centered global map (down-sampled prior knowledge of likely weed locations, already-detected weeds, and field boundaries) concatenated with a local map of the current field-of-view detection output and a scalar battery budget. The network uses two parallel convolutional feature extractors feeding fully connected layers that output Q-values for flying north, south, east, west, and, in experiments 4 and 5, a land action. This representation lets the same trained policy react to in-flight detections while steering toward clusters suggested by prior knowledge.

What would settle it

Run the transfer experiment on a field whose weed clusters are strongly elongated or row-aligned, outside the Gaussian-mixture ranges used in training, with a detector whose false positives are spatially correlated, for example concentrated in sun-glinted strips, and compare the learned policy's path length and fraction found against row-by-row; if the saving falls far below 66% or the agent lands before finding most weeds, the simulator-fidelity assumption is falsified.

Watch

Extended reading notes

Core claim

The central discovery is that a simulation-trained deep-Q-network search policy, using a drone-centric dual-map state (coarse global prior, detailed local detections, and remaining battery), learns a search strategy that exploits the spatial clustering of weeds. When weeds are drawn from Gaussian mixtures, the agent finds more than 80% of them in 73 to 94 steps on average, versus 209 steps for row-by-row; on real-world orthomosaic data it lands after 94 ± 31 steps having found 81.2% ± 10.4% of weeds, versus 276 steps and 90.9% ± 5.8% for row-by-row, a 66% shorter path at a 10% lower find rate. The paper also claims the policy is robust to detection false positives, false negatives, and positional noise, needs only coarse prior knowledge, and learns a sound landing action that terminates the search, with only 2% premature landings.

Load-bearing premise

The whole approach depends on the simulated weed fields and simulated detection errors matching real fields closely enough that a policy trained on them still makes good decisions in practice.

Editorial extensions

If this is right

  • On non-uniform weed distributions, the learned policy beats row-by-row coverage in path length for finding most weeds, but it does not guarantee complete coverage.
  • Detection errors up to moderate levels barely change performance; only a very high error level degrades results.
  • Coarse prior knowledge of moderate quality is enough; without any prior knowledge a row-by-row sweep is more efficient.
  • A learned land action gives a large path saving (74% shorter in simulation, 66% on real data) while sacrificing a modest fraction of found weeds.
  • The policy transfers to real-world imagery without retraining, suggesting the same simulation-trained policy can be deployed with a real onboard detector and flight controller.

Reading between the lines

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

  • The 66% saving is a point estimate on four real datasets; a practitioner should first check whether their field's weed cluster count, cluster spread, and detector error correlation fall inside the randomized training ranges, because the paper's own discussion says retraining may be needed otherwise.
  • The same dual-map search formulation should transfer to other patchy-object search tasks, such as disease hotspots, invasive plants, or scattered animal groups, where missing a small fraction of targets is acceptable; the learned land action is what makes the trade-off explicit.
  • A testable extension the authors did not run is giving the drone altitude actions so it can take a high-altitude overview as fresh prior knowledge, which they identify as future work and would relax the requirement for external prior maps.
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 a deep-Q-learning-based path planner for UAV weed localization. The agent receives a simulated or real detection map of the current field of view and an uncertain low-resolution prior map of the whole field, and outputs discrete flight actions (north, south, east, west, and optionally 'land'). The policy is trained entirely in simulation with Gaussian-mixture weed distributions and simulated perception and prior-knowledge errors. Five experiments evaluate the effect of weed distribution, detection-error level, prior-knowledge quality, stopping criteria, and transfer to four real-world orthomosaics. The central claim is that the learned search policy finds non-uniformly distributed weeds with a much shorter flight path than a row-by-row baseline; on real-world data the paper reports a 66% shorter path at a cost of a 10% lower percentage of found weeds.

