Air Traffic Controller Task Demand via Graph Neural Networks: An Interpretable Approach to Airspace Complexity

Dewi Gould; Edward Henderson; George De Ath; Nick Pepper; Richard Everson

REVIEW 3 major objections 5 minor 41 references

Air Traffic Controller Task Demand via Graph Neural Networks: An Interpretable Approach to Airspace Complexity

T0 review · 3 major / 5 minor · reviewed 2026-08-06 · deepseek-v4-flash

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

Pith's one-line read An aircraft-ablation score predicts controller workload better than traffic counts, and can pinpoint which aircraft drives it.

desk verdict Solid GNN clearance-count predictor with an interpretable per-aircraft score, but the workload validation leans on the same clearance proxy used for training; worth refereeing with a demand for independent validation. read the letter →

arxiv 2507.13423 v1 pith:Z4RCNSXM submitted 2025-07-17 cs.LG cs.AI

Edward Henderson , Dewi Gould , Richard Everson , George De Ath , Nick Pepper This is my paper · ORCID

classification cs.LGcs.AI

keywords airtrafficcontroltaskdemandairspacecomplexitygraphneuralnetworksinterpretabilityclearancepredictionablationstudyattention

verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

The paper tries to establish that near-term air traffic controller task demand can be read off from a graph neural network trained on static traffic snapshots, and that the network can say which aircraft are driving the workload. Its proposed measure is a per-aircraft ablation score: remove one aircraft from the scenario graph, recompute the predicted number of clearances, and take the difference. The authors report that this score, summed over aircraft, correlates at 0.91 with the clearances actually issued in the next ten minutes, whereas established indicators such as TLPD complexity and graph indicators correlate more strongly with raw traffic count than with clearances. A sympathetic reader would care because the method turns a subjective and hard-to-measure quantity (controller mental workload) into an explicit, per-aircraft, data-driven attribution that does not require hand-crafted route or hotspot knowledge. If the claim is right, controller training and sector redesign could point at the specific aircraft causing demand rather than just at busy airspace.

What carries the argument

The load-bearing object is the scenario graph $G=(V,E)$. Each aircraft is a node carrying sixteen features, including lateral position, flight level, ground speed, climb rate, communication state, and predicted time to sector exit; edges are placed between aircraft whose vertical flight-level ranges (current, cleared, and exit flight levels, with a ten-flight-level buffer) overlap, and each edge carries separation distance and closing speed. Two graph attention layers propagate context across these edges, and two prediction heads output total and per-aircraft clearance counts, trained jointly with a quantile (pinball) loss that predicts 10th, 50th, and 90th percentiles to tolerate the noise in the clearance target. The defining mechanism is the ablation identity $\phi_i = C(G) - C(G \setminus \{v_i\})$, the change in predicted total clearances when aircraft $i$ is removed; this converts the network's prediction into an interpretable, per-aircraft attribution of task demand.

What would settle it

Take scenarios with low clearance counts but known high controller workload, for instance ones where controllers report being busy while issuing few instructions, and compare the GNN ablation scores against controllers' own ratings of which aircraft demanded attention; if the ablation scores do not track those ratings in low-clearance cases, the proxy assumption fails.

Watch

Extended reading notes

Core claim

The central claim is that the number of trajectory-affecting clearances an ATCO will issue in the next ten minutes, namely heading, level, and speed instructions, is a learnable surrogate for task demand, and that the marginal effect of each aircraft on that prediction is a valid per-aircraft complexity score. The GNN, built on a graph where nodes are aircraft and edges encode overlapping vertical flight-level ranges, predicts clearance totals with a mean absolute error of $1.83 \pm 0.05$ on a held-out week of London Middle Sector traffic, compared with $3.44 \pm 0.09$ for a minimum-clearance heuristic and $1.87 \pm 0.04$ and $1.86 \pm 0.04$ for random forest and XGBoost regressors. In the qualitative comparison, summed task demand correlates $0.91$ with issued clearances and $0.74$ with traffic count, while TLPD complexity correlates $0.79$ with GNN task demand, $0.76$ with clearances, and $0.94$ with traffic count. The authors interpret the contrast as evidence that their score tracks interactive complexity rather than merely reflecting aircraft count.

Load-bearing premise

The whole approach assumes that the number of clearances issued in the next ten minutes is a faithful proxy for how much work an air traffic controller actually experiences, even though the paper itself notes that a complex situation can sometimes be solved with few, well-designed clearances.

