Pith. sign in

REVIEW 3 major objections 4 minor 58 references

Autonomous Collaborative Scheduling of Time-dependent UAVs, Workers and Vehicles for Crowdsensing in Disaster Response

T0 review · 3 major / 4 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 The paper claims that a training-free, local-information scheduler for UAVs, workers, and vehicles raises post-disaster crowdsensing completion by 14–64% over four baselines while keeping each decision under 10 seconds.

desk verdict Plausible distributed heuristic, broken local-Nash proof, and broad but under-reported experiments—salvageable with a serious revision. read the letter →

arxiv 2506.04276 v1 pith:ZZGXKDDP submitted 2025-06-04 cs.MA cs.AI

Lei Han , Yitong Guo , Pengfei Yang , Zhiyong Yu , Liang Wang , Quan Wang , Zhiwen Yu This is my paper
classification cs.MAcs.AI
keywords mobilecrowdsensingmulti-agentautonomousschedulingUAV-worker-vehiclecollaborationlocalNashequilibriumgamedimensionalityreductioninmatchingsoftmaxactionselectiondisasterresponsetime-dependentagents
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

Disaster response needs fast, reliable sensing information, but post-disaster conditions break the usual assumptions: communication is patchy, no historical data exists for offline training, and UAVs, workers, and vehicles can cooperate only when their paths actually meet. HoAs-PALN is an online scheduler in which every agent chooses its next move from what it can sense locally, with no central controller and no trained model. It cuts the hard five-way matching problem into two three-way matchings—task execution versus recharging—and stabilizes decisions through a local Nash check, meaning agents resample their softmax-chosen actions until no one nearby can improve its reward by acting alone. Across real-world bike, taxi, and ride-hailing traces plus simulated grids, the paper reports average task-completion gains of 64.12%, 46.48%, 16.55%, and 14.03% over GREEDY, K-WTA, MADL, and MARL, with each online decision under 10 seconds.

What carries the argument

The machinery is the pairing of a dimensionality-reduction rule with a local Nash-equilibrium check. The matching problem is a five-dimensional assignment over UAVs, workers, vehicles, task points, and charge points; the paper replaces it with two three-dimensional assignments, 'UAV-worker-task point' and 'UAV-vehicle-charge point', ranked by expected UAV benefit. The scheduling loop then uses softmax over distances to turn positions into action probabilities, defines local rewards as the number of matched tasks or the power added to UAVs, and iteratively resamples until every agent's reward indicator is 1, meaning no unilateral change within its communication range improves its reward. The load-bearing theoretical object is Lemma 3's claim that the gap between global task completion and the sum of local rewards is linear in the coupling strength ε, from which the paper derives an approximation factor α approaching 1 as ε tends to 0.

What would settle it

Run HoAs-PALN on the same dataset with several communication radii, compute the global completion rate and the sum of local rewards, and plot their difference against the coupling strength; if the plot does not follow a straight line through zero, the approximation factor in Eq. (31) is not a well-defined function of ε and Lemma 3's guarantee fails.

Watch

Extended reading notes

Core claim

The paper's central claim is that autonomous, time-dependent collaboration among heterogeneous sensing agents can be scheduled effectively in real time without historical data or global information. To make this tractable, HoAs-PALN decomposes the five-dimensional matching among UAVs, workers, vehicles, task points, and charge points into two three-dimensional matching problems, using each UAV's estimated expected benefit of executing a task versus being charged. It then models the multi-agent interaction as a local Nash game: each agent evaluates its reward within its communication range, checks whether any neighbor can improve by changing action, and resamples actions via softmax distance-to-probability mapping until no one can improve locally. The paper further claims that when the coupling strength between communication coverages is low, the local Nash solution approximates the global Nash solution, and that this approximation factor is driven by the ratio of coupling strength to task completion.

Load-bearing premise

The load-bearing premise is that the gap between what the whole system accomplishes and what the agents count locally is exactly a constant multiple of the coupling strength, so that when communication ranges barely overlap the local Nash solution is provably close to a global one.

Editorial extensions

If this is right

  • Post-disaster sensing can be deployed without pre-trained models or a functioning central server, because HoAs-PALN only needs each agent's current local information.
  • Repeated online decisions stay within a 10-second budget on the tested datasets, fast enough for minute-scale rescue planning.
  • The task/charge decomposition keeps decision cost low even as the number of agents or tasks grows, because the five-dimensional assignment never has to be solved directly.
  • Lower communication coupling strength yields solutions closer to the global Nash equilibrium, so widely dispersed agents can be treated as nearly decoupled.
  • Softmax-based probabilistic action selection beats pure random selection under the same game structure, giving higher completion rates, shorter decision times, and shorter movement distances.