Significance. If the claims hold, the paper makes a useful contribution to agricultural UAV search: it demonstrates a sim-to-real transfer of an RL search policy, provides an open-source simulation and training code, evaluates over 1000 episodes per condition with Welch t-tests, and honestly discusses practical limitations such as the inability to guarantee finding all weeds. The robustness analysis across detection-error levels and prior-knowledge qualities is a strength, as is the use of a learned stopping action. However, the headline real-world saving is overstated because the cost of acquiring the prior-knowledge map is omitted, and the simulation error model is not calibrated to the real detector; both issues affect the quantitative conclusions.

major comments (3)
  1. [§2.4.5, Experiment 5; §3.5; §5] The reported 66% shorter real-world path does not account for the high-altitude row-by-row flight used to generate the prior-knowledge map. Section 2.4.5 states that the prior map is created by 'a high-altitude row-by-row flight path with a field-of-view of 24x24 grid cells' over a 48x48 field; that survey requires roughly 96 m of straight-line coverage plus turn connectors. The DQN search path is 94 ± 31 steps, so the end-to-end DQN cost is about 190–214 steps, while the row-by-row baseline is 276 steps. The resulting saving is approximately 22–31%, not 66%. Please report the end-to-end comparison including prior acquisition, or explicitly state that the prior is assumed to come for free from an external source. As written, the abstract and conclusion overstate the real-world benefit.
  2. [§2.1, Table 1, Table 3] The default simulation parameters are internally inconsistent with the stated 'moderate' detection-error level. Table 1 lists rdt,fp = 0.05 and rdt,fn = 0.0001, whereas Table 3 defines the moderate level as rdt,fp = 0.0001 and rdt,fn = 0.05. The values are effectively swapped, and the false-positive rate differs by a factor of 500. Because experiments 1, 3, 4, and 5 use the default parameters, this inconsistency affects the interpretation of all results that are not part of Experiment 2. Please correct the table or clarify which parameter set is actually used.
  3. [§2.1, §2.4.5, §4.6] The simulation error model is not calibrated to the real YOLOv8-based detection system. The simulator assumes independent false positives, false negatives, and Gaussian positional offsets per detection, but Experiment 5 does not report the actual false-positive or false-negative rates of YOLOv8-nano on the four orthomosaics, nor does it examine whether real errors are spatially correlated (e.g., missed clusters or systematic localization bias). The transferability claim therefore rests on an unverified assumption that the simulated error statistics approximate the real detector. Please report the detector's confusion statistics on the real datasets and, if feasible, test the policy under spatially correlated error models.
minor comments (5)
  1. [Table 1] The parameter name 'rpn,fn' appears to be a typo for 'rpk,fn'; also the values of rdt,fp and rdt,fn in Table 1 should be cross-checked against Table 3.
  2. [Tables 5–7] The row-by-row baseline standard deviations at 100, 200, and 300 steps are inconsistent across tables (e.g., 0.37 ± 0.08 in Table 5 versus 0.37 ± 0.30 in Table 6), although the baseline is the same; please unify these values.
  3. [Figure 13] The caption for panel (a) says 'very low' but the detection-error level is 'very high'; please correct the caption.
  4. [Throughout] The term 'UA V' contains an unintended space; please standardize to 'UAV'.
  5. [§2.4.4] For the 'no new weeds for N steps' stopping criterion, the sentence 'at least 2 weeds need to be detected' is ambiguous about whether the threshold counts the total number of detected weeds or the number of weeds in the most recent detection event; please clarify.

Circularity Check

0 steps flagged · score 2.0 of 10

No significant circularity: the claimed 66% path saving is measured against an external row-by-row baseline; the only self-citation supplies datasets and detector training data, not the result.

full rationale

The paper's central results are empirical rather than derivational. The DQN policy is learned from simulated interactions using standard Q-learning (Sections 2.3 and 2.3.1), and the headline real-world result is directly measured: 'the DQN policy found 81.2 ± 10.4% of the weeds before landing in 94 ± 31 flight steps, whereas the row-by-row flight path found 90.9 ± 5.8% of the weeds in 276 ± 0 flight steps.' The baseline is an external coverage planner (Fields2Cover), not a fitted constant or a consequence of the paper's own definitions. The only self-referential element is the use of datasets and detector training data from van Essen et al. (2025) in Section 2.4.5; that citation provides raw orthomosaics and annotations, not the search policy or the speed comparison, so it is transparent and not load-bearing. The prior-knowledge map is generated by a high-altitude row-by-row survey, and whether that survey's flight cost should be included in the practical comparison is a legitimate accounting and fairness concern, but it is not circularity: the 94 ± 31 versus 276 ± 0 step comparison is measured from the executed paths, not derived from the prior map. No equation in the paper reduces a predicted quantity to a fitted parameter, to the prior-knowledge input, or to a self-citation chain. Therefore no circular step is identified; the paper is self-contained against an external baseline and the minor self-citation does not force any conclusion.