Editorial extensions

If this is right

Summed over aircraft, the ablation score behaves like a short-term forecast: its peaks lead the instantaneous graph indicators by 3-5 minutes, so it could serve as an early-warning complexity signal.
Because the score is per aircraft, a training replay tool could highlight which aircraft the model expects to consume controller attention in the next ten minutes.
The model matches or beats random forest and XGBoost while using graph structure, and degrades more than any single feature when edges are randomized, suggesting the pairwise interaction structure itself carries the signal.
In the high-clearance tail (at least 12 clearances in ten minutes), the GNN under-predicts by a median of -2.0 clearances versus -3.9 and -3.6 for the regressors, meaning it is most accurate precisely where workload is likely highest.
The approach transfers to other sectors without recalibrating expert hotspot weights, because it learns interaction features from data instead of relying on hand-crafted factors.

Reading between the lines

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

Editorial extension: the single-aircraft ablation is a one-player marginal contribution; computing Shapley values over coalitions of aircraft would reveal whether pairwise interactions materially change the attributed demand, and the paper's edge-ablation experiments suggest interactions do matter.
Editorial extension: the same recipe of training a predictor on instruction counts and then ablating entities could transfer to other human-in-the-loop settings such as drone fleet management or rail traffic control, wherever an instruction log is available as supervision.
Editorial extension: because the GNN score is only 0.74 correlated with traffic count while TLPD is 0.94, the score should be more sensitive to geometrically different traffic with the same count; constructing matched-count synthetic scenarios with different conflict geometries would test this directly.
Editorial extension: the paper leaves expert validation as future work; a direct study in which controllers label priority aircraft would settle whether the ablation score tracks human attention or only clearance likelihood.

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

Plain take: the paper's core predictive claim holds up. A GNN on a flight-level-overlap graph predicts next-10-minute clearance counts on a held-out week with MAE 1.83, beating an ATCO-inspired minimum-clearance baseline (3.44, p<1e-12) and edging RF/XGBoost (1.87/1.86, p=0.04/0.05). The graph connectivity ablation is convincing: random edges or no edges degrade performance more than any single feature. That part is a competent, useful result.

The new thing is the per-aircraft task-demand score phi_i = C(G) - C(G\{v_i}). The idea of ablating agents to attribute workload is reasonable, and the three worked examples (Fig 8) are readable and align with operational intuition. The authors are also honest that clearance count is only a proxy for workload and that no ATCO-labelled dataset was collected.

The soft spot is exactly where the reader put it. The quantitative validation in Table 5 correlates phi with the same clearance counts the model was trained to predict. It is not data leakage—the test week is held out—but it is construct circularity: the model is optimised for that target, so a 0.91 correlation with that target is not independent evidence that phi measures workload. The 0.74 correlation with traffic count also undercuts the claim that the score captures interaction-driven complexity beyond raw traffic. To support the interpretability claim, the authors need a separate criterion: ATCO ratings, eye-tracking, or expert-prioritised aircraft.

Minor but real issues: the reserved test set is reported as 1,550 scenarios in Sec. II.A and 2,279 later; that needs fixing. No code or data is released, and the comparison to TLPD and Isufaj et al. is qualitative, with hand-picked examples. The improvement over RF/XGBoost is marginal, though the long-tail performance and signed error difference are in the GNN's favour.

Who this is for: people working on ATC complexity metrics, controller workload modelling, or GNN interpretability in operational settings. It is a useful case study and a fair baseline for future work. It deserves a serious referee: the methods are competently executed, the limitations are acknowledged, and the gap is a correctable validation deficit, not an internal inconsistency. I would send it to peer review with a request for independent workload validation, a resolved test-set inconsistency, and ideally code or data release.

Referee Report

3 major / 5 minor

Summary. This paper proposes a Graph Neural Network (GNN) that predicts the number of air traffic control clearances expected in the ten minutes following a static traffic snapshot in London Middle Sector. The graph representation encodes aircraft state as node features and potential vertical-range interactions as edge features, with GATv2 message-passing layers and dual graph-level and node-level prediction heads trained with quantile loss. From the trained model, the authors derive a per-aircraft task demand score by ablation: φ_i = C(G) − C(G \ {v_i}). The clearance-count prediction is compared with a minimum-clearance heuristic, random forest, and XGBoost, and the task-demand score is compared qualitatively and quantitatively with TLPD complexity and four graph indicators from Isufaj et al. over a four-hour operational period.

