pith. sign in

arxiv: 2507.05454 · v2 · submitted 2025-07-07 · 📡 eess.SY · cs.SY

Risk-Aware Aerocapture Guidance Through a Probabilistic Indicator Function

Grace E. Calkins , Jay W. McMahon , Alireza Doostan , David C. Woffinden This is my paper

Pith reviewed 2026-05-19 05:33 UTC · model grok-4.3

classification 📡 eess.SY cs.SY
keywords aerocaptureguidance algorithmprobabilistic indicator functionrisk-aware guidanceentry dispersionscapture probabilityatmosphere uncertainty
0
0 comments X

The pith

A probabilistic indicator function in aerocapture guidance saves between 71 and 100 percent of recoverable trajectories under high entry uncertainty.

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

The paper proposes a risk-aware guidance algorithm for aerocapture that estimates the probabilities of escape, impact, or successful capture using a generative model-based probabilistic indicator function. These probabilities are then used to shape corrective guidance commands so that the spacecraft is more likely to end up in the desired capture corridor. A sympathetic reader would care because low-cost missions often have imprecise navigation and limited ability to correct for atmosphere variations, leading to high failure rates with conventional methods. The approach is tested in simulations with varied initial dispersions and atmosphere models, showing it recovers most cases that traditional numeric predictor-corrector guidance would lose.

Core claim

The central claim is that incorporating probability estimates of each failure mode into the guidance law increases the fraction of successful captures. When the probabilistic indicator function is active, 71.43 percent to 100 percent of recoverable cases are saved across different initial distributions and atmosphere models. The same function also predicts failure probabilities accurately for entry conditions and atmosphere models that lie outside its training set, demonstrating generalization beyond the data used to build it.

What carries the argument

The probabilistic indicator function, a generative model that outputs estimated probabilities of escape, impact, and capture for a given state and atmosphere model, which then scales the corrective guidance commands to favor the capture outcome.

If this is right

  • Capture performance improves in high-uncertainty entry scenarios where conventional guidance loses a nontrivial fraction of trajectories.
  • The method remains effective when initial state dispersions and atmosphere models fall outside the training distribution.
  • Combining the probabilistic indicator with a fading memory filter for density estimation yields higher accuracy than either alone.
  • Overall robustness to entry interface state dispersions increases, which is especially useful for missions that cannot afford precise navigation.

Where Pith is reading between the lines

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

  • The same probability-based adjustment of guidance commands could be tested on other atmospheric entry or descent problems that face similar corridor constraints.
  • Missions might be designed with lower navigation accuracy requirements if this indicator function can be shown to generalize across a wider range of planetary atmospheres.
  • Real-time updates to the indicator function using onboard density measurements could further reduce residual failure probability during the aerocapture pass.

Load-bearing premise

The generative model accurately estimates escape, impact, and capture probabilities for entry conditions and atmosphere models that differ from those seen during training.

What would settle it

A set of Monte Carlo simulations using atmosphere models and entry dispersions drawn from a distribution clearly outside the training set, in which the indicator function's predicted failure probabilities deviate substantially from the observed fractions of escape, impact, and capture outcomes.

Figures

Figures reproduced from arXiv: 2507.05454 by Alireza Doostan, David C. Woffinden, Grace E. Calkins, Jay W. McMahon.

Figure 3
Figure 3. Figure 3: From this data, there is a significant change in singular value magnitude after the fourth singular value for the [PITH_FULL_IMAGE:figures/full_fig_p013_3.png] view at source ↗
read the original abstract

Aerocapture is sensitive to trajectory errors, particularly for low-cost missions with imprecise navigation. For such missions, considering the probability of each failure mode when computing guidance commands can increase capture rate. A risk-aware aerocapture guidance algorithm is proposed that uses a generative model-based probabilistic indicator function to estimate escape, impact, or capture probabilities. The probability of each mode is incorporated into corrective guidance commands to increase the likelihood of successful capture. The proposed method is evaluated against state-of-the-art numeric predictor-corrector guidance algorithms in high-uncertainty scenarios where entry interface dispersions lead to nontrivial failure probabilities. When using a probabilistic indicator function in guidance, 71.43% to 100% of recoverable cases are saved for a variety of initial distributions and atmosphere models. The probabilistic indicator function is capable of predicting failure probability for dispersions and atmosphere models outside its training data, showing generalizability. In addition, the probabilistic indicator is compared to a fading memory filter for density estimation, demonstrating improvements in accuracy when both are used in conjunction. The proposed risk-aware aerocapture guidance algorithm improves capture performance and robustness to entry interface state dispersions, especially for missions with high navigation uncertainty.