Reading between the lines

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

  • The same reduction-and-local-game recipe could transfer to other heterogeneous fleets (ground robots, boats, mixed sensors) whenever scheduling splits into a service-matching half and an energy-matching half.
  • If the linear-gap premise in Eq. (28) fails—say the gap grows quadratically with coupling strength—the Lemma 3 guarantee would no longer hold, and the local Nash mechanism would rest on empirical performance alone.
  • A testable extension is to run HoAs-PALN with deliberately varied communication radii while computing the true global optimum; the achieved-to-optimal ratio should track 1 - C·ε/Cplt for a fixed C if the lemma's premise holds.
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

3 major / 4 minor

Summary. The paper proposes HoAs-PALN, an online distributed scheduling algorithm for heterogeneous UAVs, workers, and vehicles collecting environmental information in post-disaster scenarios. The method decomposes a five-dimensional matching problem into two three-dimensional matching problems, uses softmax-based probabilistic action selection, and iteratively checks a local Nash equilibrium condition within each agent's communication range. The authors claim NP-hardness of the underlying problem, prove (in Lemma 3) that local Nash solutions approximate global Nash solutions when communication coverage coupling is weak, and report experiments on real and simulated datasets showing task-completion improvements over GREEDY, K-WTA, MADL, and MARL, with per-decision times under 10 seconds. The paper also provides a GitHub link for code and data.

Significance. If the theoretical claims were valid, the paper would make a useful contribution to decentralized disaster-response crowdsensing: an untrained, online, communication-limited scheduling algorithm with a performance guarantee and reproducible experiments. The reported empirical gains are substantial and the public code base is a positive feature. However, the two central theoretical pillars are not sound. The NP-hardness proof mischaracterizes the simplified problem, which is polynomial-time maximum bipartite matching. The local Nash approximation lemma assumes the very linear gap relation it needs to prove, making the claimed guarantee vacuous. These errors undermine the advertised mechanism, not just peripheral details, and the experimental section cannot compensate for the absence of a valid theoretical grounding. The empirical observations may still be of interest, but the paper's core claims as stated are unsupported.

major comments (3)
  1. [Section 3, Lemma 1] The reduction in Lemma 1 is incorrect: under assumptions (1)-(3) and instant completion, Problem 2 is maximum bipartite matching between UAVs and task points, which is solvable in polynomial time, so it cannot be NP-hard. The cited 'optimal subset selection' framework ([39]) does not apply to this matching problem, and the proof therefore does not establish NP-hardness of Problem 1.
  2. [Section 4.2(3), Eq. (28)] The proof of Lemma 3 postulates Cplt_TasksPer - sum(Reward) = C*epsilon, but for any fixed configuration with epsilon > 0 one can always choose C as the ratio of the gap to epsilon, so Eq. (28) is vacuous unless C is bounded independently of the configuration. No such bound is given, and the statement 'as epsilon -> 0, alpha -> 1' does not follow because C may depend on epsilon; consequently the approximation-factor claim is not proven.
  3. [Section 4.2(3) and Table 3] Even if an approximation factor alpha were established, it would only relate the sum of local rewards to per-interval task completion; it would not show that the fixed point of Algorithm 1's local Nash iteration is close to a global Nash equilibrium or to the global optimum. Table 3 reports 'Global nash equilibrium task completion rate' values, but the manuscript provides no definition or computation method for this quantity, so the comparison in Section 5.5(2) cannot be reproduced.
minor comments (4)
  1. [Section 4.2(3), Eq. (28)] There are formatting and notation inconsistencies in Eq. (28) and the surrounding text (e.g., 'Cplt_T asksPer' and 'Cplt_TasksPer' are used interchangeably), which should be cleaned up.
  2. [Section 5.5(2), Table 3] Please clarify how the 'Global nash equilibrium task completion rate' is computed, or remove the column if it is not independently verifiable; without a method, the claimed convergence to the global Nash solution cannot be assessed.
  3. [Section 5.4] The experimental results report percentage-point improvements without error bars, confidence intervals, or significance tests; for an algorithm with iterative random resampling, run-to-run variability should be quantified.
  4. [Section 4.1(3), Lemma 2] The proof invokes the 'monotonicity property of the product of function' [40] without establishing the required monotonicity conditions for the multiplied factors; this is a gap in the proof of reasonableness of the expected-benefit comparison.

Circularity Check

1 steps flagged · score 7.0 of 10

Lemma 3's local-Nash-to-global guarantee is assumed in Eq. (28), making the central theoretical claim reduce to its own input; the empirical evaluation remains independent.

  1. self definitional [Section 4.2(3), proof of Lemma 3, Eqs. (27)-(31)]
    "Assume that the influence among the multi-agent behavioral strategies is governed by a coupling strength ε. That is, there exists a constant C such that Cplt TasksPer − Σ Reward_task^ijk = C·ε (28) ... We define the approximation factor as α = 1 − C·ε / Cplt TasksPer (31) ... As ε→0, it follows that α→1, which implies that Σ Reward_task^ijk can closely approximate Cplt TasksPer. This completes the proof of Lemma 3."

