pith. sign in

arxiv: 2604.17063 · v1 · submitted 2026-04-18 · 💻 cs.DC

Predictive Sectorization and Bayesian Optimized Consensus for Admission Control in Autonomous Airspace Operations

Aditya Dhodapkar , Avery Smidt , Aaron Verkleeren , Stacy Patterson , Carlos A. Varela This is my paper

Pith reviewed 2026-05-10 06:32 UTC · model grok-4.3

classification 💻 cs.DC
keywords airspace sectorizationautonomous air traffic controlleaderless PaxosBayesian optimizationXGBoost classificationFAA SWIM replaysadmission controldistributed coordination
0
0 comments X

The pith

A three-stage pipeline predicts optimal airspace sectors with machine learning, coordinates entries via consensus, and tunes parameters with optimization to scale autonomous air traffic control.

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

Conventional sector-based air traffic control creates scaling bottlenecks as flight volumes rise because fixed regions increase controller workload and handoff complexity. This paper develops an automated three-stage system that first uses an XGBoost model to select the best 3D grid layout from traffic features, then applies a leaderless Paxos protocol for aircraft to negotiate sector access, and finally employs Bayesian optimization to customize eight protocol settings for each airport. The approach reaches 91 percent accuracy on sector prediction and above 96 percent success on entries while keeping near-midair collision rates low, all tested on tens of thousands of FAA replay samples. A sympathetic reader would care because the method keeps human oversight while aiming to let capacity grow with demand rather than with the number of controllers. The core contribution is showing that location-agnostic features and per-environment tuning can make such coordination practical.

Core claim

The paper establishes that a pipeline of predictive sectorization, leaderless Paxos admission control, and Bayesian optimization produces a workable system for autonomous airspace operations. An XGBoost classifier maps 23 traffic features to the best 3D grid configuration at 91.38 percent accuracy on 65,000 FAA System Wide Information Management replays. Aircraft then use Paxos to agree on sector entries, sustaining above 96 percent success with low collision rates across configurations. Bayesian optimization with a Gaussian Process surrogate tunes the eight protocol parameters for each airport in 50 trials and demonstrates that qualitatively different settings are required for different air

What carries the argument

The three-stage pipeline that first classifies optimal 3D sector grids with XGBoost, then runs leaderless Paxos for distributed entry coordination, and finally applies Gaussian Process Bayesian optimization to tune protocol parameters per airport.

If this is right

  • Optimal 3D sector grids can be predicted reliably from location-agnostic traffic features alone.
  • Aircraft can coordinate sector entries autonomously using Paxos while preserving high success rates and low collision risk.
  • Eight protocol parameters must be tuned separately for each airport rather than using a single global setting.
  • The combination of prediction, consensus, and optimization allows human oversight to remain while supporting higher traffic volumes.
  • The pipeline provides a concrete mechanism for admission control that adapts to varying traffic environments.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • If replay results hold in live settings, controller workload per flight could drop because sector boundaries and coordination shift to automated processes.
  • The finding that each airport needs its own optimized configuration implies that deployment would require ongoing per-site calibration rather than a one-time setup.
  • The approach could be tested first in high-fidelity simulators that inject the kinds of sensor noise and communication delays absent from replay data.
  • Similar predictive-plus-consensus pipelines might apply to other distributed resource allocation problems such as vehicle routing or network slicing.

Load-bearing premise

That performance measured on historical FAA replay data will carry over to live operations without creating coordination failures or safety problems absent from the dataset.

What would settle it

A controlled live simulation or operational trial in which entry success falls below 90 percent or near-midair collision rates rise measurably above the replay results.

Figures

Figures reproduced from arXiv: 2604.17063 by Aaron Verkleeren, Aditya Dhodapkar, Avery Smidt, Carlos A. Varela, Stacy Patterson.

