Pith. sign in

REVIEW 4 major objections 5 minor 2 cited by

Global Search of Optimal Spacecraft Trajectories using Amortization and Deep Generative Models

T0 review · 4 major / 5 minor · reviewed 2026-08-10 · deepseek-v4-flash

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

Pith's one-line read The paper argues that a generative model trained on previously solved low-thrust transfers can supply initial guesses that solve the same family of problems at new thrust levels much faster than uniform multi-start.

desk verdict A credible amortized-warm-start demonstration on low-thrust CR3BP with an honest ablation, but the 'global search' claim is relative to the solver's discovered basins and the benchmark lacks a strong global-optimization baseline. read the letter →

arxiv 2412.20023 v1 pith:VMRBBPSZ submitted 2024-12-28 math.OC cs.LGcs.SYeess.SY

Ryne Beeson , Anjian Li , Amlan Sinha This is my paper · ORCID
classification math.OCcs.LGcs.SYeess.SY MSC 49M3790C2668T07
keywords amortizedglobalsearchconditionalvariationalautoencoderGaussianmixturepriorlow-thrusttrajectoryoptimizationcircularrestrictedthree-bodyproblemfunnelstructurebasinsofattractionmultistart
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

Preliminary spacecraft trajectory design is a parameter-dependent global search: as mission parameters change, designers must re-find a diverse set of high-quality local optima, and gradient-based solvers need good initial guesses. This paper tries to make that re-search cheap by learning, from solutions at one set of parameters, a conditional distribution over initial guesses for a new parameter value. The claim is that samples from this learned distribution land in the neighborhoods where the solver converges quickly, so an unseen thrust level can be searched drastically faster than with uniform multi-start. On an Earth-Moon low-thrust transfer, the full method raises convergence from 28% to 62.5% at thrust scale 0.15, and from 42% to 74% at 0.85, while cutting median solver time by factors of about 2.6 and 5. A sympathetic reader would care because this is a concrete path to amortizing expensive global search across a family of trajectory design problems.

What carries the argument

The load-bearing object is the $k$-local neighborhood $k\mathcal{N}_{\alpha,\beta}=\{z\in U_h \mid \pi_\gamma^k(z)\in A_{\alpha,\beta}\}$, the set of initial guesses that a numerical solver maps to a high-quality local optimum within $k$ iterations; the target distribution weights each optimum by the measure of its neighborhood, and an amortized generative model approximates a smoothed version of this weighted Dirac mixture. The conditional distribution is learned with a CVAE whose prior is a Gaussian mixture conditioned on $\alpha$ (20 components for the main problem), used for the four scalar variables of shooting time, initial and final coast times, and final mass, plus an LSTM that generates the 20 time-correlated thrust control vectors conditioned on those variables and $\alpha$. The division of labor is the mechanism: the CVAE captures the global funnel and hyperplane structure, while the LSTM captures temporal correlation in the controls, and the ablation study shows that neither component alone delivers the full speedup.

What would settle it

Run the same warm-start benchmark at a thrust scale far from the training grid, say $\alpha=0.25$ (the grid contains 0.2 and 0.3 but no 0.25), and compare 4,000 learned initial guesses against 4,000 uniform guesses using an exhaustive reference set built from 100,000 uniform solves. If the learned distribution does not put measurably more mass in the $k$-local neighborhoods of the high-quality optima, or if it misses a funnel that the exhaustive search finds, the amortization claim would be disproved for that parameter region.

Watch

Extended reading notes

Core claim

The central discovery is that the parameterized global search problem can be reformulated as sampling a conditional distribution $p(\cdot|\alpha)$ whose support approximates the $k$-local neighborhoods of the high-quality extremum set $A_{\alpha,\beta}$, and that a conditional variational autoencoder with a Gaussian-mixture prior, paired with an LSTM for the temporally correlated thrust controls, learns this distribution well enough to generalize to thrust parameters not in the training set. In the benchmark problem, the learned distribution predicts the hyperplane structure in the time-of-flight coordinates and the modes of the final-mass distribution, and its samples, when used as initial guesses for the numerical solver, converge more often and faster than uniform sampling, uniform controls with learned times, or a vanilla CVAE. The paper also documents the multi-modal funnel structure of the low-thrust circular restricted three-body problem, showing clusters of basins arranged in hyperplanes with multiple funnels.

Load-bearing premise

The speedup depends on the training data being representative: the 25,000 uniform-random solves per thrust level must already cover every region of initial guesses worth sampling, and the layout of those regions must change smoothly enough between the 12 training thrust levels that a model trained on them can interpolate to an unseen level.

Editorial extensions

If this is right

  • At thrust scale $\alpha=0.15$, 62.5% of 200 initial guesses from the full model converge within the solver budget, versus 28% for uniform multi-start; at $\alpha=0.85$ the figures are 74% versus 42%.
  • Median solver time for converged cases drops from 169.31 s to 64.14 s at $\alpha=0.15$ and from 121.48 s to 24.32 s at $\alpha=0.85$; only the full framework produced converged solutions in under 10 seconds.
  • With solver time budgets set to the method's own medians (64 s and 24 s), roughly a third of 4,000 samples converge, making the global search about ten times quicker than the naive uniform approach.
  • The learned model preserves solution diversity: converged samples at unseen thrust levels reproduce the hyperplane distribution and qualitative variety of the ground-truth reference set $A_{\alpha,\beta}$.
  • Ablation results imply that predicting the time and mass variables with the CVAE must be combined with LSTM-generated controls; replacing either with uniform sampling erodes most of the speedup.