Assumptions & free parameters 5 free parameters · 6 assumptions · 0 invented entities

The paper contributes a learned policy and an evaluation, but its central claim rests on a hand-built simulation whose distribution and error parameters are not calibrated to the real detector or field data. No new physical entities are introduced.

free parameters (5)
  • Weed cluster distribution parameters (Ndist, Sigma1, Sigma2, Sigma3, Sigma4) = strong: N(3,2), Sigma1=[[5,8],[8,15]], Sigma2=[[15,0],[0,5]]; medium: N(4,1), Sigma1-Sigma4 as in Section 2.4.1
    Hand-chosen simulation domains define what 'non-uniformly distributed weeds' means; the claimed advantage over row-by-row is conditional on these distributions.
  • Detection error levels (rdt,fp, rdt,fn, Ndt,pos) = five levels from perfect to very high, Table 3
    Simulates perception errors; not measured from the real YOLOv8 detector used in Experiment 5, so the robustness claim is conditional on this error model.
  • Prior knowledge quality levels (P, rpk,fp, rpk,fn, Npk,pos) = none through perfect, Table 4
    Defines the uncertainty in prior weed maps; the real-world prior is generated by a high-altitude flight, which may not match these simulated levels.
  • Reward weights (rdt=1.0, rstep=-0.5, rnfz=-1.0, rcrash=-150.0) = rdt=1.0, rstep=-0.5, rnfz=-1.0, rcrash=-150.0
    Hand-chosen; reward design shapes the learned policy and the landing behavior, so the path-length/found-weeds trade-off depends on them.
  • Battery budget binit and step cost = binit=75, bstep=0.2
    Imposes a flight-time limit that interacts with the stopping behavior and the percentage of weeds found.
assumptions (6)
  • domain assumption The simulated field, with Gaussian-mixture weed clusters and simulated perceptual errors, is an adequate proxy for real agricultural fields.
    Used throughout training and evaluation (Section 2.1, 2.4.5); if the simulation is not faithful, the transfer result in Experiment 5 is not generalizable.
  • domain assumption Weed locations are accurately modeled by mixtures of multivariate Gaussians, with cluster counts and covariances inside the training ranges.
    Stated in Section 2.1 and relied on in Discussion 4.5; real weed patches can be more complex.
  • domain assumption Detection and prior-knowledge errors are independent and Gaussian (positional offsets).
    Modeled in Section 2.1; spatial correlations in real detector errors are not represented.
  • domain assumption The drone executes discrete 1-cell flight actions exactly and knows its position to better than grid-cell accuracy.
    Assumption (a) in the introduction and Section 4.6; not tested in a live flight.
  • standard math Deep Q-learning with the chosen architecture and hyperparameters yields a useful approximation of the optimal search policy.
    Background algorithm usage (Section 2.3); no convergence guarantee is proven for this MDP, but this is standard practice.
  • domain assumption Some prior knowledge of weed locations is available and informative.
    Experiment 3 shows that without prior knowledge the RL agent underperforms row-by-row; the method's value depends on this premise.

how reviews work

0 comments
Cite this review

Pith. "Pith review of UAV-based path planning for efficient localization of non-uniformly distributed weeds using prior knowledge: A reinforcement-learning approach." pith.science (2026). https://pith.science/paper/ZEPIP327

@misc{pith2026241211717,
  author       = {Pith},
  title        = {Pith review of: UAV-based path planning for efficient localization of non-uniformly distributed weeds using prior knowledge: A reinforcement-learning approach},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/ZEPIP327}},
  note         = {Machine review of arXiv:2412.11717}
}
read the original abstract