Figure 1
Figure 1. Figure 1: Pipeline overview. Our contributions are: 1) A two stage XGBoost classifier achieving 91.38% ac￾curacy on 25 class sectorization prediction using 23 arXiv:2604.17063v1 [cs.DC] 18 Apr 2026 [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Conflict resolution breakdown for JFK real world [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: NMACs vs. aircraft count for JFK real world traffic. [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗
Figure 6
Figure 6. Figure 6: BO convergence trajectory for LAX (31 aircraft). [PITH_FULL_IMAGE:figures/full_fig_p005_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: BO convergence trajectory for DFW (60 aircraft). [PITH_FULL_IMAGE:figures/full_fig_p005_7.png] view at source ↗
read the original abstract

Conventional air traffic control divides airspace into specific regions, creating a scaling bottleneck as traffic grows. Choosing how to partition airspace is not straightforward because grid size affects workload, handoff frequency, and the capacity of whatever coordination mechanism operates within each sector. We present a three stage pipeline that automates sectorization and sector coordination while preserving human oversight. First, a two stage XGBoost classifier predicts the optimal 3D grid configuration from 23 location-agnostic traffic features, achieving 91.38% accuracy on a 65,000 sample dataset derived from Federal Aviation Administration System Wide Information Management replays. Second, a leaderless Paxos consensus protocol lets aircraft coordinate sector entries among themselves, maintaining above 96% entry success with low near mid-air collision rates across all tested configurations. Third, Bayesian Optimization with a Gaussian Process surrogate tunes eight protocol parameters per airport in 50 trials, revealing that each traffic environment requires a qualitatively different configuration. The resulting pipeline offers a practical path toward scalable, autonomous airspace management as traffic demand outpaces controller capacity.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 2 minor

Summary. The manuscript presents a three-stage pipeline for automating airspace sectorization and coordination while preserving human oversight. A two-stage XGBoost classifier predicts optimal 3D grid configurations from 23 location-agnostic traffic features, achieving 91.38% accuracy on a 65,000-sample dataset from FAA SWIM replays. A leaderless Paxos protocol enables aircraft to coordinate sector entries with >96% success and low NMAC rates. Bayesian optimization with a Gaussian process surrogate then tunes eight protocol parameters per airport over 50 trials. The authors conclude that the pipeline offers a practical path toward scalable autonomous airspace management as traffic demand grows.

Significance. The integration of established ML and distributed-systems components for a safety-critical application is a reasonable engineering contribution. Reproducible use of XGBoost, Paxos, and Gaussian-process BO is a strength that allows the work to focus on the domain-specific pipeline rather than inventing new primitives. If the reported metrics were shown to generalize beyond historical replays, the approach could meaningfully address controller scaling limits; however, the current evidence base leaves that generalization unproven.

major comments (2)
  1. [Abstract] Abstract: the 91.38% accuracy and >96% entry-success figures are reported without error bars, baseline comparisons (e.g., against static grid or rule-based sectorization), or any description of train/test split, cross-validation, or hold-out procedure on the 65,000 samples; these omissions make it impossible to judge whether the central performance claims are statistically reliable or merely overfit to the replay distribution.
  2. [Abstract] Abstract and evaluation narrative: the entire pipeline is validated exclusively on historical FAA SWIM replays; no experiments or analysis address robustness to real-time effects (variable latency, sensor noise, dynamic join/leave, or traffic patterns absent from the 65k dataset), which directly undermines the claim of a 'practical path' to live autonomous operations.
minor comments (2)
  1. [Abstract] The two-stage structure of the XGBoost classifier is mentioned but never illustrated or pseudocoded; a diagram or explicit feature-to-stage mapping would improve clarity.
  2. [Abstract] The statement that 'each traffic environment requires a qualitatively different configuration' is asserted after 50 BO trials but is not supported by any quantitative comparison of the resulting parameter sets across airports.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive feedback on the statistical presentation of results and the scope of our validation. We address each major comment below and will revise the manuscript to improve clarity and temper claims where appropriate.

read point-by-point responses

  1. Referee: [Abstract] Abstract: the 91.38% accuracy and >96% entry-success figures are reported without error bars, baseline comparisons (e.g., against static grid or rule-based sectorization), or any description of train/test split, cross-validation, or hold-out procedure on the 65,000 samples; these omissions make it impossible to judge whether the central performance claims are statistically reliable or merely overfit to the replay distribution.

    Authors: We agree that the abstract omits key details needed to evaluate the reliability of the reported figures. In the revised version we will expand the abstract to report error bars (standard deviation across folds), explicitly describe the train/test split and cross-validation procedure used on the 65,000 samples, and include brief baseline comparisons against static grid and rule-based sectorization. These additions will be drawn from the evaluation section and will not alter the underlying results. revision: yes

  2. Referee: [Abstract] Abstract and evaluation narrative: the entire pipeline is validated exclusively on historical FAA SWIM replays; no experiments or analysis address robustness to real-time effects (variable latency, sensor noise, dynamic join/leave, or traffic patterns absent from the 65k dataset), which directly undermines the claim of a 'practical path' to live autonomous operations.