Reading between the lines

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

  • Beyond the paper: if the basin topology varies smoothly with $\alpha$, the same amortization should transfer to other continuous mission parameters such as time of flight, final mass constraint, or the three-body mass ratio, with data cost growing with the effective dimension of the funnel structure.
  • Beyond the paper: the learned distribution could be paired with a level-2 basin-hopping step to traverse adjacent funnels, since the paper explicitly leaves intermediate-level algorithms as future work; a testable extension is whether hopping from learned samples finds the funnel global minimum faster than hopping from uniform samples.
  • Beyond the paper: the training-data coverage premise can be tested directly by comparing the support of the learned distribution against an extremely large uniform reference solve at a held-out $\alpha$; if the learned model assigns low mass to a funnel that the uniform solve discovers, the acceleration claim would not extend to that region.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

4 major / 5 minor

Summary. This paper introduces AmorGS, an amortized global search framework for parameterized spacecraft trajectory optimization. The authors formalize the learning target as a conditional distribution over k-local neighborhoods of high-quality local extrema A_{\alpha,\beta} (Eq. 7 and Eq. 14), represent this distribution with a CVAE with GMM prior for the time and mass variables plus an LSTM for the temporally correlated control sequence, and validate on De Jong's fifth function and a minimum-fuel low-thrust Earth-Moon CR3BP transfer. For the two held-out thrust parameters alpha=0.15 and alpha=0.85, the full framework raises the fraction of initial guesses that converge within the solver budget relative to uniform multi-start (Table 6: 62.5% vs 28% at alpha=0.15; 74% vs 42% at alpha=0.85) and reduces median solver time. The paper also reports an ablation study isolating the CVAE and LSTM contributions and provides an analysis of hyperplane and funnel structure of the solution set.

Significance. The central speedup claim is evaluated on alpha values not used in training, and the ablation design is informative; if the results are robust, the method is a practical warm-starting tool for preliminary low-thrust trajectory design. The 300,000-solution dataset and the funnel-structure analysis of the CR3BP problem are substantial contributions. I read the 'global search' claim as referring to the set of local extrema discoverable by the uniform multi-start pipeline used to define A_{\alpha,\beta}, not to coverage of the full solution set of the optimal control problem; this scope distinction, together with the lack of uncertainty quantification in the headline success rates, is the main reason the manuscript needs revision.

major comments (4)
  1. [Sec. V.B.1 and Eq. (7)] The reference set A_{\alpha,\beta} and the training data are both generated by the same uniform multi-start from which SNOPT converges within 500 s or 1,000 iterations. The learned conditional distribution can therefore place mass only on funnels that this pipeline already discovered; if uniform sampling misses entire disconnected basins, no learned model can recover them. The manuscript itself defines A_{\alpha,\beta} solver-dependently, so this is not a circularity in the derivation, but it is a scope limitation that conflicts with the title and with phrases such as 'global search' coverage. The success-rate and diversity evidence in Secs. V.B.7 and V.B.8 is relative to this discovered subset. Please either explicitly scope the claims to 'global search relative to the uniform multi-start/SNOPT pipeline' or provide independent coverage evidence, for example by comparing against monotonic basin hopping or a substantially larger uniform sample on the held-out alpha cases.
  2. [Sec. V.B.7, Table 6] The headline success rates are based on 200 samples per method and are reported without confidence intervals or standard errors. Differences such as 28% versus 29.5% for the alpha=0.15 case are well within sampling noise, and even the larger differences should carry binomial confidence intervals or repeated-seed trials. The follow-up experiment with 4,000 samples in Fig. 22 uses time limits equal to the AmorGS medians and therefore does not directly quantify the uncertainty of the Table 6 percentages; please report confidence intervals or standard errors for the Table 6 metrics.
  3. [Sec. V.B.1 and Sec. V.B.7] The held-out evaluation is interpolation-only: alpha=0.15 is bracketed by training values 0.13 and 0.16, and alpha=0.85 by 0.8 and 0.9. This is a genuine out-of-sample test in that these alpha values are not in the training set, but generalization to values outside the trained interval [0.1, 1] or to a coarser training grid is not demonstrated. Add at least one extrapolation case, or state explicitly in the abstract and conclusions that the method is validated for interpolation within the trained parameter range, which is the honest scope of the current experiments.
  4. [Sec. V.B.7, ablation study] The only non-learned baseline is uniform multi-start, which is also the distribution used to collect the training data. A simple continuation baseline, e.g., warm-starting the alpha=0.15 case with samples or solutions from the nearest training alphas 0.13 and 0.16, would isolate the benefit of the learned amortized distribution from the benefit of knowing that nearby alphas have similar solution structure. Without such a baseline, part of the observed speedup may reflect neighbor information inherited from the training data rather than the generative model itself. Please add this inexpensive baseline or restrict the claim to a comparison against uniform multi-start.
minor comments (5)
  1. [Sec. III.B.7] The time-of-flight value appears as both 38.146 TU and 38.156 TU; the numbers should be made consistent.
  2. [Sec. V.B.1] The phrase 'T o thoroughly exploit' contains a spacing typo and should be corrected.
  3. [References] Reference [53] contains a malformed DOI ('arXiv.org.2410.02976') and should be corrected.
  4. [Sec. V.B.1 and Sec. V.B.2] The mass threshold beta=415 kg and the number of GMM components K=20 are selected from the same data used in the evaluation; a short sensitivity discussion with respect to these hyperparameters would help the reader assess how robust the Table 6 results are to these choices.
  5. [Sec. V.B.6 and Fig. 20] For alpha=0.85 the LSTM predictions include higher time-of-flight modes that are absent from the ground truth; a brief explanation of whether these are spurious modes or low-probability sampling artifacts would clarify the quality of the learned distribution.

