Swarm Characteristic Classification using Robust Neural Networks with Optimized Controllable Inputs

Abram Clark; Donald W. Peltier III; Isaac Kaminer; Marko Orescanin

REVIEW 4 major objections 6 minor 25 references

Swarm Characteristic Classification using Robust Neural Networks with Optimized Controllable Inputs

T0 review · 4 major / 6 minor · reviewed 2026-08-09 · deepseek-v4-flash

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

Pith's one-line read CNNs trained on enriched swarm data classify adversary tactics under uncertainty, and optimized defender motion pushes correct-classification probability near its maximum.

desk verdict Robustness half is solid and worth citing; the trajectory optimization half is a good idea that is undermined by evaluating only the NN's own confidence (STP) rather than actual classification accuracy. read the letter →

arxiv 2502.03619 v1 pith:7RPXEWXN submitted 2025-02-05 cs.LG

Donald W. Peltier III , Isaac Kaminer , Abram Clark , Marko Orescanin This is my paper · ORCID

classification cs.LG

keywords swarmtacticsneuralnetworktimeseriesclassificationdataaugmentationrobustnessoptimalcontroldefendertrajectoryoptimizationautonomousswarmsmeasurementnoise

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

Building on the authors' earlier work, this paper claims that a neural network trained on engagements spanning the expected range of operational conditions — defender count from 1 to 15, five defender motion patterns, and measurement noise from 0 to 50 — classifies a swarming adversary's tactic more accurately under uncertainty than a network trained on any single condition. The paper further claims that defender trajectories are controllable inputs: solving an optimal control problem that maximizes the Sum of True Predictions (the network's probability assigned to the true tactic, summed over all tactics) yields defender motions that raise classification confidence while respecting speed, acceleration, collision, and operating-area constraints. If correct, this lets defenders identify an adversary swarm's tactics with fewer assets than the original 10-vs-10 setting required, and lets a planner read off the minimum number of defenders needed for a chosen confidence threshold. The evidence is built from simulated swarm-on-swarm engagements with four known adversary tactics (Greedy, Greedy+, Auction, Auction+).

What carries the argument

The central object is the Sum of True Predictions (STP), defined as the trace of the matrix $P_{\text{stack}}$ whose entry $P_{kl}$ is the trained neural network's predicted probability of tactic $l$ when tactic $k$ is the true tactic generating the adversary response; thus $\mathrm{STP}(P_D) = \sum_{k=1}^{n} P_{kk}$, where $P_D$ is the matrix of all defender trajectories. STP is the objective of an optimal control problem: the defender dynamics are a double integrator, the decision variables are the defender trajectories $P_D$, each tactic's adversary response $P_{A_k}(P_D)$ feeds the network, and the constraints fix initial positions and bound velocity, acceleration, inter-defender separation, and the operating area. Maximizing STP by nonlinear programming steers the adversaries into motions that the network reads confidently. The supporting mechanism for robustness is dataset enrichment: sub-datasets spanning each Variable of Interest are generated independently and joined into a 'Combined' dataset with consistent scaling, and a CNN is trained on it.

What would settle it

Evaluate the robust 'Combined' network on held-out engagements whose defender count, defender motion, and noise level all fall outside every training range (e.g., 16–20 defenders, a sixth motion type, and noise levels 51–60); if it does not beat the best single-condition network on that held-out region, the robustness claim is false.

Watch

Extended reading notes

Core claim

The authors establish two things. First, a 'combined' training dataset that varies defender number, defender motion, and measurement noise yields a CNN whose test accuracy stays high across the whole range of those variables, whereas networks trained on a single setting develop bias toward that setting and degrade on dissimilar conditions. Second, defender motion can be optimized for a trained network: treating the adversary's trajectory as a deterministic function of the defender trajectory for each tactic, and maximizing the trace of the network's probability-output matrix (the Sum of True Predictions) over candidate defender trajectories — subject to velocity, acceleration, collision-avoidance, and operating-area constraints — produces defender motions that raise the true-tactic probability. With a larger training set the optimized Sum of True Predictions reached 399 of a possible 400, and repeating the optimization for 1 to 10 defenders yields a 'Best STP @ ND' curve that identifies the minimum defender count needed for a required confidence level.