Significance. The paper has real strengths: it uses a substantial operational dataset (46,570 trajectories, 116,614 clearances), reserves a held-out test week, uses quantile regression to handle noisy targets, reports permutation feature importance, and includes a graph-connectivity ablation showing that graph structure matters for prediction. The clearance-prediction result is solid: MAE 1.83 versus 3.44 for the heuristic (p < 1e-12). The interpretable per-aircraft attribution idea is potentially useful for controller training and airspace redesign. However, the validation of the central task-demand claim is currently circular with respect to the training target, and no external workload ground truth is used, so the paper's headline claim of a 'more reliable estimator of ATCO task demand' is not yet established.

major comments (3)

[III.C, Table 5] The central claim that the GNN-based task demand is a more reliable estimator of ATCO task demand than TLPD or graph indicators rests on Table 5, which correlates the ablation score with the same clearance counts used to train the predictive model. Because φ_i is defined from a model optimised to predict N_clearances, a Pearson correlation of 0.91 with N_clearances is largely mechanical. The comparison is therefore not an equal test: TLPD complexity was independently calibrated by sector experts, whereas the GNN task-demand score is constructed from a model that was trained to reproduce the very signal used for validation. Without an external criterion—such as ATCO workload ratings, expert priority judgements, or at least a clearly separated construct-validity argument—the claimed superiority over existing metrics is unsupported.
[II.B.5 and III.C, Eq. (3), Fig. 7] Figure 7 and Table 5 use Σ_i φ_i as the 'GNN task demand', where φ_i = C(G) − C(G \ {v_i}). Since C is a nonlinear GNN, Σ_i [C(G) − C(G \ {v_i})] is not generally equal to C(G); no additivity property is proved or empirically checked. If the sum is intended to represent the total predicted task demand, the paper must either verify that the marginal contributions approximately sum to the total prediction, or use a cooperative-game allocation (e.g., Shapley values) that satisfies efficiency. Without this, the summed score used in the longitudinal comparison and in Table 5 is an unvalidated transformation of the model output.
[III.C, Fig. 8] The per-aircraft attribution is validated only with three hand-selected qualitative examples; no ATCO-labelled data or quantitative expert evaluation is provided, and the conclusion explicitly acknowledges that no such dataset was collected. Given that interpretable per-aircraft task demand is one of the three stated contributions, the absence of any external validation of the attributions leaves the interpretability claim under-supported. The paper should either temper the claim that the score 'attributes task demand to specific aircraft' or add an expert-evaluation study.

minor comments (5)

[II.A vs II.B.2] The held-out test set is described as 1,550 scenarios in Section II.A and as 2,279 held-out scenarios in Section II.B.2; please reconcile these numbers, as the discrepancy affects the reproducibility of the reported results.
[III.B, Table 4] The GNN's advantages over random forest and XGBoost are marginal (p = 0.04 and p = 0.05) with overlapping 95% confidence intervals; the abstract's 'significantly outperforms' should be attributed to the minimum-clearance baseline, with the comparison to tree regressors described more cautiously.
[II.A.1 and II.B.2] Several design choices—10-minute forecast horizon, 10-FL vertical buffer, 3-minute scenario sampling interval, quantile levels, and weighted-sampling percentile cut points—are fixed without sensitivity analysis; a brief robustness discussion would strengthen the claim that the graph representation, rather than these choices, drives the results.
[III.C, Table 5] The Pearson correlations are computed over a single four-hour period without confidence intervals or significance tests, and the text reports that correlations with graph indicators are maximised at a 3-5 minute lag, yet Table 5 appears to report unlagged values; please clarify whether the lag adjustment was applied and report the lag-adjusted correlations.
[II.A.1, footnote] The footnote 'A buffer of ten flight levels is included in this computation' should specify exactly how the buffer is applied (e.g., added to each aircraft's vertical range before the overlap test) and state that the units are flight levels.

Circularity Check

1 steps flagged · score 6.0 of 10

The task-demand score is a function of the GNN trained to predict clearance counts, so its validation against clearance counts is partly by construction.

fitted input called prediction [Sec. II.B.5, Eq. (3); Sec. III.C, Table 5]
"Once trained, we used the GNN clearance prediction model to infer the influence of individual aircraft on overall ATCO task demand... This provides a direct task demand score for each aircraft, phi_i: phi_i = C(G) - C(G \ {v_i})... Table 5 contains the Pearson correlation coefficients between the task demand estimation methods, the number of clearances issued and the traffic count... Our approach therefore appears to be a more reliable estimator of ATCO task demand."