    Eq. (28) asserts exactly the linear gap relation the lemma is supposed to establish. Because C is an unconstrained existential, any configuration with ε>0 satisfies Eq. (28) by taking C = (Cplt TasksPer − ΣReward)/ε, so the equation carries no independent content. The proof never bounds C or shows that C·ε → 0 as ε→0; the subsequent definition of α and the claim α→1 merely restate the assumed gap. Thus the approximation conclusion is an input, not a derivation. The proof also never defines the global Nash solution it claims to approximate, so the lemma's advertised content is assumed rather than proved.

full rationale

The only step that reduces to its own input is Lemma 3. It is load-bearing because the paper's abstract and Section 4.2 use the local Nash equilibrium determination mechanism to 'ensure scheduling decision performance,' and Lemma 3 is the sole formal bridge from Algorithm 1's local stopping rule to global performance. That bridge is Eq. (28), which postulates the very linear gap that the lemma then 'proves' in Eqs. (29)-(31); with C unconstrained, the result is vacuous. This is a genuine self-definitional circularity in the central theoretical claim. Other weaknesses are correctness gaps rather than circularities: Lemma 2 invokes 'the monotonicity property of the product of function [40]' without establishing the needed monotonicity; Lemma 1 mislabels a one-to-one matching problem as 'optimal subset selection' while importing NP-hardness from [39]; and Table 3 reports 'Global nash equilibrium task completion rate' values without stating any method for computing a global Nash solution. The empirical comparisons against GREEDY, K-WTA, MADL, and MARL are self-contained, the datasets and code are released, and the self-citation [10] is used only for baseline implementation details, not as evidence for the central claim. Those parts are not circular and prevent the score from reaching 8-10.

Assumptions & free parameters 2 free parameters · 3 assumptions · 0 invented entities

The algorithm itself has no learned or fitted parameters, but the theoretical support relies on an unspecified constant C, an implicit softmax temperature, and a proxy objective that is not proven equivalent to the stated goal. No new physical entities are introduced.

free parameters (2)
  • C in Lemma 3 = undefined
    Introduced in Eq. (28) as the constant relating the global-local gap to the coupling strength. It is not derived, estimated, or justified, and it carries the entire weight of the approximation lemma.
  • Softmax inverse temperature = 1 (implicit)
    Eq. (21) uses score = -distance with no temperature parameter. The implicit temperature of 1 controls the exploration-exploitation balance and is chosen without justification or ablation.
assumptions (3)
  • ad hoc to paper Per-interval task completion is a valid proxy for the global LimitTime objective.
    Section 3 states that the objective can be converted from Cplt_Tasks to Cplt_TasksPer, but no equivalence or approximation bound is provided.
  • ad hoc to paper There exists a constant C such that Cplt_TasksPer - sum(Reward) = C*epsilon.
    Eq. (28) in Section 4.2(3) postulates the exact linear gap that Lemma 3 then uses to define the approximation factor. This is the central circular step.
  • ad hoc to paper Product monotonicity applies to the correlation argument in Lemma 2.
    The proof of Lemma 2 claims both factors of R_task and R_charge are correlated with the objective and then applies a monotonicity property of products. The required same-direction monotonicity is never established, and the claim is heuristic rather than derived.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Autonomous Collaborative Scheduling of Time-dependent UAVs, Workers and Vehicles for Crowdsensing in Disaster Response." pith.science (2026). https://pith.science/paper/ZZGXKDDP

@misc{pith2026250604276,
  author       = {Pith},
  title        = {Pith review of: Autonomous Collaborative Scheduling of Time-dependent UAVs, Workers and Vehicles for Crowdsensing in Disaster Response},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/ZZGXKDDP}},
  note         = {Machine review of arXiv:2506.04276}
}
read the original abstract

Natural disasters have caused significant losses to human society, and the timely and efficient acquisition of post-disaster environmental information is crucial for the effective implementation of rescue operations. Due to the complexity of post-disaster environments, existing sensing technologies face challenges such as weak environmental adaptability, insufficient specialized sensing capabilities, and limited practicality of sensing solutions. This paper explores the heterogeneous multi-agent online autonomous collaborative scheduling algorithm HoAs-PALN, aimed at achieving efficient collection of post-disaster environmental information. HoAs-PALN is realized through adaptive dimensionality reduction in the matching process and local Nash equilibrium game, facilitating autonomous collaboration among time-dependent UAVs, workers and vehicles to enhance sensing scheduling. (1) In terms of adaptive dimensionality reduction during the matching process, HoAs-PALN significantly reduces scheduling decision time by transforming a five-dimensional matching process into two categories of three-dimensional matching processes; (2) Regarding the local Nash equilibrium game, HoAs-PALN combines the softmax function to optimize behavior selection probabilities and introduces a local Nash equilibrium determination mechanism to ensure scheduling decision performance. Finally, we conducted detailed experiments based on extensive real-world and simulated data. Compared with the baselines (GREEDY, K-WTA, MADL and MARL), HoAs-PALN improves task completion rates by 64.12%, 46.48%, 16.55%, and 14.03% on average, respectively, while each online scheduling decision takes less than 10 seconds, demonstrating its effectiveness in dynamic post-disaster environments.