Load-bearing premise

The optimization's value depends on the defender already having a complete and accurate model of how the adversary will respond for every tactic in the library, which is exactly the knowledge the classifier is meant to supply in the field.

Editorial extensions

If this is right

A planner can use the 'Best STP @ ND' curve to choose the minimum number of defenders needed for a desired classification confidence, replacing the original study's fixed 10-defender baseline.
The combined-dataset training recipe extends to other deployment variations the paper lists as future work, such as formation shape, intraswarm dispersion, weapon and sensor range, and velocity or acceleration limits.
Because the optimal control formulation includes operating-area and collision constraints, the classifier becomes usable in restricted airspace, waterways, or other geometrically constrained environments.
Investing in more training instances compounds: the larger 'Combined DM+' network shows stronger input-to-output gradients, and the optimizer exploits those gradients to reach near-maximum STP from multiple initial defender motions.
A network trained on a single defender motion is biased toward similar motions (e.g., Star and Semi perform alike, Straight performs worst), so combined training is a prerequisite for the optimization framework to work across motion types.

Reading between the lines

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

Inferred: in a real engagement the true tactic is unknown, so the optimizer cannot enumerate the adversary responses $P_{A_k}(P_D)$ from a simulator. A practical variant would optimize defender motion against the classifier's own saliency gradients, treating the network as the response model; that closed loop is not tested here.
Inferred: the same 'probe by controlled motion, then read intent' logic transfers to other multi-agent settings — e.g., a vehicle choosing a small maneuver to disambiguate another driver's intent — whenever a differentiable response model is available.
Inferred: the minimum-defender result is demonstrated for a single engagement seed; a robust planner would want the 'Best STP @ ND' curve averaged over many seeds and engagement geometries before committing assets.
Inferred: comparing the combined network against an ensemble of single-condition networks with equal total training budget would isolate whether the gain comes from data diversity or simply from more training examples.

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.

Desk Editor's Note

read the letter

Colleague,

The thing to know: the robustness half is credible and useful; the optimization half is not yet validated. The authors extend their TAES 2024 work by training CNNs on datasets that vary defender count, defender motion, and measurement noise. The combined-condition NNs consistently beat condition-specific NNs on held-out sub-dataset test sets. That is a clean, reproducible result, with code public. It gives a single classifier that works across engagements it wasn't trained on, which is real value.

The optimization framework is the newer piece: cast defender trajectory design as an optimal control problem that maximizes the sum of true predictions (STP) of the trained NN, subject to airspace and kinematic constraints. The formulation is sensible, and the plots show STP going up (e.g., from 345 to 367, and up to 399/400 with a larger NN). But that is the problem: the only success measure is STP, the NN's own softmax confidence for the true tactic. They never show that optimized defender motion actually changes the NN's argmax or improves top-1 accuracy on held-out engagements. STP is not calibrated accuracy. Section IV.E asserts "Required STP ≥ 390" guarantees "minimum NN prediction accuracy of 90%" with no calibration curve. That is a load-bearing claim for the minimum-defender analysis.

There is also a deeper conceptual issue: computing STP requires PAk(PD), the adversary response for every possible tactic, given the defender trajectory. The defender must already know the adversary behavior model the classifier is supposed to be discovering. In the simulation environment this is available; in the field it isn't. So the deployment story is weaker than the abstract implies.

Minor caveats: the robustness figures have no error bars, and the minimum-defender curve comes from a single engagement seed. The NN architecture changes between experiments (input length, filters), though that is defensible as fair comparison with prior work.

Who should read this? Anyone working on data augmentation for time-series classification or on using ML for autonomous swarm defense. The robustness finding deserves a serious referee. The optimization needs substantial additional validation before it should be promoted. I would send it to review, with the clear instruction that the STP-only evaluation and the missing calibration are major revision items.

Referee Report

4 major / 6 minor

Summary. The paper proposes two linked contributions for classifying swarm tactics from simulated trajectory data. First, it enriches the training data of a CNN time-series classifier by varying three 'variables of interest'—defender number, defender motion, and measurement noise—and shows that a CNN trained on the combined dataset outperforms CNNs trained on individual sub-datasets when tested on held-out sub-dataset test sets. Second, it casts defender trajectory design as an optimal control problem whose objective is to maximize the trained CNN's Sum of True Predictions (STP), i.e., the sum of softmax probabilities assigned to the true tactic under all possible adversary tactics. The authors use this framework to study how optimized defender motion affects STP and to estimate the minimum number of defenders needed to reach a required STP threshold.