Circularity Check

0 steps flagged · score 2.0 of 10

No circular derivation; the held-out-alpha evaluation is a genuine out-of-sample test, and the minor self-referential benchmark design does not reduce any prediction to its training inputs.

full rationale

The paper's claimed derivation chain is not circular. The conditional distribution p(·|alpha) is learned from solved instances A_{alpha,beta} at 12 training values of alpha (Eq. 39) and evaluated at alpha=0.15 and 0.85, which are not in that training set. The success metric—convergence of the numerical solver to A_{alpha,beta} within 500 s or 1000 iterations (Table 6)—is an external quantity computed by SNOPT, not a re-statement of the training objective. The threshold beta=415 kg is fixed from inspection of alpha=1.0 data and applied uniformly across cases; it does not encode the held-out alpha structure. The GMM component count K=20 is selected from the funnel analysis of training data, but the model must still learn the locations of hyperplanes and mass/control distributions at held-out alpha, and Figs. 17-20 show genuine mismatch (e.g., the alpha=0.85 case predicts low-probability high-time-of-flight modes not present in ground truth). The only self-citations (refs. [48], [53]-[55]) are pointers to prior applications of the same framework; the CVAE/GMM/LSTM architecture is fully specified in Tables 3-5 and the loss is derived in Eq. (37), so no load-bearing step reduces to an unverified self-citation. The benchmark is self-relative in that ground-truth A_{alpha,beta} is generated by the same uniform multi-start pipeline that produced the training data; if that pipeline systematically misses entire funnel regions, the learned model cannot recover them. That is a coverage limitation of the 'global search' claim, not a circular reduction: the held-out alpha values are genuine out-of-sample inputs, and the speedup is measured against the same solver and transcription. The score of 2 reflects only the minor self-referential benchmark design and the presence of non-load-bearing self-citations, not any equation-level reduction of a prediction to its inputs.

Assumptions & free parameters 6 free parameters · 6 assumptions · 0 invented entities

The central claim rests on the standard mathematics of optimal control and variational inference, plus three domain assumptions that are not proven in the paper: the basin topology varies continuously enough with alpha to interpolate from the 12 training values; uniform multi-start with 25,000 solves per alpha adequately covers the high-quality basins; and the fixed transcription and solver define a meaningful ground truth. The free parameters (beta, K, eta_L, latent dimension, N, delta) are chosen from data or by hand and directly shape both the learned distribution and the evaluation. No new physical entities are introduced.

free parameters (6)
  • mass threshold beta = 415 kg
    Defines the high-quality set A_{alpha,beta} in Eq. (7); filters training data and defines successful convergence in the benchmark.
  • number of GMM components K = 20
    Chosen to match the number of observed time-of-flight modes (Sec III.B.3); the prior in Eq. (34) depends on it.
  • ELBO weighting eta_L = 1e-4
    Hand-set in Eq. (37) to rebalance reconstruction against KL divergence; affects the learned latent representation.
  • latent dimension = 4
    Latent variables z are 4-dimensional to match the four CV AE output variables (tau_s, tau_i, tau_f, m_f).
  • number of control segments N = 20
    Transcription parameter in Eq. (28) that sets the dimension of the control decision variables and the temporal resolution.
  • hyperplane tolerance delta = 0.25 TU
    Used in Eq. (32) to define membership in a hyperplane mode for the funnel analysis.
assumptions (6)
  • standard math Standard optimal control and NLP theory: KKT conditions, existence of solutions, and smoothness of problem data in Sec II.A.
    Invoked throughout Sec II to define A_alpha and solver mappings.
  • domain assumption The parameterized basin topology varies continuously with alpha, so interpolation from discrete training alphas generalizes to unseen alpha.
    This is the premise of amortization; introduced in Sec II.E and relied on in Sec V.B for alpha 0.15 and 0.85.
  • domain assumption Uniform multi-start with 25,000 samples per alpha provides sufficient coverage of the high-quality basins A_{alpha,beta}.
    Training data in Sec V.B.1 are generated this way; undetected basins cannot be learned.
  • domain assumption The forward-backward shooting transcription with N=20 segments and SNOPT solver settings defines the ground-truth solution set.
    All claims about A_{alpha,beta} are relative to this transcription and solver (Sec III.B.2, Sec V.B.1).
  • ad hoc to paper Euclidean distance on U_h is an acceptable proxy for the designer's implicit diversity metric.
    Stated in Sec II.D.2 as a convenient surrogate without validation.
  • standard math Variational inference theory, including the ELBO and reparameterization trick, is correct.
    Used in Sec IV.C to derive the CVAE loss function.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Global Search of Optimal Spacecraft Trajectories using Amortization and Deep Generative Models." pith.science (2026). https://pith.science/paper/VMRBBPSZ