Figures

Figures reproduced from arXiv: 2506.04276 by the authors.

Figure 1
Figure 1. An example of time-dependent autonomous collabo [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Framework of HoAs-PALN [PITH_FULL_IMAGE:figures/full_fig_p005_2.png] view at source ↗
Figure 3
Figure 3. Schematic diagram of dimensionality reduction in the [PITH_FULL_IMAGE:figures/full_fig_p006_3.png] view at source ↗
Figures from the paper (9 more)
Figure 4
Figure 4. Figure 4: Probabilistic behavior selection of agents. [PITH_FULL_IMAGE:figures/full_fig_p008_4.png]
Figure 5
Figure 5. Figure 5: Schematic diagram of sensing task matching process. [PITH_FULL_IMAGE:figures/full_fig_p009_5.png]
Figure 6
Figure 6. Figure 6: The distribution of task points and charges for the four types of datasets (each grid cell is 1 km * 1 km). [PITH_FULL_IMAGE:figures/full_fig_p010_6.png]
Figure 7
Figure 7. Figure 7: Experimental results with different Interval. improvements of 65.25%, 55.83%, and 59.83% over GREEDY; 46.05%, 47.97%, and 48.65% over K-WTA; 19.79%, 21.08%, and 18.77% over MADL; and 18.45%, 18.44%, and 17.58% over MARL, respectively. These results demonstrate that, al…
Figure 9
Figure 9. Figure 9: Experimental results with different area number, [PITH_FULL_IMAGE:figures/full_fig_p013_9.png]
Figure 11
Figure 11. Figure 11: Experimental results with different task cost and [PITH_FULL_IMAGE:figures/full_fig_p013_11.png]
Figure 10
Figure 10. Figure 10: Experimental results with different agents number. [PITH_FULL_IMAGE:figures/full_fig_p013_10.png]
Figure 13
Figure 13. Figure 13: Multi-agent coordination in 3D urban disaster sce [PITH_FULL_IMAGE:figures/full_fig_p014_13.png]
Figure 12
Figure 12. Figure 12: Experimental results with different Radius. As Radius increases from 6 to 10, HoAs-PALN consis￾tently outperforms all baselines, see [PITH_FULL_IMAGE:figures/full_fig_p014_12.png]

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

58 extracted references · 57 canonical work pages

  1. [39]

    An analysis of approximations for maximizing submodular set functions—i,

    G. L. Nemhauser, L. A. Wolsey, and M. L. Fisher, “An analysis of approximations for maximizing submodular set functions—i,” Mathematical programming, vol. 14, pp. 265–294, 1978

    work page 1978

  2. [1]

    2023 global natural disaster assessment report,

    Global Disaster Data Platform, “2023 global natural disaster assessment report,” 2023, accessed: 2025-04-17. [Online]. Available: https://www.gddat.cn/newGlobalWeb

    work page 2023

  3. [2]

    2023 global natural disaster assessment report,

    United Nations Office for Disaster Risk Reduction (UNDRR), “2023 global natural disaster assessment report,” 2023, accessed: 2025-04-17. [Online]. Available: https://www.preventionweb.net/ publication/2023-global-natural-disaster-assessment-report

    work page 2023

  4. [3]

    Earthquake,

    W. contributors, “Earthquake,” 2023, accessed: 2025-04-17. [Online]. Available: https://en.wikipedia.org/wiki/Earthquake

    work page 2023

  5. [4]

    Applica- tions of wireless sensor networks: an up-to-date survey,

    D. Kandris, C. Nakas, D. Vomvas, and G. Koulouras, “Applica- tions of wireless sensor networks: an up-to-date survey,”Applied system innovation, vol. 3, no. 1, p. 14, 2020

    work page 2020

  6. [5]

    Applications of wireless sensor networks and internet of things frameworks in the industry rev- olution 4.0: A systematic literature review,

    M. Majid, S. Habib, A. R. Javed, M. Rizwan, G. Srivastava, T. R. Gadekallu, and J. C.-W. Lin, “Applications of wireless sensor networks and internet of things frameworks in the industry rev- olution 4.0: A systematic literature review,”Sensors, vol. 22, no. 6, p. 2087, 2022

    work page 2022

  7. [6]