Significance. If the claims are upheld, the paper would provide a practical recipe for making a pretrained swarm-tactic classifier robust to operational variability and a principled way to steer defender motion to improve classifier confidence. Strengths of the paper include the systematic treatment of three separate uncertainty axes, the use of held-out sub-dataset test sets for the robustness comparisons, and the release of a public code repository. The robustness result is internally consistent and, if replicated with error bars, would be a solid empirical contribution. The optimization contribution is conceptually interesting but, as detailed below, is currently validated only by the classifier's own confidence scores rather than by actual classification accuracy, which is a load-bearing gap for the central deployment claim.

major comments (4)

[Section IV.D and Abstract] The central claim that optimized defender trajectories 'maximize the probability of correct NN tactic classification' is not tested. All optimization results (Figures 9, 11, 12, 14) report STP, which is the trained NN's own softmax output, but no experiment reports whether optimized defender motion improves top-1 accuracy or changes the argmax label on held-out engagements. STP is not a calibrated estimate of classification accuracy: a high diagonal entry can coexist with a wrong argmax if another class receives even higher probability. The paper needs to report accuracy, confusion matrices, or calibration curves for optimized versus initial trajectories on held-out simulation data, not just STP values.
[Section IV.E, Figure 15] The minimum-defender analysis is based on a single engagement example and a single trained model, with no repeated seeds or error bars. The claim that 'Required STP ≥ 390' ensures 'minimum NN prediction accuracy of 90%' is asserted without any calibration experiment. Since the STP-to-accuracy mapping is not established, the derived minimum defender counts are not supported. The authors should repeat the optimization across multiple engagement initializations and, ideally, multiple training seeds, and show a calibration curve between STP and actual classification accuracy.
[Section II.D and Section II.A] The optimization objective requires the defender to compute PAk(PD), the exact adversary trajectory response for every possible tactic k, for any candidate defender trajectory PD. This presumes perfect knowledge of adversary dynamics and initial states over the whole observation window. In an actual deployment, the true tactic is unknown, so the defender cannot enumerate the needed responses without already possessing the model the classifier is meant to discover. This limitation should be stated explicitly, and the framework should be stress-tested under model mismatch—for example, optimize using nominal adversary dynamics and then evaluate STP and accuracy under perturbed dynamics or partially observed states.
[Section IV.A-C] The robustness comparisons rest on a single training run per architecture and dataset combination. Given the central claim that combined-dataset CNNs 'consistently' outperform single-condition CNNs, the absence of error bars or multiple random seeds weakens the statistical support. The paper should either provide repeated-training variability or temper the 'consistently' language to reflect the single-seed evidence.

minor comments (6)

[Section II.D] The constraint notation is introduced informally; for example, 'PD(t) ∈ X' is not defined as a set, and the min/max velocity and acceleration constraints are written as separate inequalities without explicit variable definitions in the displayed problem. A table of symbols or a more complete mathematical statement would improve readability.
[Section III.A and Table II] The truncation of datasets to the minimum time length (e.g., Combined ND has max time steps 45 while the original has 58) is mentioned but its effect on classification is not analyzed. The authors should state how much time is lost and whether this truncation affects the comparison across VOIs.
[Section III.C] The statement that 'random generator seeds above 1,200 or 10,000 were used' is vague. Please specify exactly which seeds were used for each optimization experiment and how they were chosen, to support reproducibility.
[Figure 4] The texts '5 NN' and '10 DN NN' are unclear; it should say 'five networks' and '10-defender NN' respectively. Also, the accuracy values printed on the plot are difficult to read and should be moved or formatted.
[Figure 8] The legend distinguishes solid and dotted lines but the text refers to 'ensemble of 51 NN'; please clarify whether the dotted line is the ensemble of all noise-specific NNs or the combined-dataset NN, and make the legend consistent with the caption.
[Section IV.D, Figure 13] The saliency maps are described as supporting the claim that the 'Better NN' has improved gradients, but no quantitative measure of saliency is given. A simple metric (e.g., mean absolute gradient across inputs) would make the comparison more rigorous.