@misc{pith2026241220023,
  author       = {Pith},
  title        = {Pith review of: Global Search of Optimal Spacecraft Trajectories using Amortization and Deep Generative Models},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/VMRBBPSZ}},
  note         = {Machine review of arXiv:2412.20023}
}
read the original abstract

Preliminary spacecraft trajectory optimization is a parameter dependent global search problem that aims to provide a set of solutions that are of high quality and diverse. In the case of numerical solution, it is dependent on the original optimal control problem, the choice of a control transcription, and the behavior of a gradient based numerical solver. In this paper we formulate the parameterized global search problem as the task of sampling a conditional probability distribution with support on the neighborhoods of local basins of attraction to the high quality solutions. The conditional distribution is learned and represented using deep generative models that allow for prediction of how the local basins change as parameters vary. The approach is benchmarked on a low thrust spacecraft trajectory optimization problem in the circular restricted three-body problem, showing significant speed-up over a simple multi-start method and vanilla machine learning approaches. The paper also provides an in-depth analysis of the multi-modal funnel structure of a low-thrust spacecraft trajectory optimization problem.

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 2 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

  1. Initial Guess Generation for Low-Thrust Trajectory Design with Robustness to Missed-Thrust-Events

    math.OC 2025-01 conditional novelty 5.0 of 10

    Warm-starting robust low-thrust trajectory optimizers with solutions to earlier non-robust problems improves feasibility and solution quality, but cumulative gains are mixed when seed-generation costs are included.

  2. Complexity Bounds and Approaches to Learning Projected Gradient Descent Solver Iterates

    cs.LG 2026-07 conditional novelty 4.0 of 10

    For one-sided box-constrained quadratics, keeping the last k+1 projected-gradient iterates yields a Rademacher-complexity generalization bound controlled by the contraction factor and neighborhood radius.

Reference graph

Works this paper leans on

64 extracted references · 31 canonical work pages · cited by 2 Pith papers

  1. [1]

    Global Optimization by Basin-Hopping and the Lowest Energy Structures of Lennard-Jones Clusters Containing up to 110 Atoms,

    Wales, D. J., and Doye, J. P . K., “Global Optimization by Basin-Hopping and the Lowest Energy Structures of Lennard-Jones Clusters Containing up to 110 Atoms,” The Journal of Physical Chemistry A , Vol. 101, No. 28, 1997, pp. 5111–5116. https://doi.org/10.1021/jp970984n, URL https://doi.org/10.1021/jp970984n

    doi:10.1021/jp970984n 1997

  2. [2]

    Global Optimization on Funneling Landscapes,

    Leary, R. H., “Global Optimization on Funneling Landscapes,” Journal of Global Optimization , Vol. 18, 2000, pp. 367–383. https://doi.org/10.1023/A:1026500301312

    doi:10.1023/a:1026500301312 2000

  3. [3]

    Exploring Energy Landscapes,

    Wales, D. J., “Exploring Energy Landscapes,” Annual Review of Physical Chemistry , Vol. 69, No. 1, 2018, pp. 401–425. https://doi.org/10.1146/annurev-physchem-050317-021219, URL https://doi.org/10.1146/annurev-physchem-050317-021219, pMID: 29677468

    doi:10.1146/annurev-physchem-050317-021219 2018

  4. [4]

    https://doi.org/10.1137/1.9781611972672, URL https://epubs.siam.org/doi/abs/10.1137/1.9781611972672

    Locatelli, M., and Schoen, F., Global Optimization, Society for Industrial and Applied Mathematics, Philadelphia, PA, 2013. https://doi.org/10.1137/1.9781611972672, URL https://epubs.siam.org/doi/abs/10.1137/1.9781611972672

    doi:10.1137/1.9781611972672 2013

  5. [5]

    Optimal Interplanetary Spacecraft Trajectories via a Pareto Genetic Algorithm,

    Hartmann, J. W., Coverstone-Carroll, V . L., and Williams, S. N., “Optimal Interplanetary Spacecraft Trajectories via a Pareto Genetic Algorithm,” The Journal of the Astronautical Sciences , Vol. 46, No. 3, 1998, pp. 267–282. https: //doi.org/10.1007/BF03546237, URL https://doi.org/10.1007/BF03546237

    doi:10.1007/bf03546237 1998

  6. [6]

    Near-optimal low-thrust orbit transfers generated by a genetic algorithm,

    Rauwolf, G. A., and Coverstone-Carroll, V . L., “Near-optimal low-thrust orbit transfers generated by a genetic algorithm,” Journal of Spacecraft and Rockets , Vol. 33, No. 6, 1996, pp. 859–862. https://doi.org/10.2514/3.26850, URL https: //doi.org/10.2514/3.26850

    work page doi:10.2514/3.26850 1996

  7. [7]

    Primer Vector Theory Applied to Global Low-Thrust Trade Studies,

    Russell, R. P ., “Primer Vector Theory Applied to Global Low-Thrust Trade Studies,” Journal of Guidance, Control, and Dynamics, Vol. 30, No. 2, 2007, pp. 460–472. https://doi.org/10.2514/1.22984, URL https://doi.org/10.2514/1.22984

    doi:10.2514/1.22984 2007

  8. [8]

    Lawden, D., Optimal Trajectories for Space Navigation, Butterworths, London, 1963

    work page 1963

Show all 64 references
  1. [9]

    Global search for low-thrust transfers to the Moon in the planar circular restricted three-body problem,

    Oshima, K., Campagnola, S., and Y anao, T., “Global search for low-thrust transfers to the Moon in the planar circular restricted three-body problem,” Celestial Mechanics and Dynamical Astronomy , Vol. 128, No. 2, 2017, pp. 303–322. https://doi.org/10.1007/s10569-016-9748-2, U...