The score phi_i is defined through C(G), the GNN's predicted clearance count, and that network is trained to predict the number of clearances issued in the following ten-minute window. Table 5 then uses corr(phi, N_clearances) = 0.91 as the main quantitative support for the claim that the GNN score is 'a more reliable estimator of ATCO task demand.' This is not an independent test: the derived score is validated against the same regression target used to fit C. TLPD and the graph indicators were never fitted to clearance counts, so comparing their clearance correlations against phi's is asymmetric. Independent validation would require ATCO workload labels or expert priority ratings, which the paper itself lists as future work in Sec. IV.

full rationale

The paper is otherwise self-contained: there are no load-bearing self-citations, and the GNN's raw clearance-prediction accuracy on a held-out week, compared against the minimum-clearance baseline, random forest, and XGBoost, is genuinely independent evidence for the graph-based prediction model. The partial circularity is confined to the interpretable task-demand contribution. Because Eq. (3) defines phi from a model trained to predict clearance counts, and Table 5 validates phi against those same counts, the central claim that the ablation score is a 'more reliable estimator of ATCO task demand' rests on an in-construct comparison rather than on an external workload criterion. The authors openly acknowledge that clearance counts are an imperfect proxy and that no ATCO-labelled dataset was collected, which mitigates but does not remove the circularity: the only quantitative support for the task-demand construct is the training target itself. I therefore assign a partial-circularity score of 6 rather than a higher score, because the clearance-prediction component is independently validated and the paper is transparent about the proxy limitation.

Assumptions & free parameters 8 free parameters · 6 assumptions · 1 invented entities

The central claim depends on several assumptions: clearance count as a proxy for workload, the graph construction rule, and the validity of the ablation score. Most are domain assumptions grounded in ATC practice, but the ablation score and the proxy assumption are load-bearing and only partially validated.

free parameters (8)

Forecast horizon = 10 minutes
The prediction window for clearances; chosen as 'initial research' with no sensitivity analysis (Sec. II.A).
Edge creation vertical buffer = 10 flight levels
Footnote in Sec. II.A.1: 'A buffer of ten flight levels is included in this computation.' Affects graph connectivity and all downstream results.
Sector boundary buffer = 0.3 degrees (~10-18 NM)
Sec. II.A: aircraft within this buffer of LMS are included in scenarios; chosen by the authors.
Scenario sampling interval = 3 minutes
Scenarios captured every 3 minutes; affects dataset size and temporal autocorrelation (Sec. II.A).
Quantile levels = 0.1, 0.5, 0.9
Used in pinball loss (Eq. 1); standard but chosen.
Weighted sampling percentile cut points = 10th, 50th, 90th percentiles of target
Sec. II.B.2: sampling strategy to balance rare high-clearance scenarios; cut points fit to training distribution.
GNN hyperparameters = lr=0.001, batch=8, 50 epochs, 2 GATv2 layers, hidden 64/32
Sec. II.B.2 implementation details; no systematic hyperparameter search reported.
Isufaj et al. re-implementation thresholds = 48 NM horizontal, 4,400 ft vertical
Sec. II.B.6: thresholds computed from data with buffer; used in the baseline graph indicators comparison.

assumptions (6)

domain assumption The number of clearances issued in the next 10 minutes is a valid proxy for ATCO task demand.
Explicitly stated in Sec. II.B.5 and Sec. III.B. The authors acknowledge it is weaker for low clearance counts, but it is the ground truth for training and validation.
domain assumption Edges based on overlapping vertical flight level ranges capture operationally relevant aircraft interactions.
Sec. II.A.1: 'if the vertical flight level ranges of two aircraft overlap, an edge is created.' Supported indirectly by the random-edge ablation, but the specific overlap rule is assumed.
ad hoc to paper The ablation difference phi_i = C(G) - C(G\{v_i}) measures an aircraft's marginal contribution to task demand.
Eq. 3 defines the score; the paper assumes this difference is a meaningful attribution without formal justification, and sums these scores without proving additivity.
domain assumption BADA trajectory simulations accurately estimate time to sector exit.
Used to compute the 'time to exit' node feature (Table 3); if the BADA model is inaccurate, the feature is noisy.
domain assumption The held-out week (Aug 11-17, 2019) is representative and independent of the training period.
Sec. II.A: a single week is reserved for testing; assumes no temporal distribution shift.
domain assumption The re-implementation of Isufaj et al. graph indicators is faithful.
Sec. III.C: thresholds (48 NM, 4,400 ft) are computed from data; the authors check correlation consistency but do not compare to the original authors' code.