Circularity Check

1 steps flagged · score 6.0 of 10

The optimization half of the paper reports the optimizer's own objective (STP) as classification performance; the robustness half is externally evaluated and not circular.

fitted input called prediction [Section II.D (STP definition and optimal-control objective); Section IV.D-IV.E (oSTP results and STP-to-accuracy threshold)]
"We use the Sum of True Predictions (STP) to define the performance of the NN to be maximized. ... min_{PD} {J = -STP(PD)} ... if a required classifier performance minimum threshold is set, such as a Required STP ≥ 390 to ensure a minimum NN prediction accuracy of 90%, the Best STP plot can guide the selection of the optimal defender number and initial trajectory."

STP is defined as the trained NN's own softmax output for the true tactic, and it is simultaneously the quantity maximized by the defender-trajectory optimization. Thus the reported oSTP improvements (e.g., oSTP=399/400) are the optimizer's objective values, not independent measurements of classification performance. The paper then converts an STP threshold into a 'minimum NN prediction accuracy of 90%' without calibration, even though STP is a sum of four per-tactic confidences and can rise while the argmax label remains wrong. The central deployment claim -- that optimized defender motion maximizes the probability of correct NN tactic classification -- therefore reduces, by construction, to maximizing the NN's own confidence, which is the same function used to score the result.

full rationale

The dataset-enrichment half of the paper is not circular: combined VOI datasets are trained and evaluated on held-out test sets (Tables II-III; Figures 4-8), so the robustness finding is an external, reproducible comparison. The self-citations to [20] provide the simulator and baseline architecture but do not by themselves force the accuracy results. The circularity is concentrated in the optimization half: STP is both the objective function (min J = -STP) and the only reported success metric, and Section IV.E equates an STP threshold with accuracy by fiat. Because no experiment reports top-1 accuracy or any calibration of STP on optimized trajectories, the paper's headline claim about maximizing the probability of correct classification is not independently validated; it is a restatement of the optimized objective. This is a partial circularity (score 6) rather than a total one, since the robust-classifier results stand on independent test data.

Assumptions & free parameters 3 free parameters · 5 assumptions · 0 invented entities

The central results rest primarily on simulation-based domain assumptions rather than on new physical entities. The most consequential assumption is the perfect knowledge of adversary response models needed by the optimization, which is not a real-world capability demonstrated here.

free parameters (3)

CNN architecture hyperparameters per experiment (input length, filters, kernel size, pool size, dropout) = Table I: input 20 or 50, filters 32 or 64, kernels 7 or 7,5,3, pool 5 or 3, dropout 0.1 or 0.4
Chosen by hand from prior work, not fitted to the target claim; they set model capacity and training dynamics and vary across the three VOI experiments.
Engagement reference values (10v10, star motion, noise 0, 20-step input) = ND=10, NA=10, motion=Star, noise=0
Baseline scenario used to hold all other variables fixed while varying one VOI; these choices shape the combined datasets and the reported accuracy levels.
STP threshold for 'required' confidence = 390 STP
Operational threshold chosen by the authors for the minimum-defender analysis; a different threshold changes the conclusion about how many defenders are needed.

assumptions (5)

domain assumption The adversary swarm uses exactly one of four predefined tactics (Greedy, Greedy+, Auction, Auction+), identical across agents, with no tactic switching.
Section II.A; the classifier's label space and the optimization's PAk enumeration depend on this finite, stationary set.
domain assumption Complete adversary trajectories are known and the NN input is the full 2D position/velocity history of all adversaries.
Section II.A; this removes perception and data association errors from the problem.
ad hoc to paper Adversary response dynamics PAk(PD) are known and computable for every tactic during optimization.
Section II.D; the STP objective requires evaluating all four simulated responses at every optimizer iteration, which presumes a perfect adversary model.
domain assumption The simulator used to generate training data faithfully represents real swarm engagements at a macro Newtonian level.
Section II.A and V.B; all results are on simulated data, so real-world transfer rests on this premise.
ad hoc to paper The trained neural network's probability outputs are calibrated so that maximizing STP corresponds to maximizing classification accuracy.
Section II.D and IV.D; the paper does not demonstrate calibration, so STP is a confidence measure, not a ground-truth measure.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Swarm Characteristic Classification using Robust Neural Networks with Optimized Controllable Inputs." pith.science (2026). https://pith.science/paper/7RPXEWXN