    2017 doi

  2. [10]

    Advanced Global Optimisation Tools for Mission Analysis and Design,

    Lizia, P . D., and Radice, G., “Advanced Global Optimisation Tools for Mission Analysis and Design,” FInal Report AO4532/18139/04/NL/MV , European Space Agency, 11 2004. URL https://www.esa.int/gsp/ACT/doc/ARI/ARI%20Study% 20Report/ACT-RPT-MAD-ARI-03-4101b-GlobalOptimisation-G...

    2004

  3. [11]

    Advanced Global Optimisation Tools for Mission Analysis and Design,

    Myatt, D. R., Becerra, V . M., Nasuto, S. J., and Bishop, J. M., “Advanced Global Optimisation Tools for Mission Analysis and Design,” Final Report AO4532/18138.04/04/NL/MV , European Space Agency, 11 2004. URL https://www.esa.int/gsp/ACT/ doc/ARI/ARI%20Study%20Report/ACT-RPT-...

    2004

  4. [12]

    Preliminary Design of Multiple Gravity-Assist Trajectories,

    Vasile, M., and Pascale, P . D., “Preliminary Design of Multiple Gravity-Assist Trajectories,” Journal of Spacecraft and Rockets , Vol. 43, No. 4, 2006, pp. 794–805. https://doi.org/10.2514/1.17413, URL https://doi.org/10.2514/1.17413

    2006 doi

  5. [13]

    Analysis of Some Global Optimization Algorithms for Space Trajectory Design,

    Vasile, M., Minisci, E., and Locatelli, M., “Analysis of Some Global Optimization Algorithms for Space Trajectory Design,” Journal of Spacecraft and Rockets , Vol. 47, No. 2, 2010, pp. 334–344. https://doi.org/10.2514/1.45742, URL https://doi.org/10.2514/1.45742

    2010 doi

  6. [14]

    Genetic algorithms, path relinking, and the flowshop sequencing problem,

    Reeves, C. R., and Y amada, T., “Genetic algorithms, path relinking, and the flowshop sequencing problem,” Evol. Comput., Vol. 6, No. 1, 1998, pp. 45–60. https://doi.org/10.1162/evco.1998.6.1.45, URL https://doi.org/10.1162/evco.1998.6.1.45

    1998 doi

  7. [15]

    A global optimization method for the design of space trajectories,

    Addis, B., Cassioli, A., Locatelli, M., and Schoen, F., “A global optimization method for the design of space trajectories,” Computational Optimization and Applications , Vol. 48, No. 3, 2011, pp. 635–652. https://doi.org/10.1007/s10589-009-9261-6, URL https://doi.org/10.1007/...

    2011 doi

  8. [16]

    Global Trajectory Optimisation Problems Database,

    European Space Agency Advanced Concepts Team, “Global Trajectory Optimisation Problems Database,” , 2024. URL https://www.esa.int/gsp/ACT/projects/gtop/

    2024

  9. [17]

    T uning Monotonic Basin Hopping: Improving the Efficiency of Stochastic Search as Applied to Low-Thrust Trajectory Optimization,

    Englander, J. A., and Englander, A. C., “T uning Monotonic Basin Hopping: Improving the Efficiency of Stochastic Search as Applied to Low-Thrust Trajectory Optimization,” 24th International Symposium on Space Flight Dynamics , Laurel, Maryland,

  10. [18]

    Preliminary Design of Low-Thrust Interplanetary Missions,

    Sims, J. A., and Flanagan, S. N., “Preliminary Design of Low-Thrust Interplanetary Missions,” AAS/AIAA Astrodynamics Specialist Conference, Girdwood, Alaska, 1999

    1999

  11. [19]

    Walking the Filament of Feasibility,

    Englander, A. C., and Englander, J. A., “Walking the Filament of Feasibility,” AIAA/AAS Astrodynamics Specialist Meeting,

  12. [20]

    Hopping with an Adaptive Hop Probability Distribution,

    Englander, A., Englander, J., and Carter, M., “Hopping with an Adaptive Hop Probability Distribution,”AAS/AIAA Astrodynamics Specialist Conference, South Lake Tahoe, CA, 2020

    2020

  13. [21]

    Machine Learning Techniques for Approximation of Objective Functions in Trajectory Optimisation,

    Ampatzis, C., and Izzo, D., “Machine Learning Techniques for Approximation of Objective Functions in Trajectory Optimisation,” IJCAI-09 Workshop on Artificial Intelligence, 2009

    2009

  14. [22]

    Machine learning for global optimization,

    Cassioli, A., Di Lorenzo, D., Locatelli, M., Schoen, F., and Sciandrone, M., “Machine learning for global optimization,” Computational Optimization and Applications , Vol. 51, No. 1, 2012, pp. 279–303. https://doi.org/10.1007/s10589-010-9330-x, URL https://doi.org/10.1007/s105...