    Collective interactive sensing,

    H. Ma and Y. Liu, “Collective interactive sensing,”Communications of the China Computer Federation, vol. 19, no. 8, 2023, (In Chinese)

    work page 2023

  8. [7]

    Crowdsensing 2.0,

    Z. Yu, H. Ma, B. Guo, and Z. Yang, “Crowdsensing 2.0,”Commu- nications of the ACM, vol. 64, no. 11, pp. 76–80, 2021

    work page 2021

Show all 58 references
  1. [8]

    Energy-efficient 3d vehicular crowdsourcing for disaster response by distributed deep reinforcement learning,

    H. Wang, C. H. Liu, Z. Dai, J. Tang, and G. Wang, “Energy-efficient 3d vehicular crowdsourcing for disaster response by distributed deep reinforcement learning,” inProceedings of the 27th ACM SIGKDD Conference on Knowledge Discovery & Data Mining, 2021, pp. 3679–3687

    2021

  2. [9]

    Aoi-minimal uav crowdsensing by model-based graph convo- lutional reinforcement learning,

    Z. Dai, C. H. Liu, Y. Ye, R. Han, Y. Yuan, G. Wang, and J. Tang, “Aoi-minimal uav crowdsensing by model-based graph convo- lutional reinforcement learning,” inIEEE INFOCOm 2022-IEEE Conference on Computer Communications. IEEE, 2022, pp. 1029– 1038

    2022

  3. [10]

    Collaborative route planning of uavs, workers, and cars for crowdsensing in disaster response,

    L. Han, C. Tu, Z. Yu, Z. Yu, W. Shan, L. Wang, and B. Guo, “Collaborative route planning of uavs, workers, and cars for crowdsensing in disaster response,”IEEE/ACM Transactions on Networking, 2024

    2024

  4. [11]

    (2025) Autonomous-collaborative- scheduling-of-time-dependent-agent

    CocaColaZero. (2025) Autonomous-collaborative- scheduling-of-time-dependent-agent. GitHub. [On- line]. Available: https://github.com/CocaColaZero/ Autonomous-Collaborative-Scheduling-of-Time-dependent-Agent

    2025

  5. [12]

    Cooperative path planning of uavs & ugvs for a persistent surveillance task in urban environments,

    Y. Wu, S. Wu, and X. Hu, “Cooperative path planning of uavs & ugvs for a persistent surveillance task in urban environments,” IEEE Internet of Things Journal, vol. 8, no. 6, pp. 4906–4919, 2020

    2020

  6. [13]

    Real-time global action planning for unmanned ground vehicle exploration in three-dimensional spaces,

    X. Zuo, J. Zhou, F. Yang, F. Su, H. Zhu, and L. Li, “Real-time global action planning for unmanned ground vehicle exploration in three-dimensional spaces,”Expert Systems with Applications, vol. 215, p. 119264, 2023

    2023

  7. [14]

    A coordinated behavior planning and trajectory plan- ning framework for multi-ugvs in unstructured narrow interaction scenarios,

    Z. Zang, X. Zhang, J. Song, Y. Lu, Z. Li, H. Dong, Y. Li, Z. Ju, and J. Gong, “A coordinated behavior planning and trajectory plan- ning framework for multi-ugvs in unstructured narrow interaction scenarios,”IEEE Transactions on Intelligent Vehicles, 2024

    2024

  8. [15]

    Radiation search operations using scene understanding with autonomous uav and ugv,

    G. Christie, A. Shoemaker, K. Kochersberger, P . Tokekar, L. McLean, and A. Leonessa, “Radiation search operations using scene understanding with autonomous uav and ugv,”Journal of Field Robotics, vol. 34, no. 8, pp. 1450–1468, 2017

    2017

  9. [16]

    Collabo- rative mapping of an earthquake damaged building via ground and aerial robots,

    N. Michael, S. Shen, K. Mohta, V . Kumar, K. Nagatani, Y. Okada, S. Kiribayashi, K. Otake, K. Yoshida, K. Ohnoet al., “Collabo- rative mapping of an earthquake damaged building via ground and aerial robots,” inField and service robotics: results of the 8th international confer...

