Pith. sign in

REVIEW 3 major objections 4 minor 40 references

Risk-Sensitive Orbital Debris Collision Avoidance using Distributionally Robust Chance Constraints

T0 review · 3 major / 4 minor · reviewed 2026-08-11 · deepseek-v4-flash

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

Pith's one-line read The paper shows that a closed-form trace inequality, derived from a distributionally robust Conditional Value-at-Risk constraint, can enforce a guaranteed collision-probability bound for a satellite against space debris using only the…

desk verdict Important problem and a sensible application of known CVaR machinery, but the central safety guarantee is not established because the ellipsoidal safe set is never constructed from the collision geometry; in the paper's own 41 m conjunction the debris mean lies inside the 100 m threshold, so the reformulated constraint cannot imply the stated collision probability bound. read the letter →

arxiv 2412.17358 v1 pith:RGXDLONN submitted 2024-12-23 eess.SY cs.ROcs.SY

Kanghyun Ryu , Jean-Baptiste Bouvier , Shazaib Lalani , Siegfried Eggl , Negar Mehr This is my paper · ORCID
classification eess.SYcs.ROcs.SY
keywords distributionallyrobustoptimizationchanceconstraintsConditionalValue-at-RiskcollisionavoidanceorbitaldebrisuncertaintypropagationmodelpredictivecontrolCross-EntropyMethod
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 aims to show that a satellite can enforce a collision-probability bound against space debris using only the mean and covariance of the debris position, without knowing its full non-Gaussian distribution. It does this through a distributionally robust chance constraint: instead of requiring the true distribution to satisfy the probability bound, it requires the worst-case distribution among all distributions with the same first two moments to satisfy a Conditional Value-at-Risk bound. The central result is the closed-form identity $\sup_{P \in \mathcal{P}} \mathrm{CVaR}^{\!P}_{\varepsilon}(l(r_d)) = -1 + (1/\varepsilon)\operatorname{Tr}\{\Sigma_d E\}$, which turns the chance constraint into a simple trace inequality that can be inserted into a model predictive controller. The authors validate the approach in a simulated, real-world-inspired close approach, using three different uncertainty propagation methods, and show that the resulting controller maintains the minimum safe distance while consuming fuel proportional to the strictness of the probability bound. A sympathetic reader would care because it offers a path from hard-to-obtain full distributions to tractable, conservative collision avoidance.

What carries the argument

The load-bearing object is the distributionally robust Conditional Value-at-Risk constraint built on a moment ambiguity set $\mathcal{P}^k = \{P : \mathbb{E}_P[r_d^k] = \mu_d^k,\ \mathbb{E}_P[(r_d^k - \mu_d^k)(r_d^k - \mu_d^k)^T] = \Sigma_d^k\}$. The paper's Theorem 1 gives the closed-form value $\sup_{P \in \mathcal{P}^k} \mathrm{CVaR}^{\!P}_{\varepsilon}(l^k(r_d^k)) = -1 + (1/\varepsilon)\operatorname{Tr}\{\Sigma_d^k E^k\}$ for the quadratic safety cost $l^k(r) = (r - \mu_d^k)^T E^k (r - \mu_d^k) - 1$; this identity is what converts a worst-case risk measure into a simple trace inequality that can be evaluated from moment estimates alone. The CVaR-to-chance-constraint chain (CVaR bound implies VaR bound implies probability bound) supplies the conservatism that makes the trace inequality a sufficient condition for collision avoidance. The ellipsoid defined by $E^k$ is the other essential piece: it convexifies the nonconvex collision-free set so the closed form applies, at the price of replacing the true distance condition with a containment assumption.

What would settle it

Run a Monte Carlo check on a strongly non-Gaussian debris distribution with the same mean and covariance, and compare the empirical fraction of times the distance falls below the threshold with the bound $\varepsilon$ while the trace inequality is satisfied. A violation would show that the ellipsoid does not actually contain the collision-free set.