@misc{pith2026250203619,
  author       = {Pith},
  title        = {Pith review of: Swarm Characteristic Classification using Robust Neural Networks with Optimized Controllable Inputs},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/7RPXEWXN}},
  note         = {Machine review of arXiv:2502.03619}
}

read the original abstract

Having the ability to infer characteristics of autonomous agents would profoundly revolutionize defense, security, and civil applications. Our previous work was the first to demonstrate that supervised neural network time series classification (NN TSC) could rapidly predict the tactics of swarming autonomous agents in military contexts, providing intelligence to inform counter-maneuvers. However, most autonomous interactions, especially military engagements, are fraught with uncertainty, raising questions about the practicality of using a pretrained classifier. This article addresses that challenge by leveraging expected operational variations to construct a richer dataset, resulting in a more robust NN with improved inference performance in scenarios characterized by significant uncertainties. Specifically, diverse datasets are created by simulating variations in defender numbers, defender motions, and measurement noise levels. Key findings indicate that robust NNs trained on an enriched dataset exhibit enhanced classification accuracy and offer operational flexibility, such as reducing resources required and offering adherence to trajectory constraints. Furthermore, we present a new framework for optimally deploying a trained NN by the defenders. The framework involves optimizing defender trajectories that elicit adversary responses that maximize the probability of correct NN tactic classification while also satisfying operational constraints imposed on the defenders.

Figures

Figures reproduced from arXiv: 2502.03619 by the authors.

**Figure 2.** Five distinct defender motions elicit unique adversary [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗

**Figure 3.** Overview of the trajectory optimization process for [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗

Figures from the paper (9 more)

**Figure 4.** Figure 4: Defender Number Experiment: comparison of 5 NN; [PITH_FULL_IMAGE:figures/full_fig_p007_4.png]

**Figure 5.** Figure 5: Accuracy improves slightly when a NN is trained [PITH_FULL_IMAGE:figures/full_fig_p007_5.png]

**Figure 7.** Figure 7: Defender Motion Experiment: comparison of 6 NN; [PITH_FULL_IMAGE:figures/full_fig_p008_7.png]

**Figure 9.** Figure 9: Defender trajectory optimization results for one [PITH_FULL_IMAGE:figures/full_fig_p008_9.png]

**Figure 8.** Figure 8: Comparison of NN performance during inference [PITH_FULL_IMAGE:figures/full_fig_p008_8.png]

**Figure 10.** Figure 10: Best defender initial trajectory (IT) for various [PITH_FULL_IMAGE:figures/full_fig_p009_10.png]

**Figure 11.** Figure 11: Optimized STP (oSTP) and defender trajectories [PITH_FULL_IMAGE:figures/full_fig_p009_11.png]

**Figure 14.** Figure 14: Optimization comparison using two different neural [PITH_FULL_IMAGE:figures/full_fig_p010_14.png]

**Figure 16.** Figure 16: Subplots on right show the optimized adversarial [PITH_FULL_IMAGE:figures/full_fig_p010_16.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

25 extracted references · 24 canonical work pages

[1]

Ukrainian drone disables Russian warship near Russia’s Novorossiysk port,

T. Balmforth, “Ukrainian drone disables Russian warship near Russia’s Novorossiysk port,” Reuters, Aug. 2023

work page 2023
[2]

AI’s ‘Oppenheimer moment’: autonomous weapons enter the battlefield,

N. Robins-Early, “AI’s ‘Oppenheimer moment’: autonomous weapons enter the battlefield,” The Guardian, July 2024

work page 2024
[3]

Kallenborn, Are drone swarms weapons of mass destruction? The Counterproliferation Papers Future Warfare Series, no

Z. Kallenborn, Are drone swarms weapons of mass destruction? The Counterproliferation Papers Future Warfare Series, no. 60, Maxwell Air Force Base, Alabama, USA: U.S. Air Force Center for Strategic Deterrence Studies, Air University, 2020

work page 2020
[4]

Swarms over the Strait,

S. Pettyjohn, M. Campbell, and H. Dennis, “Swarms over the Strait,” tech. rep., Center for a New American Security, June 2024

work page 2024
[5]

Artificial Intelligence and the Future of Warfare,

M. L. Cummings, “Artificial Intelligence and the Future of Warfare,” tech. rep., International Security Department and US and the Americas Programme, July 2017

work page 2017
[6]