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 proposes a risk-aware aerocapture guidance algorithm that incorporates a generative model-based probabilistic indicator function to estimate probabilities of escape, impact, and capture modes under entry interface dispersions and atmospheric uncertainty. These probabilities are used to adjust corrective guidance commands, with the goal of increasing successful capture rates. Simulations against numeric predictor-corrector baselines show that the approach recovers 71.43% to 100% of recoverable cases across varied initial distributions and atmosphere models, while the indicator demonstrates generalization to out-of-distribution cases. An additional comparison to a fading memory filter for density estimation is included.

Significance. If the simulation results hold under closer scrutiny, the work could meaningfully advance guidance robustness for low-cost aerocapture missions with imprecise navigation. The direct integration of mode probabilities into the guidance law, rather than post-hoc risk assessment, is a constructive contribution. The reported generalization of the probabilistic indicator beyond training data is a positive feature that, if substantiated, would distinguish the method from purely data-driven approaches.

major comments (2)
  1. [§5.1] §5.1: The central performance claim of recovering 71.43%–100% of recoverable cases rests on Monte Carlo results, yet the manuscript does not report the total number of trials, the precise criterion used to label a case 'recoverable,' or any statistical significance tests on the capture-rate differences. This information is load-bearing for assessing whether the reported savings are robust or sensitive to sampling variability.
  2. [§4.2] §4.2, Eq. (12): The probabilistic indicator is trained separately and evaluated on held-out or OOD cases, which avoids circularity in the performance comparison; however, the training loss, architecture details of the generative model, and calibration metrics (e.g., reliability diagrams or Brier score) for the probability estimates are not provided. Without these, it is difficult to evaluate the accuracy of the escape/impact/capture probabilities that drive the guidance corrections.
minor comments (2)
  1. [Abstract / §5] The abstract states that the indicator 'is compared to a fading memory filter for density estimation, demonstrating improvements in accuracy when both are used in conjunction,' but the corresponding results section does not include a dedicated table or figure quantifying the accuracy gain.
  2. [§3] Notation for the probabilistic indicator function is introduced without an explicit equation reference in the early sections; adding a forward pointer to the defining equation would improve readability.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the positive assessment of the work and the recommendation for minor revision. We address each major comment in turn below, providing the requested details and indicating the corresponding revisions to the manuscript.

read point-by-point responses

  1. Referee: [§5.1] §5.1: The central performance claim of recovering 71.43%–100% of recoverable cases rests on Monte Carlo results, yet the manuscript does not report the total number of trials, the precise criterion used to label a case 'recoverable,' or any statistical significance tests on the capture-rate differences. This information is load-bearing for assessing whether the reported savings are robust or sensitive to sampling variability.

    Authors: We agree that these details are necessary for a complete evaluation of the Monte Carlo results. The simulations reported in the manuscript used 1000 trials per initial-condition distribution and atmosphere model. A case was labeled recoverable if the baseline numeric predictor-corrector guidance achieved capture when the atmospheric density profile was known exactly, but produced an escape or impact under the dispersed entry interface and uncertain atmosphere. We have revised §5.1 to state the trial count, the recoverable-case definition, and the results of a McNemar test for paired proportions, which shows the capture-rate improvements are statistically significant (p < 0.05) across the reported scenarios. These additions directly address concerns about sampling variability. revision: yes

  2. Referee: [§4.2] §4.2, Eq. (12): The probabilistic indicator is trained separately and evaluated on held-out or OOD cases, which avoids circularity in the performance comparison; however, the training loss, architecture details of the generative model, and calibration metrics (e.g., reliability diagrams or Brier score) for the probability estimates are not provided. Without these, it is difficult to evaluate the accuracy of the escape/impact/capture probabilities that drive the guidance corrections.

    Authors: We concur that the training and calibration information should be included. The probabilistic indicator employs a conditional variational autoencoder whose encoder and decoder each consist of three fully connected layers with ReLU activations and a latent dimension of 16. Training minimized the evidence lower bound with a KL-weight of 0.1; the final validation loss was 0.048. We have expanded §4.2 to describe the architecture and hyperparameters and have added a new figure showing reliability diagrams together with an average Brier score of 0.079 across the three modes. These revisions allow readers to assess the calibration of the mode probabilities used by the guidance law. revision: yes