Watch

Extended reading notes

Core claim

On the paper's own terms, the discovery is that distributionally robust collision avoidance reduces to checking the inequality $-1 + (1/\varepsilon)\operatorname{Tr}\{\Sigma_d^k E^k\} \le 0$ at each time step. Here $E^k$ is a positive-definite matrix defining an ellipsoid $\{r : (r - \mu_d^k)^T E^k (r - \mu_d^k) \le 1\}$ that under-approximates the true collision-free set centered at the estimated debris mean, and $\Sigma_d^k$ is the covariance of the debris position estimate. Theorem 1 states that this inequality is equivalent to $\sup_{P \in \mathcal{P}^k} \mathrm{CVaR}^{\!P}_{\varepsilon}(l^k(r_d^k)) \le 0$, where $\mathcal{P}^k$ is the set of all distributions with mean $\mu_d^k$ and covariance $\Sigma_d^k$, and $l^k$ is the quadratic safety cost. Because CVaR dominates VaR, this condition is sufficient for every distribution in the ambiguity set to satisfy $\mathrm{Prob}(r_d^k \in \mathcal{R}_{\mathrm{free}}^k) \ge 1 - \varepsilon$, which in turn guarantees the original chance constraint whenever the true distribution is moment-matched. The authors then embed this closed-form constraint in a fuel-minimizing model predictive control problem solved by the Cross-Entropy Method, and demonstrate in simulation that the controller meets the failure-probability bound across linear, unscented, and Monte Carlo uncertainty propagation.

Load-bearing premise

The safety guarantee rests on the assumption that the chosen ellipsoid lies entirely inside the true collision-free region, an under-approximation the paper states but does not prove from the satellite position and distance threshold.

Editorial extensions

If this is right

  • A collision-probability bound of $1-\varepsilon$ can be enforced using only moment estimates, so any uncertainty propagation method that outputs a mean and covariance (linear Gaussian, unscented transform, Monte Carlo) can be plugged into the same controller.
  • Smaller allowable collision probability $\varepsilon$, larger debris covariance, or a smaller ellipsoid all push the trace term up, so the controller responds by keeping a larger distance and spending more $\Delta v$; the simulations show this trade-off explicitly.
  • When no sampled control sequence satisfies the closed-form constraint, selecting sequences by the discounted trajectory risk steers the Cross-Entropy Method toward safer maneuvers rather than terminating with an infeasible plan.
  • The closed-form constraint is conservative: it holds for every distribution with the given mean and covariance, so the true collision probability is no larger than $\varepsilon$ if the ellipsoid containment assumption holds.

Reading between the lines

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

  • A direct extension the authors leave implicit: the method becomes a plug-in safety layer for any nonlinear propagator that reports moments, such as polynomial chaos or Gaussian mixture models, since the trace inequality depends only on those moments.
  • The closed form also suggests a design rule: since $-1 + (1/\varepsilon)\operatorname{Tr}\{\Sigma_d E\} \le 0$ is equivalent to $\operatorname{Tr}\{\Sigma_d E\} \le \varepsilon$, the controller should trigger a maneuver as soon as the estimated covariance and chosen ellipsoid exceed that threshold; this threshold could be used for early-warning screening before full trajectory optimization.
  • A testable extension would be to run a second Monte Carlo evaluation on the same scenario, computing the empirical collision probability under a strongly non-Gaussian distribution with matching moments, to measure how much conservatism both the CVaR approximation and the ellipsoidal under-approximation add.
  • The safety guarantee is modular: the trace inequality is about the ellipsoid, not the satellite geometry, so connecting $E^k$ to the actual distance threshold through a proven containment ellipsoid would turn the method into a certified collision-avoidance guarantee rather than a heuristic safeguard.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

3 major / 4 minor

Summary. The paper proposes a distributionally robust chance-constrained model predictive control approach for satellite collision avoidance with space debris. The debris position uncertainty is represented by a moment ambiguity set containing all distributions with a given mean and covariance, and the collision chance constraint is replaced by a Conditional Value-at-Risk (CVaR) constraint whose closed form, Eq. (15), depends only on the trace of the covariance times an ellipsoidal shape matrix. The resulting optimal control problem is solved with a constrained Cross-Entropy Method, and the approach is tested on a Starlink-THEA conjunction scenario using linear Gaussian, unscented transform, and Monte Carlo uncertainty propagation. The central claim is that the closed-form constraint (16e) guarantees that the true collision probability is at most epsilon for any debris distribution with the estimated mean and covariance.