Hicks Underscores U.S. Innovation in Unveiling Strategy to Counter China’s Military Buildup,

J. Clark, “Hicks Underscores U.S. Innovation in Unveiling Strategy to Counter China’s Military Buildup,” U.S. Department of Defense , Aug. 2023

work page 2023
[7]

Report on Applied Research Directions and Future Opportunities for Swarm Systems in Defence,

A. Hepworth, “Report on Applied Research Directions and Future Opportunities for Swarm Systems in Defence,” Tech. Rep. No. 11, Australian Army Research Center, 2022

work page 2022
[8]

The Ethics of Autonomous Cars,

P. Lin, “The Ethics of Autonomous Cars,” The Atlantic , Oct. 2013

work page 2013

Show all 25 references

[9]

A Survey on Autonomous Vehicles Interactions with Human and other Vehicles,

B. Jafary, E. Rabiei, M. Diaconeasa, H. Masoomi, L. Fiondella, and A. Mosleh, “A Survey on Autonomous Vehicles Interactions with Human and other Vehicles,” (Los Angeles, CA), Sept. 2018

2018
[10]

Implementation of Artificial Intelligence on Air Traffic Control - A Systematic Literature Review,

R. Emha Abdillah, H. Moenaf, L. Fadullah Rasyid, S. Achmad, and R. Sutoyo, “Implementation of Artificial Intelligence on Air Traffic Control - A Systematic Literature Review,” in 2024 18th International Conference on Ubiquitous Information Management and Communica- tion (IMCOM...

2024
[11]

Artificial intelligence, machine learning and deep learning in advanced robotics, a review,

M. Soori, B. Arezoo, and R. Dastres, “Artificial intelligence, machine learning and deep learning in advanced robotics, a review,” Cognitive Robotics, vol. 3, pp. 54–70, Jan. 2023

2023
[12]

Path planning and collision avoidance for autonomous surface vehicles II: a comparative study of algorithms,

A. Vagale, R. Bye, R. Oucheikh, O. Osen, and T. Fossen, “Path planning and collision avoidance for autonomous surface vehicles II: a comparative study of algorithms,” Journal of Marine Science and Technology, Feb. 2021

2021
[13]

Coordinating Hundreds of Cooperative, Autonomous Vehicles in Warehouses,

P. R. Wurman, R. D’Andrea, and M. Mountz, “Coordinating Hundreds of Cooperative, Autonomous Vehicles in Warehouses,” AI Magazine , vol. 29, pp. 9–9, Mar. 2008. Number: 1. 12

2008
[14]

Path Planning for Autonomous Mobile Robots: A Review,

J. R. S ´anchez Ib´a˜nez, C. Perez-del Pulgar, and A. Garcia, “Path Planning for Autonomous Mobile Robots: A Review,” Sensors, vol. 21, p. 7898, Nov. 2021

2021
[15]

AI-driven warehouse automation: A comprehensive review of systems,

O. Amoo, E. Sodiya, U. Umoga, and A. Atadoga, “AI-driven warehouse automation: A comprehensive review of systems,” GSC Advanced Re- search and Reviews , vol. 18, pp. 272–282, Feb. 2024

2024
[16]

Optimal motion planning in rapid-fire combat situations with attacker uncertainty,

C. Walton, P. Lambrianides, I. Kaminer, J. Royset, and Q. Gong, “Optimal motion planning in rapid-fire combat situations with attacker uncertainty,” Naval Research Logistics, vol. 65, pp. 101–119, Mar. 2018

2018
[17]

Defense against Adversarial Swarms with Parameter Uncertainty,

C. Walton, I. Kaminer, Q. Gong, A. H. Clark, and T. Tsatsanifos, “Defense against Adversarial Swarms with Parameter Uncertainty,” Sensors, vol. 22, p. 4773, Jan. 2022

2022
[18]

Modeling Large-Scale Adversarial Swarm Engagements using Optimal Control,

T. Tsatsanifos, A. H. Clark, C. Walton, I. Kaminer, and Q. Gong, “Modeling Large-Scale Adversarial Swarm Engagements using Optimal Control,” in 2021 60th IEEE Conference on Decision and Control , pp. 1244–1249, Dec. 2021. ISSN: 2576-2370