    2012 doi

  15. [23]

    Fast Evaluation of Low-Thrust Transfers via Multilayer Perceptions,

    Zhu, Y .-h., and Luo, Y .-Z., “Fast Evaluation of Low-Thrust Transfers via Multilayer Perceptions,”Journal of Guidance, Control, and Dynamics, Vol. 42, No. 12, 2019, pp. 2627–2637. https://doi.org/10.2514/1.G004080, URL https://doi.org/10.2514/1. G004080

    2019 doi

  16. [24]

    Global Optimization Requires Global Information,

    Stephens, C. P ., and Baritompa, W., “Global Optimization Requires Global Information,” Journal of Optimization The- ory and Applications , Vol. 96, No. 3, 1998, pp. 575–588. https://doi.org/https://doi.org/10.1023/A:1022612511618, URL https://login.ezproxy.princeton.edu/login...

    1998

  17. [25]

    Schoen, F., Two-Phase Methods for Global Optimization, Springer US, Boston, MA, 2002, pp. 151–177. https://doi.org/10. 1007/978-1-4757-5362-2_5, URL https://doi.org/10.1007/978-1-4757-5362-2_5

    2002 doi

  18. [26]

    E., and Ho, Y .-C., Applied Optimal Control: Optimization, Estimation, and Control , Hemisphere Publication Corporation, 1975

    Bryson, A. E., and Ho, Y .-C., Applied Optimal Control: Optimization, Estimation, and Control , Hemisphere Publication Corporation, 1975

    1975

  19. [27]

    https://doi.org/10.2307/j.ctvcm4g0s, URL https://press.princeton.edu/books/hardcover/9780691151878/calculus-of-variations- and-optimal-control-theory

    Liberzon, D., Calculus of Variations and Optimal Control Theory: A Concise Introduction , Princeton University Press, 2012. https://doi.org/10.2307/j.ctvcm4g0s, URL https://press.princeton.edu/books/hardcover/9780691151878/calculus-of-variations- and-optimal-control-theory

    2012

  20. [29]

    Direct Trajectory Optimization Using Nonlinear Programming and Collocation,

    Hargraves, C., and Paris, S. W ., “Direct Trajectory Optimization Using Nonlinear Programming and Collocation,” Journal of Guidance, Control, and Dynamics , Vol. 10, No. 4, 1987, pp. 338–342. https://doi.org/10.2514/3.20223

    1987 doi

  21. [30]

    Discrete Approximation to Optimal Trajectories Using Direct Transcription and Nonlinear Programming,

    Enright, P . J., and Conway, B. A., “Discrete Approximation to Optimal Trajectories Using Direct Transcription and Nonlinear Programming,” Journal of Guidance, Control, and Dynamics , Vol. 15, No. 4, 1992, pp. 994–1002. https: //doi.org/10.2514/3.20934

    1992 doi

  22. [31]

    Survey of Numerical Methods for Trajectory Optimization,

    Betts, J. T., “Survey of Numerical Methods for Trajectory Optimization,” Journal of Guidance, Control, and Dynamics , Vol. 21, No. 2, 1998, pp. 193–207. https://doi.org/10.2514/2.4231, URL https://doi.org/10.2514/2.4231

    1998 doi

  23. [32]

    Legendre pseudospectral approximations of optimal control problems,

    Ross, M., and Fahroo, F., “Legendre pseudospectral approximations of optimal control problems,” , 2003. https://doi.org/10. 1007/978-3-540-45056-6_21, URL http://hdl.handle.net/10945/66382

    2003

  24. [33]

    T., Practical Methods for Optimal Control and Estimation Using Nonlinear Programming, Second Edition ,2 nd ed., Society for Industrial and Applied Mathematics, 2010

    Betts, J. T., Practical Methods for Optimal Control and Estimation Using Nonlinear Programming, Second Edition ,2 nd ed., Society for Industrial and Applied Mathematics, 2010. https://doi.org/10.1137/1.9780898718577, URL https://epubs.siam.org/ doi/abs/10.1137/1.9780898718577

    2010 doi

  25. [34]

    A., and Paris, S

    Conway, B. A., and Paris, S. W ., Spacecraft Trajectory Optimization Using Direct Transcription and Nonlinear Programming, Cambridge Aerospace Series, Cambridge University Press, 2010, pp. 37–78. https://doi.org/10.1017/CBO9780511778025. 44

    2010 doi

  26. [35]

    Minima of Functions of Several Variables with Inequalities as Side Constraints,

    Karush, W ., “Minima of Functions of Several Variables with Inequalities as Side Constraints,” Master’s thesis, University of Chicago, Chicago, Illinois, 1939

    1939

  27. [36]

    L., Sullivan, F

    Peressini, A. L., Sullivan, F. E., and J. J. Uhl, J., The Mathematics of Nonlinear Programming , Undergraduate Texts in Mathematics, Springer, 1988

    1988

  28. [37]

    A global optimization algorithm,

    Becker, R., and Lago, G., “A global optimization algorithm,” Proceedings of the 8th Allerton Conference on Circuits and Systems Theory, Monticello, Illinois, 1970, pp. 3–12

    1970

  29. [38]