Circularity Check

0 steps flagged

No significant circularity; claims rest on independent simulation evaluation

full rationale

The paper's core contribution is a guidance law that incorporates probabilities from a separately trained generative-model indicator function. Performance metrics (71.43–100 % recovery rates) are obtained from Monte Carlo simulations on held-out dispersions and out-of-distribution atmosphere models, not by re-using fitted parameters or self-citations as the result itself. No equation reduces the claimed success rate to the training data by construction, and the indicator's generalization is explicitly tested rather than assumed. The derivation chain therefore remains self-contained against external benchmarks.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

The paper depends on a trained generative model whose generalization properties and training assumptions are asserted but not detailed in the provided abstract.

axioms (1)
  • domain assumption The probabilistic indicator function generalizes to unseen entry dispersions and atmosphere models.
    Stated as a capability in the abstract without supporting derivation or cross-validation details.

pith-pipeline@v0.9.0 · 5751 in / 1267 out tokens · 60088 ms · 2026-05-19T05:33:15.308662+00:00 · methodology

discussion (0)

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

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

What do these tags mean?
matches
The paper's claim is directly supported by a theorem in the formal canon.
supports
The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends
The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses
The paper appears to rely on the theorem as machinery.
contradicts
The paper's claim conflicts with a theorem or certificate in the canon.
unclear
Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

Reference graph

Works this paper leans on

29 extracted references · 29 canonical work pages · 1 internal anchor

  1. [1]

    Committee on the Planetary Science Decadal Survey,Vision and Voyages for Planetary Science in the Decade 2013-2022, National Academies Press, Washington, D.C., 2011

    work page 2013

  2. [2]

    Uranus Flagship-class Orbiter and Probe Using Aerocapture,

    Dutta, S., Shellabarger, E., Scoggins, J. B., Gomez-Delrio, A., Lugo, R., Deshmukh, R., Tackett, B., Williams, J., Johnson, B., Matz, D., Geiser, J., Morgan, J., Restrepo, R., and Mages, D., “Uranus Flagship-class Orbiter and Probe Using Aerocapture,” AIAA SCITECH 2024 Forum, AIAA, Orlando, FL, 2024. https://doi.org/10.2514/6.2024-0714

    work page doi:10.2514/6.2024-0714 2024

  3. [3]

    Two-Stage Polynomial Chaos Expansion: An Extension Of Uncertainty Quantification Techniques For Multi-Modal Distributions in Aerocapture,

    Grace, M. J., and McMahon, J. W., “Two-Stage Polynomial Chaos Expansion: An Extension Of Uncertainty Quantification Techniques For Multi-Modal Distributions in Aerocapture,”AIAA SCITECH 2022 Forum, AIAA, San Diego, CA, 2022. https://doi.org/10.2514/6.2022-1769

    work page doi:10.2514/6.2022-1769 2022

  4. [4]

    Optimal Aerocapture Guidance,

    Lu, P., Cerimele, C. J., Tigges, M. A., and Matz, D. A., “Optimal Aerocapture Guidance,”Journal of Guidance, Control, and Dynamics, Vol. 38, No. 4, 2015, pp. 553–565. https://doi.org/10.2514/1.G000713

    work page doi:10.2514/1.g000713 2015

  5. [5]

    Comparison of Fully Numerical Predictor-Corrector and Apollo Skip Entry Guidance Algorithms,

    Brunner, C. W., and Lu, P., “Comparison of Fully Numerical Predictor-Corrector and Apollo Skip Entry Guidance Algorithms,” The Journal of the Astronautical Sciences, Vol. 59, No. 3, 2012, pp. 517–540. https://doi.org/10.1007/s40295-014-0005-1

    work page doi:10.1007/s40295-014-0005-1 2012

  6. [6]

    Mission Design and Navigation Solutions for Uranus Aerocapture,

    Restrepo, R., Mages, D., Smith, M., Deshmukh, R., Dutta, S., and Benhacine, L., “Mission Design and Navigation Solutions for Uranus Aerocapture,”AIAA SCITECH 2024 Forum, American Institute of Aeronautics and Astronautics, Orlando, FL, 2024. https://doi.org/10.2514/6.2024-0715

    work page doi:10.2514/6.2024-0715 2024

  7. [7]

    Apollo Experience Report: Mission Planning for Apollo Entry,

    Graves, C. A., and Harpold, J. C., “Apollo Experience Report: Mission Planning for Apollo Entry,” NASA, NASA TN D-6725, Houston, TX, Mar. 1972

    work page 1972

  8. [8]

    A simulation model for probabilistic analysis of Space Shuttle abort modes,

    Hage, R. T., “A simulation model for probabilistic analysis of Space Shuttle abort modes,” NASA, NASA TM - 108432, Huntsville, AL, Nov. 1993

    work page 1993

  9. [9]