UAVs are becoming popular in agriculture, however, they usually use time-consuming row-by-row flight paths. This paper presents a deep-reinforcement-learning-based approach for path planning to efficiently localize weeds in agricultural fields using UAVs with minimal flight-path length. The method combines prior knowledge about the field containing uncertain, low-resolution weed locations with in-flight weed detections. The search policy was learned using deep Q-learning. We trained the agent in simulation, allowing a thorough evaluation of the weed distribution, typical errors in the perception system, prior knowledge, and different stopping criteria on the planner's performance. When weeds were non-uniformly distributed over the field, the agent found them faster than a row-by-row path, showing its capability to learn and exploit the weed distribution. Detection errors and prior knowledge quality had a minor effect on the performance, indicating that the learned search policy was robust to detection errors and did not need detailed prior knowledge. The agent also learned to terminate the search. To test the transferability of the learned policy to a real-world scenario, the planner was tested on real-world image data without further training, which showed a 66% shorter path compared to a row-by-row path at the cost of a 10% lower percentage of found weeds. Strengths and weaknesses of the planner for practical application are comprehensively discussed, and directions for further development are provided. Overall, it is concluded that the learned search policy can improve the efficiency of finding non-uniformly distributed weeds using a UAV and shows potential for use in agricultural practice.

Figures

Figures reproduced from arXiv: 2412.11717 by the authors.

Figure 1
Figure 1. High-level drone control using Reinforcement Learning (RL) in combination with a local weed map from a perception [PITH_FULL_IMAGE:figures/full_fig_p004_1.png] view at source ↗
Figure 2
Figure 2. Two examples of the simulation environment with the field-of-view indicated by the blue rectangle around the drone, [PITH_FULL_IMAGE:figures/full_fig_p006_2.png] view at source ↗
Figure 3
Figure 3. Example of the local (a) and global (b) map. Red indicates the already detected weeds, green the area outside the [PITH_FULL_IMAGE:figures/full_fig_p008_3.png] view at source ↗
Figures from the paper (15 more)
Figure 4
Figure 4. Figure 4: Training procedure for the Deep Q-Network. [PITH_FULL_IMAGE:figures/full_fig_p010_4.png]
Figure 5
Figure 5. Figure 5: Network architecture for the DQN using the global and local map and the movement budget indicating the input [PITH_FULL_IMAGE:figures/full_fig_p011_5.png]
Figure 6
Figure 6. Figure 6: Example of a field with (a) a strong distribution, (b) a medium distribution, and (c) a uniform distribution of weeds. [PITH_FULL_IMAGE:figures/full_fig_p013_6.png]
Figure 7
Figure 7. Figure 7: Example of field-of-view (a) and the corresponding simulated detection network output for a very high (b), high (c), [PITH_FULL_IMAGE:figures/full_fig_p014_7.png]
Figure 8
Figure 8. Figure 8: Example of the prior knowledge of world (a) for quality level none (no prior knowledge) (b), low (c), moderate (d), [PITH_FULL_IMAGE:figures/full_fig_p015_8.png]
Figure 9
Figure 9. Figure 9: Applying the simulation-trained Reinforcement Learning policy to real-world image data by generating an image [PITH_FULL_IMAGE:figures/full_fig_p016_9.png]
Figure 10
Figure 10. Figure 10: Effect of a strong, a medium, and a uniform weed distribution on both the number of found weeds and the path [PITH_FULL_IMAGE:figures/full_fig_p017_10.png]
Figure 11
Figure 11. Figure 11: Single flight paths of the RL agent (a-c) and the baseline row-by-row flight path (d-f) for strong (a,d), medium [PITH_FULL_IMAGE:figures/full_fig_p018_11.png]
Figure 12
Figure 12. Figure 12: Effect of the different levels of detection errors on the percentage of found weeds for the policy learned by DQN, [PITH_FULL_IMAGE:figures/full_fig_p018_12.png]
Figure 13
Figure 13. Figure 13: Single flight paths of the RL agent for detection error levels very low (a), low (b), moderate (c), high (d), and [PITH_FULL_IMAGE:figures/full_fig_p019_13.png]
Figure 14
Figure 14. Figure 14: Effect of the different prior knowledge quality levels on the percentage of found weeds for the policy learned by [PITH_FULL_IMAGE:figures/full_fig_p020_14.png]
Figure 15
Figure 15. Figure 15: Single flight paths of the RL agent for prior knowledge quality level none (no prior knowledge) (a), low (b), moderate [PITH_FULL_IMAGE:figures/full_fig_p021_15.png]
Figure 16
Figure 16. Figure 16: Histogram showing the relation between the action value after the softmax layer for the landing action and the [PITH_FULL_IMAGE:figures/full_fig_p022_16.png]
Figure 17
Figure 17. Figure 17: Percentage of found weeds and the flight-path length for the policy learned by DQN and the baseline row-by-row [PITH_FULL_IMAGE:figures/full_fig_p022_17.png]
Figure 18
Figure 18. Figure 18: Flight path on dataset 1 (a,e), 2 (b,f), 3 (c,g) and 4 (d,h) for the DQN policy (a–d) and the baseline row-by-row [PITH_FULL_IMAGE:figures/full_fig_p023_18.png]