Significance. If the central claim were established, the paper would offer a valuable practical contribution: a chance-constrained collision avoidance method that requires only moment information, is agnostic to the uncertainty propagation technique, and avoids the sample complexity of direct Monte Carlo chance constraint evaluation. The use of the known CVaR closed form and the CEM solver are reasonable algorithmic choices, and the comparative simulations across three propagation methods are informative about the sensitivity of the resulting maneuvers. However, the paper's headline safety guarantee is not established, because the ellipsoidal safe set used in Eq. (14) is never connected to the actual collision-free set of Eq. (4), and in the paper's own scenario the two sets cannot be related in the required way.

major comments (3)
  1. [Section VI.A, Eq. (14)] The ellipsoidal safe set is introduced without any construction rule: the shape matrix E^k is never defined as a function of the satellite position r_s^k and the collision threshold d_thres, and no subset containment R^k_free ⊆ R_free is proven. Since the closed-form constraint (16e) only enforces l(r) ≤ 0 for the ellipsoid, the chain of implications from (16e) to the original chance constraint (6e) is broken unless such containment is established. The text labels this an under-approximation, but no geometric argument is given.
  2. [Section VII.A and Fig. 4] In the reported conjunction, the nominal close-approach distance is 41 m while d_thres = 100 m, so the debris mean mu_d^k lies strictly inside the collision ball {r : ||r - r_s^k|| ≤ 100 m}. For any E^k ≻ 0, the set {r : (r - mu_d^k)^T E^k (r - mu_d^k) ≤ 1} contains an open ball around mu_d^k and therefore contains points with ||r - r_s^k|| < 100 m. Hence R^k_free cannot be a subset of R_free in this scenario, and constraint (16e) does not imply the claimed chance constraint (6e). The paper offers no scaling or orientation rule for E^k that could avoid this obstruction.
  3. [Section VIII] The simulations validate only the minimum satellite-debris distance and the total Delta-v; they never estimate the empirical collision probability Prob(r_d in R_free) under the propagated debris distributions, nor do they check the chance constraint (5). Thus the central safety claim is not tested even in cases where the ellipsoidal geometry might be more favorable. The reported minimum distances above 100 m show that the samples did not collide in those runs, but they do not quantify the probabilistic guarantee that the paper claims.
minor comments (4)
  1. [Eq. (12)] Equation (12) contains a notation error: the expression sup_{P in P} VaR^{P*}_epsilon(l(r_d)) appears to use the true distribution P* inside the supremum; it should read VaR^P_epsilon(l(r_d)).
  2. [Definition 2 and Eq. (14)] Definition 2 states that R_free is a sublevel set {r : l(r) ≤ 0}, while Eq. (4) defines R_free as the superlevel set of the distance function, and Eq. (14) later redefines R^k_free as an ellipsoid centered on the debris mean. The change of definition should be made explicit, and a different symbol should be used for the ellipsoidal under-approximation to avoid conflating it with the true collision-free set in (6e).
  3. [Algorithm 1] The last line of Algorithm 1, 'Output first control input u* {u^{0:K}_s}_elite,' is garbled and should read 'Output the first control input u^{0*}_s from the elite set.'
  4. [Section VI.B, Eq. (17)] When no feasible control sequence is found, the algorithm selects the elite set by minimizing Trajectory Risk; this fallback procedure does not provide any safety guarantee, since the selected sequence may still violate constraint (16e). The paper should explicitly acknowledge this limitation, especially because the scenario's geometry already prevents the constraint from implying the original chance constraint.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: the central CVaR bound is taken from external reference [17], and the only self-citation concerns the CEM optimizer rather than the safety guarantee.

full rationale

The paper's load-bearing inequality (15), sup_{P in P^k} CVaR^P_eps(l(r_d^k)) = -1 + (1/eps) Tr{Sigma_d^k E^k}, is attributed to the external reference [17] ('Proof. See Proof of Corollary 1.3 in [17]'), not to the authors' own prior work, so the distributionally robust CVaR reformulation is independent support rather than a self-citation chain. The equivalence chain (11)-(12) is a standard VaR/CVaR implication derived from the definitions in (8)-(10). No parameter is fitted to the quantity that is later 'predicted': the ambiguity set (13) uses externally supplied mean and covariance, and the constraint (16e) is the same closed-form expression (15) restated. The only self-citation [18] appears in Section VI.B ('Following [18, 35], we use a variant of CEM to solve constrained MPC problems') and is paired with an external citation [35]; it concerns the sampling optimizer, not the collision-safety claim, so it is not load-bearing. The genuine weakness is a correctness/validation gap, not circularity: E^k in Eq. (14) is never defined as a function of r_s^k or d_thres, and no subset containment of the ellipsoid {l^k(r) <= 0} in the true collision-free set (4) is shown; consequently, in the 41 m conjunction with d_thres = 100 m, the ellipsoid centered on the debris mean cannot be contained in the exterior-of-ball safe set. Because this gap does not reduce the theorem's conclusion to its assumptions by construction, and because the reported minimum-distance checks in Figs. 4-5 are external to the moments-only constraint, the circularity score is 0.