    Authors: The evaluation is indeed performed on historical replays, and the simulations do not explicitly model sensor noise, variable network latency, or traffic patterns outside the dataset. The Paxos component does handle dynamic aircraft arrivals and departures within the replay scenarios. We will revise the abstract and conclusion to remove or qualify the phrase 'practical path to live autonomous operations,' instead framing the work as a simulation study on historical data that demonstrates the feasibility of the integrated pipeline. Real-time robustness testing is noted as future work. revision: partial

Circularity Check

0 steps flagged

No significant circularity; pipeline uses independent external algorithms

full rationale

The derivation applies standard XGBoost classification to FAA SWIM replay features for sectorization prediction, deploys the established leaderless Paxos protocol for coordination, and employs off-the-shelf Gaussian-process Bayesian optimization for per-airport parameter tuning. Reported metrics (91.38% accuracy, >96% entry success) are empirical outcomes on the 65k-sample dataset rather than quantities forced by construction from the inputs or from self-citations. No equations reduce predictions to fitted parameters by definition, no uniqueness theorems are imported from the authors' prior work, and no ansatz is smuggled via self-citation. Minor self-citation of external algorithms, if present, is not load-bearing for the central claims.

Axiom & Free-Parameter Ledger

1 free parameters · 2 axioms · 0 invented entities

The central claim rests on the empirical performance of standard ML and consensus components plus the assumption that simulation results generalize; no new physical entities are postulated.

free parameters (1)
  • eight protocol parameters
    Tuned separately per airport via 50-trial Bayesian optimization with Gaussian process surrogate.
axioms (2)
  • standard math Leaderless Paxos maintains agreement and safety properties under the assumed network and failure model
    Invoked for the coordination stage without re-derivation.
  • domain assumption XGBoost classifier trained on the 23 traffic features will generalize to unseen traffic patterns
    Required for the 91.38% accuracy to support the practical-path claim.

pith-pipeline@v0.9.0 · 5496 in / 1440 out tokens · 48456 ms · 2026-05-10T06:32:42.179907+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

23 extracted references · 23 canonical work pages

  1. [1]

    FAA faces controller staffing challenges as air traffic operations return to pre- pandemic levels at critical facilities,

    U.S. Department of Transportation, Office of Inspector General, “FAA faces controller staffing challenges as air traffic operations return to pre- pandemic levels at critical facilities,” DOT OIG, Tech. Rep. A V2023035, Jun. 2023

    work page 2023

  2. [2]

    Air traffic controller workforce plan, FY 2025–FY 2028,

    Federal Aviation Administration, “Air traffic controller workforce plan, FY 2025–FY 2028,” FAA, Tech. Rep., 2024

    work page 2025

  3. [3]

    Network operations report 2023,

    EUROCONTROL, “Network operations report 2023,” EUROCON- TROL, Tech. Rep., 2024

    work page 2023

  4. [4]

    Dynamic density: An air traffic management metric,

    I. V . Laudeman, S. G. Shelden, R. Branstrom, and C. R. Brasil, “Dynamic density: An air traffic management metric,” NASA Ames Research Center, NASA Tech. Memorandum NASA TM-1998-112226, 1998

    work page 1998

  5. [5]

    Cognitive complexity in air traffic control: A literature review,

    EUROCONTROL Experimental Centre, “Cognitive complexity in air traffic control: A literature review,” EUROCONTROL, Tech. Rep. EEC Note 04/03, 2003

    work page 2003

  6. [6]

    Initial concepts for dynamic airspace configuration,

    P. Kopardekar, K. Bilimoria, and B. Sridhar, “Initial concepts for dynamic airspace configuration,” inAIAA Aviation Technol., Integr ., Oper . Conf. (ATIO), 2007

    work page 2007

  7. [7]

    Automatic airspace sectorisation: A survey,

    P. Flener and J. Pearson, “Automatic airspace sectorisation: A survey,” Knowl. Eng. Rev., vol. 28, no. 3, pp. 293–314, 2013

    work page 2013

  8. [8]

    Conflict-aware flight planning for avoiding near mid-air collisions,

    S. Paul, S. Patterson, and C. A. Varela, “Conflict-aware flight planning for avoiding near mid-air collisions,” in38th AIAA/IEEE Digit. Avionics Syst. Conf. (DASC), San Diego, CA, USA, 2019

    work page 2019

  9. [9]