    2014

  10. [17]

    A marsupial robotic system for surveying and inspection of freshwater ecosystems,

    M. Kalaitzakis, B. Cain, N. Vitzilaios, I. Rekleitis, and J. Moulton, “A marsupial robotic system for surveying and inspection of freshwater ecosystems,”Journal of Field Robotics, vol. 38, no. 1, pp. 121–138, 2021

    2021

  11. [18]

    When crowdsourcing meets unmanned vehicles: Toward cost-effective collaborative ur- ban sensing via deep reinforcement learning,

    L. Ding, D. Zhao, M. Cao, and H. Ma, “When crowdsourcing meets unmanned vehicles: Toward cost-effective collaborative ur- ban sensing via deep reinforcement learning,”IEEE Internet of Things Journal, vol. 8, no. 15, pp. 12 150–12 162, 2021

    2021

  12. [19]

    Evolutionary collaborative human-uav search for escaped crimi- nals,

    Y.-J. Zheng, Y.-C. Du, H.-F. Ling, W.-G. Sheng, and S.-Y. Chen, “Evolutionary collaborative human-uav search for escaped crimi- nals,”IEEE Transactions on Evolutionary Computation, vol. 24, no. 2, pp. 217–231, 2019

    2019

  13. [20]

    Vision-based target detection and localization via a team of cooperative uav and ugvs,

    S. Minaeian, J. Liu, and Y.-J. Son, “Vision-based target detection and localization via a team of cooperative uav and ugvs,”IEEE Transactions on systems, man, and cybernetics: systems, vol. 46, no. 7, pp. 1005–1016, 2015

    2015

  14. [21]

    Task scheduling of the collaborative aerial–ground system for the search and capture of multiple targets,

    Y. Wu, “Task scheduling of the collaborative aerial–ground system for the search and capture of multiple targets,”Knowledge-Based Systems, vol. 250, p. 109031, 2022

    2022

  15. [22]

    S. J. Russell and P . Norvig,Artificial intelligence: a modern approach. pearson, 2016

    2016

  16. [23]

    G. H. Plocher,The German Air Force Versus Russia, 1943. Pickle Partners Publishing, 2017

    1943

  17. [24]

    Mobile wireless rechargeable uav networks: Challenges and solutions,

    Y. Wang, Z. Su, N. Zhang, and R. Li, “Mobile wireless rechargeable uav networks: Challenges and solutions,”IEEE Communications Magazine, vol. 60, no. 3, pp. 33–39, 2022

    2022

  18. [25]

    Air-ground spatial crowdsourcing with uav carriers by geometric graph convolutional multi-agent deep reinforcement learning,

    Y. Wang, J. Wu, X. Hua, C. H. Liu, G. Li, J. Zhao, Y. Yuan, and G. Wang, “Air-ground spatial crowdsourcing with uav carriers by geometric graph convolutional multi-agent deep reinforcement learning,” in2023 IEEE 39th International Conference on Data Engi- neering (ICDE). IEEE,...

    2023

  19. [26]

    Col- laborative intelligent delivery with one truck and multiple het- erogeneous drones in covid-19 pandemic environment,

    Y. Luo, X. Deng, W. Zhang, Y. Ke, S. Wan, and Y. Qian, “Col- laborative intelligent delivery with one truck and multiple het- erogeneous drones in covid-19 pandemic environment,”IEEE Transactions on Intelligent Transportation Systems, vol. 25, no. 7, pp. 7907–7920, 2024

    2024

  20. [27]

    Energy-efficient 3-d data collection formulti-uav assisted mobile crowdsensing,

    L. Fu, Z. Zhao, G. Min, W. Miao, L. Zhao, and W. Huang, “Energy-efficient 3-d data collection formulti-uav assisted mobile crowdsensing,”IEEE Transactions on Computers, vol. 72, no. 7, pp. 2025–2038, 2022

    2025

  21. [28]

    Deep reinforce- ment learning robot for search and rescue applications: Explo- ration in unknown cluttered environments,

    F. Niroui, K. Zhang, Z. Kashino, and G. Nejat, “Deep reinforce- ment learning robot for search and rescue applications: Explo- ration in unknown cluttered environments,”IEEE Robotics and Automation Letters, vol. 4, no. 2, pp. 610–617, 2019

    2019

  22. [29]

    Cooperative aerial-ground multi-robot system for automated construction tasks,

    M. Krizmancic, B. Arbanas, T. Petrovic, F. Petric, and S. Bogdan, “Cooperative aerial-ground multi-robot system for automated construction tasks,”IEEE Robotics and Automation Letters, vol. 5, no. 2, pp. 798–805, 2020

    2020

  23. [30]

    Evolutionary human-uav cooperation for transmission network restoration,

    Y.-J. Zheng, Y.-C. Du, Z.-L. Su, H.-F. Ling, M.-X. Zhang, and S.- Y. Chen, “Evolutionary human-uav cooperation for transmission network restoration,”IEEE Transactions on Industrial Informatics, vol. 17, no. 3, pp. 1648–1657, 2020

    2020

  24. [31]

    An iterated greedy heuristic for collaborative human-uav search of missing tourists,

    Y. Xu, X. Li, X. Meng, and W. Zhang, “An iterated greedy heuristic for collaborative human-uav search of missing tourists,” Knowledge-Based Systems, vol. 286, p. 111409, 2024