invented entities (1)

Per-aircraft task demand score (phi_i)
purpose: To attribute ATCO workload to individual aircraft via ablation of the trained GNN.
Defined in Eq. 3; validated only within the paper against the training target (clearance counts) and qualitative examples. No external behavioral data (e.g., controller eye tracking or subjective ratings) is used.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Air Traffic Controller Task Demand via Graph Neural Networks: An Interpretable Approach to Airspace Complexity." pith.science (2026). https://pith.science/paper/Z4RCNSXM

@misc{pith2026250713423,
  author       = {Pith},
  title        = {Pith review of: Air Traffic Controller Task Demand via Graph Neural Networks: An Interpretable Approach to Airspace Complexity},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/Z4RCNSXM}},
  note         = {Machine review of arXiv:2507.13423}
}

read the original abstract

Real-time assessment of near-term Air Traffic Controller (ATCO) task demand is a critical challenge in an increasingly crowded airspace, as existing complexity metrics often fail to capture nuanced operational drivers beyond simple aircraft counts. This work introduces an interpretable Graph Neural Network (GNN) framework to address this gap. Our attention-based model predicts the number of upcoming clearances, the instructions issued to aircraft by ATCOs, from interactions within static traffic scenarios. Crucially, we derive an interpretable, per-aircraft task demand score by systematically ablating aircraft and measuring the impact on the model's predictions. Our framework significantly outperforms an ATCO-inspired heuristic and is a more reliable estimator of scenario complexity than established baselines. The resulting tool can attribute task demand to specific aircraft, offering a new way to analyse and understand the drivers of complexity for applications in controller training and airspace redesign.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

41 extracted references · 32 canonical work pages

[1]

Doc 4444 PANS-ATM,

International Civil Aviation Organization, “Doc 4444 PANS-ATM,” , 2007. URLhttps://www.ealts.com/documents/ ICAODoc4444AirTrafficManagement.pdf

work page 2007
[2]

CAP 493: Manual of Air Traffic Services (MATS) Part 1,

UK Civil Aviation Authority, “CAP 493: Manual of Air Traffic Services (MATS) Part 1,” , 2023. URLhttps://www.caa. co.uk/our-work/publications/documents/content/cap-493/

work page 2023
[3]

2022 European Aviation Environmental Report,

European Union Aviation Safety Agency (EASA), “2022 European Aviation Environmental Report,” , 2022. URLhttps: //www.easa.europa.eu/eco/eaer

work page 2022
[4]

Automation in Air Traffic Management,

SESAR Joint Undertaking, “Automation in Air Traffic Management,” , 2020. URLhttps://www.sesarju.eu/sites/ default/files/documents/reports/automation%20atm%20vision%20roadmap.pdf

work page 2020
[5]

Automation in Future Air Traffic Management: Effects of Decision Aid Reliability on Controller Performance and Mental Workload,

Metzger, U., and Parasuraman, R., “Automation in Future Air Traffic Management: Effects of Decision Aid Reliability on Controller Performance and Mental Workload,”Human Factors, Vol. 47, No. 1, 2005, pp. 35–49. https://doi.org/10.1518/ 0018720053653802. 17

work page 2005
[6]

Determining Air Traffic Complexity – Challenges and Future Development,

Antulov-Fantulin, B., Juričić, B., Radišić, T., and Çetek, C., “Determining Air Traffic Complexity – Challenges and Future Development,”Promet - Traffic & Transportation, Vol. 32, No. 4, 2020, pp. 475–485. https://doi.org/10.7307/ptt.v32i4.3401

doi:10.7307/ptt.v32i4.3401 2020
[7]

The Graph Neural Network Model,

Scarselli, F., Gori, M., Tsoi, A. C., Hagenbuchner, M., and Monfardini, G., “The Graph Neural Network Model,”IEEE Transactions on Neural Networks, Vol. 20, No. 1, 2009, pp. 61–80. https://doi.org/10.1109/TNN.2008.2005605

arXiv 2009
[8]

Towards an air traffic control complexity metric based on workspace constraints,

van Paassen, M. M., d’Engelbronner, J. G., and Mulder, M., “Towards an air traffic control complexity metric based on workspace constraints,”2010 IEEE International Conference on Systems, Man and Cybernetics, 2010, pp. 654–660. https://doi.org/10.1109/ICSMC.2010.5641823

work page arXiv 2010

Show all 41 references

[9]