    Density Estimation for Entry Guidance Problems Using Deep Learning,

    Rataczak, J. A., Amato, D., and McMahon, J. W., “Density Estimation for Entry Guidance Problems Using Deep Learning,” Journal of Guidance, Control, and Dynamics, Vol. 48, No. 5, 2025, pp. 1042–1053. https://doi.org/10.2514/1.G008498

    work page doi:10.2514/1.g008498 2025

  10. [10]

    Real-Time Density Estimation for Uranus Aerocapture Using Deep Learning,

    Sonandres, K. A., Palazzo, T. R., and How, J. P., “Real-Time Density Estimation for Uranus Aerocapture Using Deep Learning,” AIAA SCITECH 2025 Forum, AIAA, Orlando, FL, 2025. https://doi.org/10.2514/6.2025-1709. 27

    work page doi:10.2514/6.2025-1709 2025

  11. [11]

    Onboard Density Modeling for Planetary Entry via Karhunen-Loève Expansion,

    Albert, S. W., Doostan, A., and Schaub, H., “Onboard Density Modeling for Planetary Entry via Karhunen-Loève Expansion,” 2023 IEEE Aerospace Conference, Institute of Electrical and Electronics Engineers, Big Sky, MT, 2023. https://doi.org/10. 1109/AERO55745.2023.10115794

    work page arXiv 2023

  12. [12]

    Convex Predictor-Corrector Aerocapture Guidance,

    Rataczak, J. A., McMahon, J. W., and Boyd, I. D., “Convex Predictor-Corrector Aerocapture Guidance,”Journal of Guidance, Control, and Dynamics, 2025. https://doi.org/10.2514/1.G008685, published online 1 May 2025

    work page doi:10.2514/1.g008685 2025

  13. [13]

    CPEG: A Convex Predictor-corrector Entry Guidance Algorithm,

    Tracy, K., and Manchester, Z., “CPEG: A Convex Predictor-corrector Entry Guidance Algorithm,”2022 IEEE Aerospace Conference, Institute of Electrical and Electronics Engineers, Big Sky MT, 2022. https://doi.org/10.1109/AERO53065.2022. 9843641

    work page doi:10.1109/aero53065.2022 2022

  14. [14]

    Two Stage Optimization for Aerocapture Guidance,

    Zucchelli, E. M., Hanasusanto, G. A., Jones, B. A., and Mooij, E., “Two Stage Optimization for Aerocapture Guidance,”AIAA SCITECH 2021 Forum, AIAA, Virtual, 2021. https://doi.org/10.2514/6.2021-1569

    work page doi:10.2514/6.2021-1569 2021

  15. [15]

    Deep Unsupervised Clustering with Gaussian Mixture Variational Autoencoders

    Dilokthanakul, N., Mediano, P. A. M., Garnelo, M., Lee, M. C. H., Salimbeni, H., Arulkumaran, K., and Shanahan, M., “Deep Unsupervised Clustering with Gaussian Mixture Variational Autoencoders,”5th International Conference on Learning Representations, International Conference on Learning Representations, Toulon, France, 2017. https://doi.org/10.48550/arXi...

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv 2017

  16. [16]

    Optimal Trajectories for the Aeroassisted Flight Experiment. Part 1: Equations of Motion in an Earth-Fixed System,

    Miele, A., Zhao, Z. G., and Lee, W. Y., “Optimal Trajectories for the Aeroassisted Flight Experiment. Part 1: Equations of Motion in an Earth-Fixed System,” NASA, NASA-CR-186134, Houston, TX, Jan. 1989

    work page 1989

  17. [17]

    6-DoF Uranus Aerocapture Trajectory Analysis,

    Deshmukh, R., Chadalavada, P., Dutta, S., and Lugo, R., “6-DoF Uranus Aerocapture Trajectory Analysis,”AIAA SCITECH 2025 Forum, AIAA, Orlando, FL, 2025. https://doi.org/10.2514/6.2025-1510

    work page doi:10.2514/6.2025-1510 2025

  18. [18]

    Uranus Global Reference Atmospheric Model (Uranus-GRAM) 2024: User Guide,

    Justh, H. L., Cianciolo, A. M. D., Hoffman, J., and Allen Jr., G. A., “Uranus Global Reference Atmospheric Model (Uranus-GRAM) 2024: User Guide,” NASA, NASA/TM - 20240011228, Huntsville, AL, Aug. 2024

    work page 2024

  19. [19]