Assumptions & free parameters 3 free parameters · 6 assumptions · 0 invented entities

The load-bearing items are the moment ambiguity set assumption and the unproven ellipsoidal safe set; the latter is both a design parameter and a missing link to the physical collision threshold. No new physical entities are introduced.

free parameters (3)
  • E^k (shape matrix of the ellipsoidal safe set) = not reported
    All simulation results and the chance-constraint reformulation (16e) depend on E^k, but the paper gives no formula or value for it. The safety guarantee is vacuous until this matrix is specified.
  • P_d^0 (initial debris state covariance) = not reported
    The initial Gaussian covariance drives every uncertainty propagator in Section VII.B, but its numerical value is not provided, so the reported distances and delta-v cannot be reproduced.
  • CEM hyperparameters and discount gamma = not reported
    Algorithm 1 requires sample counts, elite sizes, iteration counts, and the discount factor in Eq. (17); none are given, and the solution quality depends on them.
assumptions (6)
  • domain assumption The true debris position distribution P* belongs to the moment ambiguity set P of distributions with mean mu_d and covariance Sigma_d.
    Invoked in Eq. (7) and Theorem 1; if the propagator's moments are wrong, the guarantee does not apply to the physical distribution.
  • ad hoc to paper The ellipsoid {r : (r - mu_d)^T E (r - mu_d) <= 1} is a subset of the true collision-free set {r : ||r - r_s|| > d_thres}.
    Stated around Eq. (14) and Figure 2 without construction or proof; the subset relation is not guaranteed when the satellite is close to the debris mean.
  • domain assumption Initial debris uncertainty is Gaussian with zero-mean white process noise.
    Problem 1 (Eq. 6d) and Section VII.A assume x_d^0 ~ N(mu_d^0, P_d^0) and w_d ~ N(0, Q).
  • domain assumption The satellite state is deterministically known via GNSS tracking.
    Section IV assumes deterministic satellite dynamics; all uncertainty is assigned to debris.
  • standard math Theorem 1, Eq. (15), from Van Parys et al. [17] is correct.
    The paper cites Corollary 1.3 of [17] without proof; this external result is the mathematical core of the method.
  • domain assumption Only Earth gravity and atmospheric drag are included in the dynamics.
    Equations (1) and (2) omit J2 and other perturbations, which limits fidelity of the real-world-inspired scenario.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Risk-Sensitive Orbital Debris Collision Avoidance using Distributionally Robust Chance Constraints." pith.science (2026). https://pith.science/paper/RGXDLONN

@misc{pith2026241217358,
  author       = {Pith},
  title        = {Pith review of: Risk-Sensitive Orbital Debris Collision Avoidance using Distributionally Robust Chance Constraints},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/RGXDLONN}},
  note         = {Machine review of arXiv:2412.17358}
}
read the original abstract

The exponential increase in orbital debris and active satellites will lead to congested orbits, necessitating more frequent collision avoidance maneuvers by satellites. To minimize fuel consumption while ensuring the safety of satellites, enforcing a chance constraint, which poses an upper bound in collision probability with debris, can serve as an intuitive safety measure. However, accurately evaluating collision probability, which is critical for the effective implementation of chance constraints, remains a non-trivial task. This difficulty arises because uncertainty propagation in nonlinear orbit dynamics typically provides only limited information, such as finite samples or moment estimates about the underlying arbitrary non-Gaussian distributions. Furthermore, even if the full distribution were known, it remains unclear how to effectively compute chance constraints with such non-Gaussian distributions. To address these challenges, we propose a distributionally robust chance-constrained collision avoidance algorithm that provides a sufficient condition for collision probabilities under limited information about the underlying non-Gaussian distribution. Our distributionally robust approach satisfies the chance constraint for all debris position distributions sharing a given mean and covariance, thereby enabling the enforcement of chance constraints with limited distributional information. To achieve computational tractability, the chance constraint is approximated using a Conditional Value-at-Risk (CVaR) constraint, which gives a conservative and tractable approximation of the distributionally robust chance constraint. We validate our algorithm on a real-world inspired satellite-debris conjunction scenario with different uncertainty propagation methods and show that our controller can effectively avoid collisions.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