A Spatial, Temporal Complexity Metric for Tactical Air Traffic Control,

Wee, H. J., Lye, S. W., and Pinheiro, J.-P., “A Spatial, Temporal Complexity Metric for Tactical Air Traffic Control,”Journal of Navigation, Vol. 71, No. 5, 2018, pp. 1040–1054. https://doi.org/10.1017/S0373463318000255

2018 doi
[10]

Probabilistic Air Traffic Complexity Analysis Considering Prediction Uncertainties in Traffic Scenarios,

Samardžić, K., Andraši, P., Radišić, T., and Novak, D., “Probabilistic Air Traffic Complexity Analysis Considering Prediction Uncertainties in Traffic Scenarios,”Aerospace, Vol. 11, No. 10, 2024. https://doi.org/10.3390/aerospace11100798

2024 doi
[11]

SpatiotemporalGraphIndicatorsforAirTrafficComplexityAnalysis,

Isufaj, R., Koca, T., andPiera, M.A., “SpatiotemporalGraphIndicatorsforAirTrafficComplexityAnalysis,”Aerospace, Vol.8, No. 12, 2021. https://doi.org/10.3390/aerospace8120364

2021 doi
[12]

Air Traffic Complexity Assessment Based on Ordered Deep Metric,

Chen, H., Zhang, L., Yuan, L., Zhu, W., and Liu, L., “Air Traffic Complexity Assessment Based on Ordered Deep Metric,” Aerospace, Vol. 9, No. 12, 2022. https://doi.org/10.3390/aerospace9120758

2022 doi
[13]

Enhancing air traffic complexity assessment through deep metric learning: A CNN-Based approach,

Chen, H., Zhou, Z., Wu, L., Fu, Y., and Xue, D., “Enhancing air traffic complexity assessment through deep metric learning: A CNN-Based approach,”Aerospace Science and Technology, Vol. 160, 2025. https://doi.org/10.1016/j.ast.2025.110090

2025
[14]

Gaze Analysis of Air Traffic Contoller Using AI-Based Conflict Detection,

Celina, V., Samardžić, K., Tukarić, I., Radišić, T., and Hermann, R. H., “Gaze Analysis of Air Traffic Contoller Using AI-Based Conflict Detection,”2024 AIAA DATC/IEEE 43rd Digital Avionics Systems Conference (DASC), 2024, pp. 1–10. https://doi.org/10.1109/DASC62030.2024.10749617

2024
[15]

The Impact of Automation on Air Traffic Controller’s Behaviors,

Wang, Y., Hu, R., Lin, S., Schultz, M., and Delahaye, D., “The Impact of Automation on Air Traffic Controller’s Behaviors,” Aerospace, Vol. 8, No. 9, 2021. https://doi.org/10.3390/aerospace8090260

2021 doi
[16]

AmachinelearningframeworkforpredictingATCconflictresolution strategiesforconformalautomation,

Guleria,Y.,Tran,P.,Pham,D.-T.,Durand,N.,andAlam,S.,“AmachinelearningframeworkforpredictingATCconflictresolution strategiesforconformalautomation,”11thSESARInnovationDays(SIDs2021),2021. URLhttps://hdl.handle.net/10356/154668

2021
[17]

Autonomous separation assurance in an high-density en route sector: A deep multi-agent reinforcement learning approach,

Brittain, M., and Wei, P., “Autonomous separation assurance in an high-density en route sector: A deep multi-agent reinforcement learning approach,”2019 IEEE intelligent transportation systems conference (ITSC), IEEE, 2019, pp. 3256–3262. https://doi.org/10.1109/ITSC.2019.8917217

2019
[18]

Scalable autonomous separation assurance with heterogeneous multi-agent reinforcement learning,

Brittain, M., and Wei, P., “Scalable autonomous separation assurance with heterogeneous multi-agent reinforcement learning,” IEEE Transactions on automation science and engineering, Vol. 19, No. 4, 2022, pp. 2837–2848. https://doi.org/10.1109/ TASE.2022.3151607

2022
[19]

Random Forests,

Breiman, L., “Random Forests,”Machine Learning, Vol. 45, No. 1, 2001, pp. 5–32. https://doi.org/10.1023/a:1010933404324

2001 doi
[20]

XGBoost: AScalableTreeBoostingSystem,

Chen,T.,andGuestrin,C.,“XGBoost: AScalableTreeBoostingSystem,”Proceedingsofthe22ndACMSIGKDDInternational Conference on Knowledge Discovery and Data Mining, Association for Computing Machinery, New York, NY, USA, 2016, pp. 785–794. https://doi.org/10.1145/2939672.2939785