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 1 Pith paper

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

  1. A drone that learns to efficiently find non-uniformly distributed objects in agricultural fields: from simulation to the real world

    cs.RO 2025-05 conditional novelty 4.5 of 10

    A simulated DQN drone planner transferred to real orthomosaic data shortens search flight paths by around 72-78% compared to full coverage while reducing recall by roughly 14-25%.

Reference graph

Works this paper leans on

44 extracted references · 37 canonical work pages · cited by 1 Pith paper

  1. [1]

    title Metashape Professional

    author Agisoft , year 2023 . title Metashape Professional . https://www.agisoft.com. note version: 2.0.3

    work page 2023

  2. [2]

    , author Manoni, T

    author Albani, D. , author Manoni, T. , author Arik, A. , author Nardi, D. , author Trianni, V. , year 2019 . title Field Coverage for Weed Mapping : Toward Experiments with a UAV Swarm , in: editor Compagnoni, A. , editor Casey, W. , editor Cai, Y. , editor Mishra, B. (Eds.), booktitle Bio-inspired Information and Communication Technologies , publisher S...

    work page 2019

  3. [3]

    , year 2022

    author Anul Haq, M. , year 2022 . title CNN Based Automated Weed Detection System Using UAV Imagery . journal Computer Systems Science and Engineering volume 42 , pages 837--849 . :10.32604/csse.2022.023016

    work page arXiv 2022

  4. [4]

    , author Koubaa, A

    author Azar, A.T. , author Koubaa, A. , author Ali Mohamed, N. , author Ibrahim, H.A. , author Ibrahim, Z.F. , author Kazim, M. , author Ammar, A. , author Benjdira, B. , author Khamis, A.M. , author Hameed, I.A. , author Casalino, G. , year 2021 . title Drone Deep Reinforcement Learning : A Review . journal Electronics volume 10 , pages 999 . :10.3390/el...

    work page doi:10.3390/electronics10090999 2021

  5. [5]

    , author Johnson, G.A

    author Cardina, J. , author Johnson, G.A. , author Sparrow, D.H. , year 1997 . title The Nature and Consequence of Weed Spatial Distribution . journal Weed Science volume 45 , pages 364--373 . note Publisher: [Cambridge University Press, Weed Science Society of America]

    work page 1997

  6. [6]

    , author Berger, G.S

    author Castro, G.G.R.D. , author Berger, G.S. , author Cantieri, A. , author Teixeira, M. , author Lima, J. , author Pereira, A.I. , author Pinto, M.F. , year 2023 . title Adaptive Path Planning for Fusing Rapidly Exploring Random Trees and Deep Reinforcement Learning in an Agriculture Dynamic Environment UAVs . journal Agriculture volume 13 , pages 354 ....

    work page doi:10.3390/agriculture13020354 2023

  7. [7]

    , author Catal, C

    author Chin, R. , author Catal, C. , author Kassahun, A. , year 2023 . title Plant disease detection using drones in precision agriculture . journal Precision Agriculture volume 24 , pages 1663--1682 . :10.1007/s11119-023-10014-y

    work page doi:10.1007/s11119-023-10014-y 2023

  8. [8]