    Cluster Analysis Using Seed Points and Density-Determined Hyperspheres as an Aid to Global Optimization,

    Torn, A. A., “Cluster Analysis Using Seed Points and Density-Determined Hyperspheres as an Aid to Global Optimization,”IEEE Transactions on Systems, Man, and Cybernetics, Vol. 7, No. 8, 1977, pp. 610–616. https://doi.org/10.1109/TSMC.1977.4309787

    1977

  30. [39]

    Stochastic global optimization methods part I: Clustering methods,

    Rinnooy Kan, A. H. G., and Timmer, G. T., “Stochastic global optimization methods part I: Clustering methods,” Mathematical Programming, Vol. 39, No. 1, 1987, pp. 27–56. https://doi.org/10.1007/BF02592070, URL https://doi.org/10.1007/BF02592070

    1987 doi

  31. [40]

    Stochastic global optimization methods part II: Multi level methods,

    Rinnooy Kan, A. H. G., and Timmer, G. T., “Stochastic global optimization methods part II: Multi level methods,” Mathematical Programming, Vol. 39, No. 1, 1987, pp. 57–78. https://doi.org/10.1007/BF02592071, URL https://doi.org/10.1007/BF02592071

    1987 doi

  32. [41]

    Simple linkage: Analysis of a threshold-accepting global optimization method,

    Locatelli, M., and Schoen, F., “Simple linkage: Analysis of a threshold-accepting global optimization method,”Journal of Global Optimization, Vol. 9, No. 1, 1996, pp. 95–111. https://doi.org/10.1007/BF00121752, URL https://doi.org/10.1007/BF00121752

    1996 doi

  33. [42]

    Random Linkage: a family of acceptance/rejection algorithms for global optimisation,

    Locatelli, M., and Schoen, F., “Random Linkage: a family of acceptance/rejection algorithms for global optimisation,” Mathematical Programming , Vol. 85, No. 2, 1999, pp. 379–396. https://doi.org/10.1007/s101070050062, URL https: //doi.org/10.1007/s101070050062

    1999 doi

  34. [43]

    Test functions for optimization needs,

    Molga, M., and Smutnicki, C., “Test functions for optimization needs,” , 2005. URL https://marksmannet.com/RobertMarks/ Classes/ENGR5358/Papers/functions.pdf, access Oct. 18, 2024

    2005

  35. [44]

    Test Functions for Multimodal Search Techniques,

    Shekel, J., “Test Functions for Multimodal Search Techniques,” Fifth Annual Princeton Conference on Information Science and Systems, 1971

    1971

  36. [45]

    Dynamically Leveraged Automated Multibody (N) Trajectory Optimization,

    Beeson, R., Sinha, A., Jagannatha, B., Bunce, D., and Carroll, D., “Dynamically Leveraged Automated Multibody (N) Trajectory Optimization,” AAS/AIAA Space Flight Mechanics Conference , American Astronautical Society, Charlotte, NC, 2022

    2022

  37. [46]

    SNOPT: An SQP Algorithm for Large-Scale Constrained Optimization,

    Gill, P . E., Murray, W., and Saunders, M. A., “SNOPT: An SQP Algorithm for Large-Scale Constrained Optimization,” SIAM Review, Vol. 47, No. 1, 2005, pp. 99–131. https://doi.org/10.1137/S0036144504446096, URL https://doi.org/10.1137/ S0036144504446096

    2005 doi

  38. [47]

    Matplotlib: A 2D graphics environment,

    Hunter, J. D., “Matplotlib: A 2D graphics environment,” Computing in Science & Engineering , Vol. 9, No. 3, 2007, pp. 90–95. https://doi.org/10.1109/MCSE.2007.55

    2007 doi

  39. [48]

    Amortized Global Search for Efficient Preliminary Trajectory Design with Deep Generative Models,

    Li, A., Sinha, A., and Beeson, R., “Amortized Global Search for Efficient Preliminary Trajectory Design with Deep Generative Models,” AAS/AIAA Astrodynamics Specialist Conference, Big Sky, Montana, 2023. URL http://arxiv.org/abs/2308.03960. 45

    2023 arXiv

  40. [49]

    Generative Adversarial Nets,

    Goodfellow, I. J., Pouget-Abadie, J., Mirza, M., Xu, B., Warde-Farley, D., Ozair, S., Courville, A., and Bengio, Y ., “Generative Adversarial Nets,”Proceedings of the International Conference on Neural Information Processing Systems , 2014, pp. 2672–2680. https://doi.org/10.48...

    2014 doi arXiv

  41. [50]

    Deep Unsupervised Learning using Nonequilibrium Thermodynamics,

    Sohl-Dickstein, J., Weiss, E., Maheswaranathan, N., and Ganguli, S., “Deep Unsupervised Learning using Nonequilibrium Thermodynamics,” Proceedings of the 32nd International Conference on Machine Learning , Vol. 37, 2015, pp. 2256–2265. https://doi.org/10.48550/arXiv.1503.03585...