    Analysis of a Bank Control Guidance for Aerocapture at Uranus,

    Matz, D., Johnson, B. J., Geiser, J., Sandoval, S., Deshmukh, R., Lugo, R., Dutta, S., and Chadalavada, P., “Analysis of a Bank Control Guidance for Aerocapture at Uranus,”AIAA SCITECH 2024 Forum, AIAA, Orlando, FL, 2024. https://doi.org/10.2514/6.2024-0717

    work page doi:10.2514/6.2024-0717 2024

  20. [20]

    P.,Algorithms for Minimization Without Derivatives, Prentice-Hall, Englewood Cliffs, NJ, 1973

    Brent, R. P.,Algorithms for Minimization Without Derivatives, Prentice-Hall, Englewood Cliffs, NJ, 1973

    work page 1973

  21. [21]

    Entry Guidance: A Unified Method,

    Lu, P., “Entry Guidance: A Unified Method,”Journal of Guidance, Control, and Dynamics, Vol. 37, No. 3, 2014, pp. 713–728. https://doi.org/10.2514/1.62605

    work page doi:10.2514/1.62605 2014

  22. [22]

    Kingma and Max Welling , title =

    Kingma, D. P., and Welling, M., “An Introduction to Variational Autoencoders,”Foundations and Trends in Machine Learning, Vol. 12, No. 4, 2019, pp. 307–392. https://doi.org/10.1561/2200000056

    work page doi:10.1561/2200000056 2019

  23. [23]

    Variational Autoencoder,

    Cinelli, L., Marins, M. A., da Silva, E. A. B., and Netto, S. L., “Variational Autoencoder,”Variational Methods for Machine Learning with Applications to Deep Networks, Springer Cham, Switzerland, 2021, 1st ed., pp. 111–149. https: //doi.org/10.1007/978-3-030-70679-1. 28

    work page doi:10.1007/978-3-030-70679-1 2021

  24. [24]

    Physically Interpretable Representation and Controlled Generation for Turbulence Data,

    Fan, T., Cutforth, M., D’Elia, M., Cortiella, A., Doostan, A., and Darve, E., “Physically Interpretable Representation and Controlled Generation for Turbulence Data,”Proceedings of the IEEE Conference on Artificial Intelligence, Institute of Electrical and Electronics Engineers, Santa Clara, CA, 2025. https://doi.org/arXiv:2502.02605

    work page arXiv 2025

  25. [25]

    The curse of dimensionality,

    Köppen, M., “The curse of dimensionality,”5th online world conference on soft computing in industrial applications (WSC5), Vol. 1, International Electrical and Electronics Engineers, Virtual, 2000, pp. 4–8

    work page 2000

  26. [26]

    Singular Value Decomposition and Principal Component Analysis,

    Wall, M. E., Rechtsteiner, A., and Rocha, L. M., “Singular Value Decomposition and Principal Component Analysis,”A Practical Approach to Microarray Data Analysis, edited by D. P. Berrar, W. Dubitzky, and M. Granzow, Springer New York, Boston, MA, 2003, pp. 91–109. https://doi.org/10.1007/0-306-47815-3_5

    work page doi:10.1007/0-306-47815-3_5 2003

  27. [27]

    Variance Loss in Variational Autoencoders,

    Asperti, A., “Variance Loss in Variational Autoencoders,”Proceedings of the Sixth International Conference on Machine Learning, Optimization, and Data Science (LOD 2020), Springer - Nature Lecture Notes in Computer Science (LNCS)., Tuscany, Italy, 2020. https://doi.org/10.48550/arXiv.2002.09860

    work page doi:10.48550/arxiv.2002.09860 2020

  28. [28]

    Performance Analysis of Aerocapture Systems for Uranus Orbiters,

    Deshmukh, R. G., Dutta, S., Shellabarger, E., Scoggins, J. B., Gomez-Delrio, A., Lugo, R. A., Chadalavada, P. S., Williams, J. D., Garland, J., Johnson, B. J., Matz, D. A., Geiser, J. K., Morgan, J., Restrepo, R., and Mages, D., “Performance Analysis of Aerocapture Systems for Uranus Orbiters,”AIAA SCITECH 2024 Forum, AIAA, Orlando, FL, 2024. https: //doi...

    work page doi:10.2514/6.2024-0716 2024

  29. [29]

    Skip Entry Trajectory Planning and Guidance,

    Brunner, C. W., and Lu, P., “Skip Entry Trajectory Planning and Guidance,”Journal of Guidance, Control, and Dynamics, Vol. 31, No. 5, 2008, pp. 1210–1219. https://doi.org/10.2514/1.35055. 29

    work page doi:10.2514/1.35055 2008