    , author Anagnostopoulos, G

    author Chronis, C. , author Anagnostopoulos, G. , author Politi, E. , author Garyfallou, A. , author Varlamis, I. , author Dimitrakopoulos, G. , year 2023 . title Path planning of autonomous UAVs using reinforcement learning . journal Journal of Physics: Conference Series volume 2526 , pages 012088 . :10.1088/1742-6596/2526/1/012088

    work page doi:10.1088/1742-6596/2526/1/012088 2023

Show all 44 references
  1. [9]

    , author Forcella, F

    author Colbach, N. , author Forcella, F. , author Johnson, G.A. , year 2000 . title Spatial and temporal stability of weed populations over five years . journal Weed Science volume 48 , pages 366--377 . :10.1614/0043-1745(2000)048[0366:SATSOW]2.0.CO;2

    2000 doi

  2. [10]

    , author Chadoeuf, R

    author Dessaint, F. , author Chadoeuf, R. , author Barralis, G. , year 1991 . title Spatial Pattern Analysis of Weed Seeds in the Cultivated Soil Seed Bank . journal Journal of Applied Ecology volume 28 , pages 721--730 . :https://doi.org/10.2307/2404578

    1991 doi

  3. [11]

    , author Yoshiyasu, Y

    author Druon, R. , author Yoshiyasu, Y. , author Kanezaki, A. , author Watt, A. , year 2020 . title Visual Object Search by Learning Spatial Context . journal IEEE Robotics and Automation Letters volume 5 , pages 1279--1286 . :10.1109/LRA.2020.2967677

    2020

  4. [12]

    , author Ye, W

    author Gao, J. , author Ye, W. , author Guo, J. , author Li, Z. , year 2020 . title Deep Reinforcement Learning for Indoor Mobile Robot Path Planning . journal Sensors volume 20 , pages 5493 . :10.3390/s20195493

    2020 doi

  5. [13]

    , year 2015

    author Girshick, R. , year 2015 . title Fast R - CNN . note ArXiv:1504.08083 [cs]

    2015 arXiv

  6. [14]

    , author Haque, A

    author Gugan, G. , author Haque, A. , year 2023 . title Path Planning for Autonomous Drones : Challenges and Future Directions . journal Drones volume 7 , pages 169 . :10.3390/drones7030169

    2023 doi

  7. [15]

    , author Mokhtar, N

    author Husnain, A. , author Mokhtar, N. , author Shah, N.M. , author Dahari, M. , author Iwahashi, M. , year 2023 . title A systematic literature review (slr) on autonomous path planning of unmanned aerial vehicles . journal Drones 2023, Vol. 7, Page 118 volume 7 , pages 118 ....

    2023 doi

  8. [16]

    , author Qiu, J

    author Jocher, G. , author Qiu, J. , author Chaurasia, A. , year 2023 . title Ultralytics YOLO . https://github.com/ultralytics/ultralytics. note version: 8.3.49

    2023

  9. [17]

    , author Jian, Z

    author Liu, C. , author Jian, Z. , author Xie, M. , author Cheng, I. , year 2021 . title A Real - Time Mobile Application for Cattle Tracking using Video Captured from a Drone , in: booktitle 2021 International Symposium on Networks , Computers and Communications ( ISNCC ) , p...

    2021

  10. [18]

    , author Brito, B

    author Lodel, M. , author Brito, B. , author Serra-Gomez, A. , author Ferranti, L. , author Babuska, R. , author Alonso-Mora, J. , year 2022 . title Where to Look Next : Learning Viewpoint Recommendations for Informative Trajectory Planning , in: booktitle 2022 International C...

    2022

  11. [19]

    , author Valente, J

    author Mier, G. , author Valente, J. , author de Bruin, S. , year 2023 . title Fields2cover: An open-source coverage path planning library for unmanned agricultural vehicles . journal IEEE Robotics and Automation Letters volume 8 , pages 2166--2172 . :10.1109/LRA.2023.3248439

    2023

  12. [20]

    , author Kavukcuoglu, K

    author Mnih, V. , author Kavukcuoglu, K. , author Silver, D. , author Rusu, A.A. , author Veness, J. , author Bellemare, M.G. , author Graves, A. , author Riedmiller, M. , author Fidjeland, A.K. , author Ostrovski, G. , author Petersen, S. , author Beattie, C. , author Sadik, ...