    2015 doi arXiv

  42. [51]

    Score-Based Generative Modeling through Stochastic Differential Equations,

    Song, Y ., Sohl-Dickstein, J., Kingma, D. P ., Kumar, A., Ermon, S., and Poole, B., “Score-Based Generative Modeling through Stochastic Differential Equations,” , 2020. https://doi.org/10.48550/arXiv.2011.13456, URL https://arxiv.org/abs/2011.13456

    2020 doi arXiv

  43. [52]

    Denoising Diffusion Probabilistic Models,

    Ho, J., Jain, A., and Abbeel, P ., “Denoising Diffusion Probabilistic Models,” 34th Conference on Neural Information Processing Systems (NeurIPS 2020) , Vol. 34, Vancouver, Canada, 2020. https://doi.org/10.48550/arXiv.2006.11239, URL https://proceedings.neurips.cc/paper/2020/fil...

    2020 doi arXiv

  44. [53]

    Learning Optimal Control and Dynamical Structure of Global Trajectory Search Problems with Diffusion Models,

    Gräbner, J., Li, A., Sinha, A., and Beeson, R., “Learning Optimal Control and Dynamical Structure of Global Trajectory Search Problems with Diffusion Models,” Submitted, 10 2024. https://doi.org/10.48550/arXiv.org.2410.02976, URL https: //arxiv.org/abs/2410.02976

    2024 doi arXiv

  45. [54]

    Semi-Supervised Global Search Capabilities for Optimal Low Thrust Spacecraft Trajectories over an Indirect Approach,

    Gräbner, J., and Beeson, R., “Semi-Supervised Global Search Capabilities for Optimal Low Thrust Spacecraft Trajectories over an Indirect Approach,” 35th AAS/AIAA Space Flight Mechanics Meeting , 2025

    2025

  46. [55]

    Efficient and Guaranteed-Safe Non-Convex Trajectory Optimization with Constrained Diffusion Model,

    Li, A., Ding, Z., Dieng, A. B., and Beeson, R., “Efficient and Guaranteed-Safe Non-Convex Trajectory Optimization with Constrained Diffusion Model,” ICLR 2024 Workshop GenAI4DM , 2024. https://doi.org/10.48550/arXiv.2403.05571, URL https://arxiv.org/abs/2403.05571v1

    2024 doi arXiv

  47. [56]

    Variational Deep Embedding: An Unsupervised and Generative Approach to Clustering,

    Jiang, Z., Zheng, Y ., Tan, H., Tang, B., and Zhou, H., “Variational Deep Embedding: An Unsupervised and Generative Approach to Clustering,” , 2017. https://doi.org/10.48550/arXiv.1611.05148, URL https://arxiv.org/abs/1611.05148v3

    2017 doi arXiv

  48. [57]

    SCALE method for single-cell ATAC-seq analysis via latent feature extraction,

    Xiong, L., Xu, K., Tian, K., Shao, Y ., Tang, L., Gao, G., Zhang, M., Jiang, T., and Zhang, Q. C., “SCALE method for single-cell ATAC-seq analysis via latent feature extraction,” Nature Communications, Vol. 10, No. 1, 2019, p. 4576. https://doi.org/10.1038/s41467-019-12630-7, ...

    2019 doi

  49. [58]

    Learning Structured Output Representation using Deep Conditional Generative Models,

    Sohn, K., Lee, H., and Y an, X., “Learning Structured Output Representation using Deep Conditional Generative Models,” Advances in Neural Information Processing Systems , Vol. 28, edited by C. Cortes, N. Lawrence, D. Lee, M. Sugiyama, and R. Garnett, Curran Associates, Inc., 2...

    2015

  50. [59]

    An Introduction to Variational Autoencoders,

    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, URL http://dx.doi.org/10.1561/2200000056. 46

    2019 doi

  51. [60]

    Long Short-Term Memory,

    Hochreiter, S., and Schmidhuber, J., “Long Short-Term Memory,” Neural Computation, Vol. 9, No. 8, 1997, pp. 1735–1780. https://doi.org/10.1162/neco.1997.9.8.1735, URL https://doi.org/10.1162/neco.1997.9.8.1735

    1997 doi

  52. [61]

    Framewise phoneme classification with bidirectional LSTM and other neural network architectures,

    Graves, A., and Schmidhuber, J., “Framewise phoneme classification with bidirectional LSTM and other neural network architectures,”Neural Networks, Vol. 18, No. 5, 2005, pp. 602–610. https://doi.org/https://doi.org/10.1016/j.neunet.2005.06.042, URL https://www.sciencedirect.com...

    2005 doi

  53. [62]

    Automatic differentiation in PyTorch,

    Paszke, A., Gross, S., Chintala, S., Chanan, G., Y ang, E., DeVito, Z., Lin, Z., Desmaison, A., Antiga, L., and Lerer, A., “Automatic differentiation in PyTorch,” NIPS 2017 Workshop Autodiff , Long Beach, California, 2017. URL https: //openreview.net/forum?id=BJJsrmfCZ

    2017

  54. [63]

    Adam: A Method for Stochastic Optimization,

    Kingma, D. P ., and Ba, J., “Adam: A Method for Stochastic Optimization,” , 2017. https://doi.org/10.48550/arXiv.1412.6980, URL https://arxiv.org/abs/1412.6980v9. 47

    2017 doi arXiv

  55. [2014]

    URL https://www.issfd.org/ISSFD_2014/ISSFD24_Paper_S7-3_Englander.pdf

  56. [2017]

    URL https://ntrs.nasa.gov/citations/20170008012

Pith tools

Reviewed August 10, 2026 · model on record in the stance chip above.