2016
[21]

Agent Prioritization for Autonomous Navigation,

Refaat, K. S., Ding, K., Ponomareva, N., and Ross, S., “Agent Prioritization for Autonomous Navigation,”2019 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), 2019, pp. 2060–2067. https://doi.org/10.1109/IROS40897. 2019.8967743

2019
[22]

CausalAgents: A Robustness Benchmark for Motion Forecasting using Causal Relationships,

Roelofs, R., Sun, L., Caine, B., Refaat, K. S., Sapp, B., Ettinger, S., and Chai, W., “CausalAgents: A Robustness Benchmark for Motion Forecasting using Causal Relationships,”arXiv, 2022. https://doi.org/10.48550/arXiv.2207.03586

2022 doi arXiv
[23]

Introduction to Airspace,

NATS, “Introduction to Airspace,” , 2023. URL https://www.nats.aero/ae-home/introduction-to-airspace/

2023
[24]

CAP 413: Radiotelephony Manual - Civil Aviation Authority,

CAA, “CAP 413: Radiotelephony Manual - Civil Aviation Authority,” , 2021. URL https://www.caa.co.uk/our-work/ publications/documents/content/cap-413/

2021
[25]

BADA: An advanced aircraft performance model for present and future ATM systems,

Nuic, A., Poles, D., and Mouillet, V., “BADA: An advanced aircraft performance model for present and future ATM systems,”International journal of adaptive control and signal processing, Vol. 24, No. 10, 2010, pp. 850–866. https: //doi.org/10.1002/acs.1176. 18

2010 doi
[26]

Hofmann-Wellenhof, B., Lichtenegger, H., and Collins, J.,Reference systems, Springer Vienna, 2001, Chap. 3, pp. 25—-38. https://doi.org/10.1007/978-3-7091-6199-9_3

2001 doi
[27]

Graph neural networks: A review of methods and applications,

Zhou, J., Cui, G., Hu, S., Zhang, Z., Yang, C., Liu, Z., Wang, L., Li, C., and Sun, M., “Graph neural networks: A review of methods and applications,”AI Open, Vol. 1, 2020, pp. 57–81. https://doi.org/10.1016/j.aiopen.2021.01.001

2020 doi
[28]

How Attentive are Graph Attention Networks?

Brody, S., Alon, U., and Yahav, E., “How Attentive are Graph Attention Networks?”International Conference on Learning Representations, 2022. URL https://openreview.net/forum?id=F72ximsx7C1

2022
[29]

Paszke,A.,Gross,S.,Massa,F.,Lerer,A.,Bradbury,J.,Chanan,G.,Killeen,T.,Lin,Z.,Gimelshein,N.,Antiga,L.,Desmaison, A., Kopf, A., Yang, E., DeVito, Z., Raison, M., Tejani, A., Chilamkurthy, S., Steiner, B., Fang, L., Bai, J., and Chintala, S.,PyTorch: An Imperative Style, High-Per...

2019
[30]

Fast Graph Representation Learning with PyTorch Geometric,

Fey, M., and Lenssen, J. E., “Fast Graph Representation Learning with PyTorch Geometric,”ICLR Workshop on Representation Learning on Graphs and Manifolds, 2019. https://doi.org/10.48550/arXiv.1903.02428

2019 doi arXiv
[31]

DecoupledWeightDecayRegularization,

Loshchilov,I.,andHutter,F.,“DecoupledWeightDecayRegularization,”InternationalConferenceonLearningRepresentations,
[32]

RegressionQuantiles,

Koenker,R.,andBassett,G.,“RegressionQuantiles,”Econometrica,Vol.46,No.1,1978,p.33.https://doi.org/10.2307/1913643

1978 doi
[33]

Beyond Pinball Loss: Quantile Methods for Calibrated Uncertainty Quantification,

Chung, Y., Neiswanger, W., Char, I., and Schneider, J., “Beyond Pinball Loss: Quantile Methods for Calibrated Uncertainty Quantification,”Advances in Neural Information Processing Systems, Vol. 34, Curran Associates, Inc., 2021, pp. 10971–10984. URL https://proceedings.neurips...