    2015

  13. [21]

    , author Peng, B

    author Narvekar, S. , author Peng, B. , author Leonetti, M. , author Sinapov, J. , author Taylor, M.E. , author Stone, P. , year 2020 . title Curriculum Learning for Reinforcement Learning Domains : A Framework and Survey . journal J. Mach. Learn. Res. volume 21 . note Publish...

    2020

  14. [22]

    , author Zhang, K

    author Niroui, F. , author Zhang, K. , author Kashino, Z. , author Nejat, G. , year 2019 . title Deep Reinforcement Learning Robot for Search and Rescue Applications : Exploration in Unknown Cluttered Environments . journal IEEE Robotics and Automation Letters volume 4 , pages...

    2019

  15. [23]

    , author Yakovlev, K.S

    author Panov, A.I. , author Yakovlev, K.S. , author Suvorov, R. , year 2018 . title Grid Path Planning with Deep Reinforcement Learning : Preliminary Results . journal Procedia Computer Science volume 123 , pages 347--353 . :10.1016/j.procs.2018.01.054

    2018 doi

  16. [24]

    , author Sun, Y

    author Pei, H. , author Sun, Y. , author Huang, H. , author Zhang, W. , author Sheng, J. , author Zhang, Z. , year 2022 . title Weed Detection in Maize Fields by UAV Images Based on Crop Row Preprocessing and Improved YOLOv4 . journal Agriculture volume 12 , pages 975 . :10.33...

    2022 doi

  17. [25]

    , author Hitz, G

    author Popovic, M. , author Hitz, G. , author Nieto, J. , author Sa, I. , author Siegwart, R. , author Galceran, E. , year 2017 . title Online informative path planning for active classification using uavs . journal Proceedings - IEEE International Conference on Robotics and A...

    2017

  18. [26]

    , author Ott, J

    author Popović, M. , author Ott, J. , author Rückin, J. , author Kochenderfer, M.J. , year 2024 . title Learning-based methods for adaptive informative path planning . journal Robotics and Autonomous Systems volume 179 , pages 104727 . :10.1016/j.robot.2024.104727

    2024

  19. [27]

    , author Hill, A

    author Raffin, A. , author Hill, A. , author Gleave, A. , author Kanervisto, A. , author Ernestus, M. , author Dormann, N. , year 2021 . title Stable-baselines3: Reliable reinforcement learning implementations . journal Journal of Machine Learning Research volume 22 , pages 1--8

    2021

  20. [28]

    , author Zhang, Y

    author Rai, N. , author Zhang, Y. , author Ram, B.G. , author Schumacher, L. , author Yellavajjala, R.K. , author Bajwa, S. , author Sun, X. , year 2023 . title Applications of deep learning in precision weed management: A review . journal Computers and Electronics in Agricult...

    2023

  21. [29]

    , author Eesaar, H

    author Rehman, M.U. , author Eesaar, H. , author Abbas, Z. , author Seneviratne, L. , author Hussain, I. , author Chong, K.T. , year 2024 . title Advanced drone-based weed detection using feature-enriched deep learning approach . journal Knowledge-Based Systems volume 305 , pa...

    2024

  22. [30]

    , author Abdollahi, A

    author Rejeb, A. , author Abdollahi, A. , author Rejeb, K. , author Treiblmaier, H. , year 2022 . title Drones in agriculture: A review and bibliometric analysis . journal Computers and Electronics in Agriculture volume 198 , pages 107017 . :10.1016/j.compag.2022.107017

    2022

  23. [31]

    , author Bernardos, A.M

    author Rey, L. , author Bernardos, A.M. , author Dobrzycki, A.D. , author Carramiñana, D. , author Bergesio, L. , author Besada, J.A. , author Casar, J.R. , year 2025 . title A Performance Analysis of You Only Look Once Models for Deployment on Constrained Computational Edge D...