    2024

  25. [32]

    Decentralized dynamic task allocation in swarm robotic systems for disaster 17 response,

    P . Ghassemi, D. DePauw, and S. Chowdhury, “Decentralized dynamic task allocation in swarm robotic systems for disaster 17 response,” in2019 international symposium on multi-robot and multi- agent systems (mrs). IEEE, 2019, pp. 83–85

    2019

  26. [33]

    Collaboration helps camera overtake lidar in 3d detection,

    Y. Hu, Y. Lu, R. Xu, W. Xie, S. Chen, and Y. Wang, “Collaboration helps camera overtake lidar in 3d detection,” inProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2023, pp. 9243–9252

    2023

  27. [34]

    Multi-uavs collaborative tracking of moving target with maximized visibility in urban environment,

    J. Wang, Y. X. Wu, Y.-Q. Chen, and S. Ju, “Multi-uavs collaborative tracking of moving target with maximized visibility in urban environment,”Journal of the Franklin Institute, vol. 359, no. 11, pp. 5512–5532, 2022

    2022

  28. [35]

    How2comm: Communication-efficient and collaboration-pragmatic multi-agent perception,

    D. Yang, K. Yang, Y. Wang, J. Liu, Z. Xu, R. Yin, P . Zhai, and L. Zhang, “How2comm: Communication-efficient and collaboration-pragmatic multi-agent perception,”Advances in Neu- ral Information Processing Systems, vol. 36, pp. 25 151–25 164, 2023

    2023

  29. [36]

    Using physiological measurements to analyze the tactical decisions in human swarm teams,

    H. Manjunatha, J. P . Distefano, A. Jani, P . Ghassemi, S. Chowdhury, K. Dantu, D. Doermann, and E. T. Esfahani, “Using physiological measurements to analyze the tactical decisions in human swarm teams,” in2020 IEEE International Conference on Systems, Man, and Cybernetics (SM...

    2020

  30. [37]

    Stronger together: Air-ground robotic collaboration using seman- tics,

    I. D. Miller, F. Cladera, T. Smith, C. J. Taylor, and V . Kumar, “Stronger together: Air-ground robotic collaboration using seman- tics,”IEEE Robotics and Automation Letters, vol. 7, no. 4, pp. 9643– 9650, 2022

    2022

  31. [38]

    Reinforcement learning for synchronization of heterogeneous multiagent systems by improvedq-functions,

    J. Li, L. Yuan, W. Cheng, T. Chai, and F. L. Lewis, “Reinforcement learning for synchronization of heterogeneous multiagent systems by improvedq-functions,”IEEE Transactions on Cybernetics, 2024

    2024

  32. [40]

    Convex optimization,

    S. Boyd, “Convex optimization,”Cambridge UP, 2004

    2004

  33. [41]

    Traffic prediction in a bike-sharing system,

    Y. Li, Y. Zheng, H. Zhang, and L. Chen, “Traffic prediction in a bike-sharing system,” inProceedings of the 23rd SIGSP ATIAL international conference on advances in geographic information systems, 2015, pp. 1–10

    2015

  34. [42]

    Urban computing: concepts, methodologies, and applications,

    Y. Zheng, L. Capra, O. Wolfson, and H. Yang, “Urban computing: concepts, methodologies, and applications,”ACM Transactions on Intelligent Systems and Technology (TIST), vol. 5, no. 3, pp. 1–55, 2014

    2014

  35. [43]

    Object tracking by the least spatiotemporal searches,

    Z. Yu, L. Han, C. Chen, W. Guo, and Z. Yu, “Object tracking by the least spatiotemporal searches,”IEEE Internet of Things Journal, vol. 8, no. 16, pp. 12 934–12 946, 2021

    2021

  36. [44]

    Gaia open dataset,

    DiDi Chuxing, “Gaia open dataset,” 2023, accessed: 2025-04-17. [Online]. Available: https://gaia.didichuxing.com

    2023

  37. [45]

    Three- dimensional stable matching problem for spatial crowdsourcing platforms,

    B. Li, Y. Cheng, Y. Yuan, G. Wang, and L. Chen, “Three- dimensional stable matching problem for spatial crowdsourcing platforms,” inProceedings of the 25th ACM SIGKDD International Conference on Knowledge Discovery & Data Mining, 2019, pp. 1643– 1653

    2019

  38. [46]

    Task scheduling in three-dimensional spatial crowdsourcing: A social welfare perspective,

    L. Wang, D. Yang, Z. Yu, F. Xiong, L. Han, S. Pan, and B. Guo, “Task scheduling in three-dimensional spatial crowdsourcing: A social welfare perspective,”IEEE Transactions on Mobile Computing, vol. 22, no. 9, pp. 5555–5567, 2022