40 extracted references · 37 canonical work pages

  1. [1]

    Detecting, tracking and imaging space debris,

    Mehrholz, D., Leushacke, L., Flury, W., Jehn, R., Klinkrad, H., and Landgraf, M., “Detecting, tracking and imaging space debris,”ESA Bulletin(0376-4265), , No. 109, 2002, pp. 128–134

    work page 2002

  2. [2]

    International Space Station swerves to avoid Russian space debris, NASA says,

    Wattles, J., and Hunt, K., “International Space Station swerves to avoid Russian space debris, NASA says,”CNN, 2022. URL https://www.cnn.com/2022/10/25/world/iss-maneuver-russia-space-junk-scn/index.html

    work page 2022

  3. [3]

    Cost and risk assessment for spacecraft operation decisions caused by the space debris environment,

    Schaub, H., Jasper, L. E., Anderson, P. V., and McKnight, D. S., “Cost and risk assessment for spacecraft operation decisions caused by the space debris environment,”Acta Astronautica, Vol. 113, 2015, pp. 66–79

    work page 2015

  4. [4]

    Operational support to collision avoidance activities by ESA’s space debris office,

    Braun, V., Flohrer, T., Krag, H., Merz, K., Lemmens, S., Bastida Virgili, B., and Funke, Q., “Operational support to collision avoidance activities by ESA’s space debris office,”CEAS Space Journal, Vol. 8, No. 3, 2016, pp. 177–189

    work page 2016

  5. [5]

    A review of uncertainty propagation in orbital mechanics,

    Luo, Y.-z., and Yang, Z., “A review of uncertainty propagation in orbital mechanics,”Progress in Aerospace Sciences, Vol. 89, 2017, pp. 23–39

    work page 2017

  6. [6]

    Convex optimization of collision avoidance maneuvers in the presence of uncertainty,

    Dutta, S., and Misra, A. K., “Convex optimization of collision avoidance maneuvers in the presence of uncertainty,”Acta Astronautica, Vol. 197, 2022, pp. 257–268

    work page 2022

  7. [7]

    K.,Spacecraft Collision Probability, American Institute of Aeronautics and Astronautics, Inc., 2008

    Chan, F. K.,Spacecraft Collision Probability, American Institute of Aeronautics and Astronautics, Inc., 2008

    work page 2008

  8. [8]

    An approach for nonlinear uncertainty propagation: Application to orbital mechanics,

    Giza, D., Singla, P., and Jah, M., “An approach for nonlinear uncertainty propagation: Application to orbital mechanics,”AIAA Guidance, Navigation, and Control Conference, 2009, p. 6082

    work page 2009

Show all 40 references
  1. [9]

    Nonlinearuncertaintypropagationforperturbedtwo-bodyorbits,

    Vishwajeet,K.,Singla,P.,andJah,M.,“Nonlinearuncertaintypropagationforperturbedtwo-bodyorbits,” JournalofGuidance, Control, and Dynamics, Vol. 37, No. 5, 2014, pp. 1415–1425

    2014

  2. [10]

    Conjugate unscented transformation-based approach for accurate conjunction analysis,

    Adurthi, N., and Singla, P., “Conjugate unscented transformation-based approach for accurate conjunction analysis,”Journal of Guidance, Control, and Dynamics, Vol. 38, No. 9, 2015, pp. 1642–1658

    2015

  3. [11]

    Convex approximations of chance constrained programs,

    Nemirovski, A., and Shapiro, A., “Convex approximations of chance constrained programs,”SIAM Journal on Optimization, Vol. 17, No. 4, 2007, pp. 969–996

    2007

  4. [12]

    Monte Carlo method for collision probability calculations using 3D satellite models,

    De Vries, W., and Phillion, D., “Monte Carlo method for collision probability calculations using 3D satellite models,”Advanced Maui Optical and Space Surveillance Technologies Conference, 2010