2021
[34]

Regularization Strategies for Quantile Regression,

Narayan, T., Wang, S., Canini, K., and Gupta, M., “Regularization Strategies for Quantile Regression,” , 2021. https: //doi.org/10.48550/arXiv.2102.05135

2021 doi arXiv
[35]

A review of boosting methods for imbalanced data classification,

Li, Q., and Mao, Y., “A review of boosting methods for imbalanced data classification,”Pattern Analysis and Applications, Vol. 17, No. 4, 2014, pp. 679—-693. https://doi.org/10.1007/s10044-014-0392-8

2014 doi
[36]

Weighting Methods for Rare Event Identification From Imbalanced Datasets,

He, J., and Cheng, M. X., “Weighting Methods for Rare Event Identification From Imbalanced Datasets,”Frontiers in Big Data, Vol. 4, 2021. https://doi.org/10.3389/fdata.2021.715320

2021
[37]

Kuhn, M., and Johnson, K.,Over-Fitting and Model Tuning, Springer New York, 2013, Chap. 4, pp. 61—-92. https: //doi.org/10.1007/978-1-4614-6849-3_4

2013 doi
[38]

Kuhn, M., and Johnson, K.,Regression Trees and Rule-Based Models, Springer New York, 2013, Chap. 8, pp. 173––220. https://doi.org/10.1007/978-1-4614-6849-3_8

2013 doi
[39]

On Random Graphs I,

Erdös, P., and Rényi, A., “On Random Graphs I,”Publicationes Mathematicae Debrecen, Vol. 6, 1959, pp. 290–297

1959
[40]

NATS selects Altran Praxis to support major air traffic control system,

ATC Network, “NATS selects Altran Praxis to support major air traffic control system,” , Sep. 2010. URL https://www.atc- network.com/atc-news/nats-selects-altran-praxis-to-support-major-air-traffic-control-system. 19

2010
[2019]

URL https://openreview.net/forum?id=Bkg6RiCqY7

Pith tools

Reviewed August 6, 2026 · model on record in the stance chip above.

[1] [1]

Doc 4444 PANS-ATM,

International Civil Aviation Organization, “Doc 4444 PANS-ATM,” , 2007. URLhttps://www.ealts.com/documents/ ICAODoc4444AirTrafficManagement.pdf

work page 2007

[2] [2]

CAP 493: Manual of Air Traffic Services (MATS) Part 1,

UK Civil Aviation Authority, “CAP 493: Manual of Air Traffic Services (MATS) Part 1,” , 2023. URLhttps://www.caa. co.uk/our-work/publications/documents/content/cap-493/

work page 2023

[3] [3]

2022 European Aviation Environmental Report,

European Union Aviation Safety Agency (EASA), “2022 European Aviation Environmental Report,” , 2022. URLhttps: //www.easa.europa.eu/eco/eaer

work page 2022

[4] [4]

Automation in Air Traffic Management,

SESAR Joint Undertaking, “Automation in Air Traffic Management,” , 2020. URLhttps://www.sesarju.eu/sites/ default/files/documents/reports/automation%20atm%20vision%20roadmap.pdf

work page 2020

[5] [5]

Automation in Future Air Traffic Management: Effects of Decision Aid Reliability on Controller Performance and Mental Workload,

Metzger, U., and Parasuraman, R., “Automation in Future Air Traffic Management: Effects of Decision Aid Reliability on Controller Performance and Mental Workload,”Human Factors, Vol. 47, No. 1, 2005, pp. 35–49. https://doi.org/10.1518/ 0018720053653802. 17

work page 2005

[6] [6]

Determining Air Traffic Complexity – Challenges and Future Development,

Antulov-Fantulin, B., Juričić, B., Radišić, T., and Çetek, C., “Determining Air Traffic Complexity – Challenges and Future Development,”Promet - Traffic & Transportation, Vol. 32, No. 4, 2020, pp. 475–485. https://doi.org/10.7307/ptt.v32i4.3401

doi:10.7307/ptt.v32i4.3401 2020

[7] [7]

The Graph Neural Network Model,

Scarselli, F., Gori, M., Tsoi, A. C., Hagenbuchner, M., and Monfardini, G., “The Graph Neural Network Model,”IEEE Transactions on Neural Networks, Vol. 20, No. 1, 2009, pp. 61–80. https://doi.org/10.1109/TNN.2008.2005605

arXiv 2009

[8] [8]

Towards an air traffic control complexity metric based on workspace constraints,

van Paassen, M. M., d’Engelbronner, J. G., and Mulder, M., “Towards an air traffic control complexity metric based on workspace constraints,”2010 IEEE International Conference on Systems, Man and Cybernetics, 2010, pp. 654–660. https://doi.org/10.1109/ICSMC.2010.5641823

work page arXiv 2010