2021
[19]

Partial Observability Analysis of an Adversarial Swarm Model,

Q. Gong, W. Kang, C. Walton, I. Kaminer, and H. Park, “Partial Observability Analysis of an Adversarial Swarm Model,” AIAA Journal of Guidance, Control, and Dynamics , vol. 43, pp. 250–261, Feb. 2020

2020
[20]

Swarm Charac- teristics Classification Using Neural Networks,

D. W. Peltier, I. Kaminer, A. Clark, and M. Orescanin, “Swarm Charac- teristics Classification Using Neural Networks,” IEEE Transactions on Aerospace and Electronic Systems , pp. 1–12, 2024

2024
[21]

Data Augmentation for Time-Series Clas- sification: An Extensive Empirical Study and Comprehensive Survey,

Z. Gao, H. Liu, and L. Li, “Data Augmentation for Time-Series Clas- sification: An Extensive Empirical Study and Comprehensive Survey,” Aug. 2024. arXiv:2310.10060 [cs]

2024 arXiv
[22]

Data Augmentation techniques in time series domain: a survey and taxonomy,

G. Iglesias, E. Talavera, A. Gonz ´alez-Prieto, A. Mozo, and S. G ´omez- Canaval, “Data Augmentation techniques in time series domain: a survey and taxonomy,” Neural Computing and Applications, vol. 35, pp. 10123– 10145, May 2023

2023
[23]

Nocedal and S

J. Nocedal and S. J. Wright, Numerical optimization. Springer series in operations research and financial engineering, New York, NY: Springer, second edition ed., 2006

2006
[24]

Math- Works, Inc., 2024

MathWorks, Inc., fmincon: Constrained nonlinear minimization . Math- Works, Inc., 2024

2024
[25]

MathWorks, Inc., 2024

MathWorks, Inc., importNetworkFromTensorFlow: Import TensorFlow network into MATLAB . MathWorks, Inc., 2024. Donald W. Peltier III received the B.S. degree in aerospace engineering from The University of Texas at Austin, Austin, TX, USA, in 2006, and the M.S. degree in aeronau...

2024

Pith tools

Reviewed August 9, 2026 · model on record in the stance chip above.

[1] [1]

Ukrainian drone disables Russian warship near Russia’s Novorossiysk port,

T. Balmforth, “Ukrainian drone disables Russian warship near Russia’s Novorossiysk port,” Reuters, Aug. 2023

work page 2023

[2] [2]

AI’s ‘Oppenheimer moment’: autonomous weapons enter the battlefield,

N. Robins-Early, “AI’s ‘Oppenheimer moment’: autonomous weapons enter the battlefield,” The Guardian, July 2024

work page 2024

[3] [3]

Kallenborn, Are drone swarms weapons of mass destruction? The Counterproliferation Papers Future Warfare Series, no

Z. Kallenborn, Are drone swarms weapons of mass destruction? The Counterproliferation Papers Future Warfare Series, no. 60, Maxwell Air Force Base, Alabama, USA: U.S. Air Force Center for Strategic Deterrence Studies, Air University, 2020

work page 2020

[4] [4]

Swarms over the Strait,

S. Pettyjohn, M. Campbell, and H. Dennis, “Swarms over the Strait,” tech. rep., Center for a New American Security, June 2024

work page 2024

[5] [5]

Artificial Intelligence and the Future of Warfare,

M. L. Cummings, “Artificial Intelligence and the Future of Warfare,” tech. rep., International Security Department and US and the Americas Programme, July 2017

work page 2017

[6] [6]

Hicks Underscores U.S. Innovation in Unveiling Strategy to Counter China’s Military Buildup,

J. Clark, “Hicks Underscores U.S. Innovation in Unveiling Strategy to Counter China’s Military Buildup,” U.S. Department of Defense , Aug. 2023

work page 2023

[7] [7]

Report on Applied Research Directions and Future Opportunities for Swarm Systems in Defence,

A. Hepworth, “Report on Applied Research Directions and Future Opportunities for Swarm Systems in Defence,” Tech. Rep. No. 11, Australian Army Research Center, 2022

work page 2022

[8] [8]

The Ethics of Autonomous Cars,

P. Lin, “The Ethics of Autonomous Cars,” The Atlantic , Oct. 2013

work page 2013