    2010

  5. [13]

    Probability of collision with special perturbation dynamics using the Monte Carlo method,

    Sabol, C., Binz, C., Segerman, A., Roe, K., and Schumacher Jr, P. W., “Probability of collision with special perturbation dynamics using the Monte Carlo method,”AAS/AIAA Astrodynamics Specialist Conference, Vol. 142, 2011, pp. 1081–1094

    2011

  6. [14]

    An optimal algorithm for Monte Carlo estimation,

    Dagum, P., Karp, R., Luby, M., and Ross, S., “An optimal algorithm for Monte Carlo estimation,”SIAM Journal on Computing, Vol. 29, No. 5, 2000, pp. 1484–1496

    2000

  7. [15]

    Distributionally robust optimization: A review,

    Rahimian, H., and Mehrotra, S., “Distributionally robust optimization: A review,”arXiv preprint arXiv:1908.05659, 2019

    1908 arXiv

  8. [16]

    Risk-sensitive and robust decision-making: a cvar optimization approach,

    Chow, Y., Tamar, A., Mannor, S., and Pavone, M., “Risk-sensitive and robust decision-making: a cvar optimization approach,” Advances in neural information processing systems, Vol. 28, 2015. 14

    2015

  9. [17]

    Distributionally robust control of constrained stochastic systems,

    Van Parys, B. P., Kuhn, D., Goulart, P. J., and Morari, M., “Distributionally robust control of constrained stochastic systems,” IEEE Transactions on Automatic Control, Vol. 61, No. 2, 2015, pp. 430–442

    2015

  10. [18]

    Integrating Predictive Motion Uncertainties with Distributionally Robust Risk-Aware Control for Safe Robot Navigation in Crowds,

    Ryu, K., and Mehr, N., “Integrating Predictive Motion Uncertainties with Distributionally Robust Risk-Aware Control for Safe Robot Navigation in Crowds,”arXiv preprint arXiv:2403.05081, 2024

    2024 arXiv

  11. [19]

    Satellite orbital conjunction reports assessing threatening encounters in space (SOCRATES),

    Kelso, T. S., and Alfano, S., “Satellite orbital conjunction reports assessing threatening encounters in space (SOCRATES),” Modeling, Simulation, and Verification of Space-based Systems III, Vol. 6221, edited by P. Motaghedi, International Society for Optics and Photonics, SPI...

    2006 doi

  12. [20]

    Non-Gaussianerrorpropagationinorbitalmechanics,

    Junkins, J.L., Akella, M.R., andAlfrined, K.T., “Non-Gaussianerrorpropagationinorbitalmechanics,” Guidanceandcontrol 1996, 1996, pp. 283–298

    1996

  13. [21]

    The partial banana mapping: A robust linear method for impact probability estimation,

    Vavilov, D. E., “The partial banana mapping: A robust linear method for impact probability estimation,”Monthly Notices of the Royal Astronomical Society, Vol. 492, No. 3, 2020, pp. 4546–4552

    2020

  14. [22]

    Nonlinear propagation of orbit uncertainty using non-intrusive polynomial chaos,

    Jones, B. A., Doostan, A., and Born, G. H., “Nonlinear propagation of orbit uncertainty using non-intrusive polynomial chaos,” Journal of Guidance, Control, and Dynamics, Vol. 36, No. 2, 2013, pp. 430–444

    2013

  15. [23]

    Analytical uncertainty propagation for satellite relative motion along elliptic orbits,

    Lee, S., Lyu, H., and Hwang, I., “Analytical uncertainty propagation for satellite relative motion along elliptic orbits,”Journal of Guidance, Control, and Dynamics, Vol. 39, No. 7, 2016, pp. 1593–1601

    2016

  16. [24]

    Space object collision probability using multidirectional Gaussian mixture models,

    Vittaldev, V., and Russell, R. P., “Space object collision probability using multidirectional Gaussian mixture models,”Journal of Guidance, Control, and Dynamics, Vol. 39, No. 9, 2016, pp. 2163–2169

    2016

  17. [25]

    Collision avoidance for satellites in formation flight,

    Slater, G. L., Byram, S. M., and Williams, T. W., “Collision avoidance for satellites in formation flight,”Journal of Guidance, Control, and Dynamics, Vol. 29, No. 5, 2006, pp. 1140–1146

    2006

  18. [26]