    2022

  39. [47]

    A distributed competi- tive and collaborative coordination for multirobot systems,

    M. Liu, Y. Li, Y. Chen, Y. Qi, and L. Jin, “A distributed competi- tive and collaborative coordination for multirobot systems,”IEEE Transactions on Mobile Computing, 2024

    2024

  40. [48]

    Distributed dynamic task allocation for moving target tracking of networked mobile robots usingk- wta network,

    K. Liu and Y. Zhang, “Distributed dynamic task allocation for moving target tracking of networked mobile robots usingk- wta network,”IEEE Transactions on Neural Networks and Learning Systems, 2024

    2024

  41. [49]

    Robust k-wta network generation, analysis, and applications to multiagent coordina- tion,

    Y. Qi, L. Jin, X. Luo, Y. Shi, and M. Liu, “Robust k-wta network generation, analysis, and applications to multiagent coordina- tion,”IEEE Transactions on Cybernetics, vol. 52, no. 8, pp. 8515–8527, 2021

    2021

  42. [50]

    Energy-efficient distributed mobile crowd sensing: A deep learning approach,

    C. H. Liu, Z. Chen, and Y. Zhan, “Energy-efficient distributed mobile crowd sensing: A deep learning approach,”IEEE Journal on Selected Areas in Communications, vol. 37, no. 6, pp. 1262–1276, 2019

    2019

  43. [51]

    Energy-efficient uav crowdsensing with multiple charging stations by deep learning,

    C. H. Liu, C. Piao, and J. Tang, “Energy-efficient uav crowdsensing with multiple charging stations by deep learning,” inIEEE INFO- COm 2020-IEEE conference on computer communications. IEEE, 2020, pp. 199–208

    2020

  44. [52]

    Ex- ploring both individuality and cooperation for air-ground spatial crowdsourcing by multi-agent deep reinforcement learning,

    Y. Ye, C. H. Liu, Z. Dai, J. Zhao, Y. Yuan, G. Wang, and J. Tang, “Ex- ploring both individuality and cooperation for air-ground spatial crowdsourcing by multi-agent deep reinforcement learning,” in 2023 IEEE 39th International Conference on Data Engineering (ICDE). IEEE, 2023...

    2023

  45. [53]

    Cadre: A cascade deep reinforcement learning framework for vision-based autonomous urban driving,

    Y. Zhao, K. Wu, Z. Xu, Z. Che, Q. Lu, J. Tang, and C. H. Liu, “Cadre: A cascade deep reinforcement learning framework for vision-based autonomous urban driving,” inProceedings of the AAAI conference on artificial intelligence, vol. 36, no. 3, 2022, pp. 3481–3489

    2022

  46. [54]

    Human-drone collaborative spatial crowdsourcing by memory- augmented and distributed multi-agent deep reinforcement learn- ing,

    Y. Wang, C. H. Liu, C. Piao, Y. Yuan, R. Han, G. Wang, and J. Tang, “Human-drone collaborative spatial crowdsourcing by memory- augmented and distributed multi-agent deep reinforcement learn- ing,” in2022 IEEE 38th International Conference on Data Engineering (ICDE). IEEE, 202...

    2022

  47. [55]

    Urbangpt: Spatio-temporal large language models,

    Z. Li, L. Xia, J. Tang, Y. Xu, L. Shi, L. Xia, D. Yin, and C. Huang, “Urbangpt: Spatio-temporal large language models,” inProceed- ings of the 30th ACM SIGKDD Conference on Knowledge Discovery and Data Mining, 2024, pp. 5351–5362

    2024

  48. [56]

    Unist: A prompt- empowered universal model for urban spatio-temporal predic- tion,

    Y. Yuan, J. Ding, J. Feng, D. Jin, and Y. Li, “Unist: A prompt- empowered universal model for urban spatio-temporal predic- tion,” inProceedings of the 30th ACM SIGKDD Conference on Knowl- edge Discovery and Data Mining, 2024, pp. 4095–4106

    2024

  49. [57]

    Unity game engine,

    Unity Technologies, “Unity game engine,” https://unity.com/, 2023. Lei Hanreceived the Ph.D. degree in computer science from Northwestern Polytechnical Uni- versity, Xi’an, China, in 2023. He currently holds the position of Postdoctoral Researcher at Xidian University. His res...

    2023

  50. [2005]

    From November 2009 to October 2010, he was funded by the German Humboldt Foundation and went to the University of Mannheim in Germany for collaborative research

    From 2006 to 2009, he was a postdoctoral researcher at Nagoya University and a special researcher at Kyoto University in Japan. From November 2009 to October 2010, he was funded by the German Humboldt Foundation and went to the University of Mannheim in Germany for collaborati...

    2006

Pith tools

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