    2025 doi

  24. [32]

    , author Chamoso, P

    author Rivas, A. , author Chamoso, P. , author González-Briones, A. , author Corchado, J.M. , year 2018 . title Detection of Cattle Using Drones and Convolutional Neural Networks . journal Sensors volume 18 , pages 2048 . :10.3390/s18072048. note publisher: MDPI AG

    2018 doi

  25. [33]

    , author Van Henten, E.J

    author Ruigrok, T. , author Van Henten, E.J. , author Kootstra, G. , year 2023 . title Improved generalization of a plant-detection model for precision weed control . journal Computers and Electronics in Agriculture volume 204 , pages 107554 . :10.1016/j.compag.2022.107554

    2023

  26. [34]

    , author Barto, A.G

    author Sutton, R.S. , author Barto, A.G. , year 2018 . title Reinforcement learning: an introduction . publisher MIT press , address Cambridge, Massachusetts, USA

    2018

  27. [35]

    , author Liang, Y

    author Tang, J. , author Liang, Y. , author Li, K. , year 2024 . title Dynamic Scene Path Planning of UAVs Based on Deep Reinforcement Learning . journal Drones volume 8 , pages 60 . :10.3390/drones8020060

    2024 doi

  28. [36]

    , author Bayerlein, H

    author Theile, M. , author Bayerlein, H. , author Nai, R. , author Gesbert, D. , author Caccamo, M. , year 2020 . title UAV Coverage Path Planning under Varying Power Constraints using Deep Reinforcement Learning , in: booktitle 2020 IEEE / RSJ International Conference on Inte...

    2020

  29. [37]

    , author Bayerlein, H

    author Theile, M. , author Bayerlein, H. , author Nai, R. , author Gesbert, D. , author Caccamo, M. , year 2021 . title UAV Path Planning using Global and Local Map Information with Deep Reinforcement Learning , in: booktitle 2021 20th International Conference on Advanced Robo...

    2021

  30. [38]

    , author Juang, J.G

    author Tu, G.T. , author Juang, J.G. , year 2023 . title Uav path planning and obstacle avoidance based on reinforcement learning in 3d environments . journal Actuators 2023, Vol. 12, Page 57 volume 12 , pages 57 . :10.3390/ACT12020057

    2023 doi

  31. [39]

    , author van Henten , E

    author van Essen , R. , author van Henten , E. , author Kooistra, L. , author Kootstra, G. , year 2025 . title Adaptive path planning for efficient object search by uavs in agricultural fields . journal Smart Agricultural Technology volume 12 , pages 101075 . :https://doi.org/...

    2025

  32. [40]

    , author Rückin, J

    author Westheider, J. , author Rückin, J. , author Popović, M. , year 2023 . title Multi- UAV Adaptive Path Planning Using Deep Reinforcement Learning , in: booktitle 2023 IEEE / RSJ International Conference on Intelligent Robots and Systems ( IROS ) , publisher IEEE , address...

    2023

  33. [41]

    , author Shu, L

    author Xu, K. , author Shu, L. , author Xie, Q. , author Song, M. , author Zhu, Y. , author Cao, W. , author Ni, J. , year 2023 . title Precision weed detection in wheat fields for agriculture 4.0: A survey of enabling technologies, methods, and research challenges . journal C...

    2023

  34. [42]

    , author Wang, X

    author Yang, W. , author Wang, X. , author Farhadi, A. , author Gupta, A. , author Mottaghi, R. , year 2018 . title Visual Semantic Navigation using Scene Priors . note ArXiv:1810.06543 [cs]

    2018 arXiv

  35. [43]

    , author Su, Y

    author Yu, J. , author Su, Y. , author Liao, Y. , year 2020 . title The path planning of mobile robot by neural networks and hierarchical reinforcement learning . journal Frontiers in Neurorobotics volume 14 , pages 63 . :10.3389/FNBOT.2020.00063/BIBTEX

    2020

  36. [44]

    , author Valente, J

    author Zhang, C. , author Valente, J. , author Wang, W. , author Guo, L. , author Tubau Comas, A. , author Van Dalfsen, P. , author Rijk, B. , author Kooistra, L. , year 2023 . title Feasibility assessment of tree-level flower intensity quantification from UAV RGB imagery: A t...

    2023 doi

Pith tools

Reviewed August 11, 2026 · model on record in the stance chip above.