    Collaborative situational awareness for conflict-aware flight plan- ning,

    ——, “Collaborative situational awareness for conflict-aware flight plan- ning,” in2020 AIAA/IEEE 39th Digit. Avionics Syst. Conf. (DASC), San Antonio, TX, USA, 2020, pp. 1–10

    work page 2020

  10. [10]

    Eventual consensus in synod: Verification using a failure-aware actor model,

    S. Paul, G. A. Agha, S. Patterson, and C. A. Varela, “Eventual consensus in synod: Verification using a failure-aware actor model,”Innov. Syst. Softw. Eng., 2022

    work page 2022

  11. [11]

    Formal verification of timely knowledge propagation in airborne networks,

    S. Paul, C. McCarthy, S. Patterson, and C. Varela, “Formal verification of timely knowledge propagation in airborne networks,”Sci. Comput. Program., vol. 239, 2025

    work page 2025

  12. [12]

    Taking the human out of the loop: A review of Bayesian optimization,

    B. Shahriari, K. Swersky, Z. Wang, R. P. Adams, and N. de Freitas, “Taking the human out of the loop: A review of Bayesian optimization,” Proc. IEEE, vol. 104, no. 1, pp. 148–175, 2016

    work page 2016

  13. [13]

    Practical Bayesian optimiza- tion of machine learning algorithms,

    J. Snoek, H. Larochelle, and R. P. Adams, “Practical Bayesian optimiza- tion of machine learning algorithms,” inAdv. Neural Inf. Process. Syst., vol. 25, 2012

    work page 2012

  14. [14]

    Formal verification of safety-critical aerospace systems,

    S. Paulet al., “Formal verification of safety-critical aerospace systems,” IEEE Aerosp. Electron. Syst. Mag., vol. 38, no. 5, pp. 72–88, 2023

    work page 2023

  15. [15]

    Predictive sectorization and Bayesian optimized consen- sus for admission control in autonomous airspace operations,

    A. Dhodapkar, “Predictive sectorization and Bayesian optimized consen- sus for admission control in autonomous airspace operations,” Master’s thesis, Rensselaer Polytechnic Institute, Troy, NY , USA, 2026, available: https://wcl.cs.rpi.edu/theses/dhodaa thesis.pdf

    work page 2026

  16. [16]

    System wide information manage- ment (SWIM) — overview,

    Federal Aviation Administration, “System wide information manage- ment (SWIM) — overview,” https://www.faa.gov/air traffic/technology/ swim, 2025

    work page 2025

  17. [17]

    SMOTE: Synthetic minority over-sampling technique,

    N. V . Chawla, K. W. Bowyer, L. O. Hall, and W. P. Kegelmeyer, “SMOTE: Synthetic minority over-sampling technique,”J. Artif. Intell. Res., vol. 16, pp. 321–357, 2002

    work page 2002

  18. [18]

    XGBoost: A scalable tree boosting system,

    T. Chen and C. Guestrin, “XGBoost: A scalable tree boosting system,” inProc. 22nd ACM SIGKDD Int. Conf. Knowl. Discovery Data Mining, 2016, pp. 785–794

    work page 2016

  19. [19]

    The part-time parliament,

    L. Lamport, “The part-time parliament,”ACM Trans. Comput. Syst., vol. 16, no. 2, pp. 133–169, 1998

    work page 1998

  20. [20]

    C. E. Rasmussen and C. K. I. Williams,Gaussian Processes for Machine Learning. MIT Press, 2006

    work page 2006

  21. [21]

    Efficient global optimiza- tion of expensive black-box functions,

    D. R. Jones, M. Schonlau, and W. J. Welch, “Efficient global optimiza- tion of expensive black-box functions,”J. Global Optim., vol. 13, no. 4, pp. 455–492, 1998

    work page 1998

  22. [22]

    A comparison of three methods for selecting values of input variables in the analysis of output from a computer code,

    M. D. McKay, R. J. Beckman, and W. J. Conover, “A comparison of three methods for selecting values of input variables in the analysis of output from a computer code,”Technometrics, vol. 21, no. 2, pp. 239– 245, 1979

    work page 1979

  23. [23]

    Decentralized air traffic control (DATC) — source code,

    A. Dhodapkar, A. Smidt, A. Verkleeren, S. Patterson, and C. Varela, “Decentralized air traffic control (DATC) — source code,” https://github. com/Aditya-Dhodapkar/Decentralized-Air-Traffic-Control---DATC, 2025. 6

    work page 2025