    Satellite collision avoidance using repeated games,

    Dolan, S., Qin, V., Ding, G., and Balakrishnan, H., “Satellite collision avoidance using repeated games,”AAS/AIAA Astrodynamics Specialist Conference, American Astronautical Society, 2023

    2023

  19. [27]

    Stochasticmulti-satellitemaneuveringwithconstraints in an elliptical orbit,

    Pacula,I.,Vinod,A.,Sivaramakrishnan,V.,Petersen,C.,andOishi,M.,“Stochasticmulti-satellitemaneuveringwithconstraints in an elliptical orbit,”American Control Conference, 2021, pp. 4261–4268

    2021

  20. [28]

    Model for linearized satellite relative motion about a J2-perturbed mean circular orbit,

    Vadali, S. R., “Model for linearized satellite relative motion about a J2-perturbed mean circular orbit,”Journal of guidance, control, and dynamics, Vol. 32, No. 5, 2009, pp. 1687–1691

    2009

  21. [29]

    Collision avoidance maneuver planning with robust optimization,

    Mueller, J. B., and Larsson, R., “Collision avoidance maneuver planning with robust optimization,”International ESA Conference on Guidance, Navigation and Control Systems, Tralee, County Kerry, Ireland, 2008

    2008

  22. [30]

    Onboard planning of collision avoidance maneuvers using robust optimization,

    Mueller, J., “Onboard planning of collision avoidance maneuvers using robust optimization,”AIAA Infotech@ Aerospace Conference and AIAA Unmanned... Unlimited Conference, 2009, p. 2051

    2009

  23. [31]

    Distributionally robust joint chance constraints with second-order moment information,

    Zymler, S., Kuhn, D., and Rustem, B., “Distributionally robust joint chance constraints with second-order moment information,” Mathematical Programming, Vol. 137, 2013, pp. 167–198

    2013

  24. [32]

    Conditional value-at-risk approximation to value-at-risk constrained programs: A remedy via Monte Carlo,

    Hong, L. J., Hu, Z., and Zhang, L., “Conditional value-at-risk approximation to value-at-risk constrained programs: A remedy via Monte Carlo,”INFORMS Journal on Computing, Vol. 26, No. 2, 2014, pp. 385–400

    2014

  25. [33]

    Optimization of conditional value-at-risk,

    Rockafellar, R. T., and Uryasev, S., “Optimization of conditional value-at-risk,”Journal of Risk, Vol. 2, 2000, pp. 21–42

    2000

  26. [34]

    A tutorial on the cross-entropy method,

    De Boer, P.-T., Kroese, D. P., Mannor, S., and Rubinstein, R. Y., “A tutorial on the cross-entropy method,”Annals of Operations Research, Vol. 134, 2005, pp. 19–67

    2005

  27. [35]

    Constrained model-based reinforcement learning with robust cross-entropy method,

    Liu, Z., Zhou, H., Chen, B., Zhong, S., Hebert, M., and Zhao, D., “Constrained model-based reinforcement learning with robust cross-entropy method,”arXiv preprint arXiv:2010.07968, 2020

    2010 arXiv

  28. [36]

    Ansys Systems Tool Kit (STK),

    ANSYS, “Ansys Systems Tool Kit (STK),” , 2023

    2023

  29. [37]

    THEA Satellite - Technical Information,

    SpaceQuest, L., “THEA Satellite - Technical Information,” , May 2018. Accessed: 2024-11-30

    2018

  30. [38]

    Jacod, J.,Discretization of processes, Springer, 2012

    2012

  31. [39]

    Performance comparison between EKF and UKF in GPS/INS low observability conditions,

    Ryu, K., Kang, J., and Lee, D., “Performance comparison between EKF and UKF in GPS/INS low observability conditions,” 21st International Conference on Control, Automation and Systems (ICCAS), IEEE, 2021, pp. 1911–1916

    2021

  32. [40]

    The unscented Kalman filter for nonlinear estimation,

    Wan, E. A., and Van Der Merwe, R., “The unscented Kalman filter for nonlinear estimation,”Proceedings of the IEEE 2000 adaptive systems for signal processing, communications, and control symposium (Cat. No. 00EX373), Ieee, 2000, pp. 153–158. 15

    2000

Pith tools

Reviewed August 11, 2026 · model on record in the stance chip above.