pith. machine review for the scientific record. sign in

arxiv: 2605.08420 · v1 · submitted 2026-05-08 · 🧮 math.OC · cs.SY· eess.SY

Recognition: 2 theorem links

· Lean Theorem

Transcription-Induced Failure Modes in 6-DOF Rocket Landing Trajectory Optimization

Prayag Sharma , Jonathan Y.M. Goh , Beh\c{c}et A\c{c}{\i}kme\c{s}e , Franck Djeumou
Authors on Pith no claims yet

Pith reviewed 2026-05-12 01:17 UTC · model grok-4.3

classification 🧮 math.OC cs.SYeess.SY
keywords transcription methodsoptimal controlrocket landingtruncation errorinvariant driftdiscretizationnonlinear programming6-DOF dynamics
0
0 comments X

The pith

Only three of fourteen transcription methods produce dynamically feasible trajectories for 6-DOF rocket landing optimization.

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

The paper establishes that discretizing the continuous dynamics of optimal control problems for use in large-scale nonlinear programming solvers introduces hidden vulnerabilities, primarily through truncation errors and drift away from invariants such as quaternion norms. To expose these issues, the authors construct a problem- and transcription-agnostic adversarial objective that amplifies local truncation-error bounds. When this test is run on a 6-DOF rocket landing problem, fourteen common transcription methods are evaluated and only three meet rigorous validation criteria for dynamic feasibility. The work further shows that these theoretical defects determine practical outcomes, including a performance inversion in which a fourth-order implicit scheme matches the accuracy of a sixth-order explicit scheme and delivers better solve speed and robustness on lateral-divert maneuvers.

Core claim

When applied to a 6-DOF rocket-landing problem, we reveal a stark reliability gap: of fourteen transcription methods tested, only three satisfy rigorous validation criteria. These results also expose a striking performance inversion: even in the absence of classical stiffness, a fourth-order implicit scheme (GL2) matches the fidelity of a sixth-order explicit method (RK6). Using B-series expansions and symplectic Runge-Kutta theorems, we isolate the specific truncation errors and quaternion-invariant drift responsible for these failures. Crucially, these theoretical vulnerabilities dictate operational performance: in practical lateral-divert scenarios, the implicit GL2 consistently outperms

What carries the argument

An adversarial objective constructed from local truncation-error bounds that forces transcription defects to become visible in the solved trajectory.

If this is right

  • Only three of the fourteen tested transcription methods satisfy the validation criteria for dynamic feasibility.
  • A fourth-order implicit scheme achieves comparable fidelity to a sixth-order explicit scheme even without classical stiffness.
  • Theoretical truncation errors and quaternion drift directly control solve speed and robustness in lateral-divert maneuvers.
  • The implicit GL2 scheme outperforms the explicit RK6 scheme in both end-to-end solve time and reliability on practical rocket landing instances.

Where Pith is reading between the lines

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

  • The same adversarial testing procedure could be applied to other optimal control problems to rank transcription methods by reliability before deployment.
  • Implicit methods may be preferable to higher-order explicit ones even in non-stiff problems if invariant preservation is the dominant error source.
  • Redesigning transcription schemes to enforce quaternion invariance at the discrete level could eliminate one major class of observed failures.

Load-bearing premise

The adversarial objective and chosen validation criteria expose genuine operational failure modes rather than artifacts introduced by the objective itself.

What would settle it

Re-running the fourteen transcription methods on the same 6-DOF rocket landing problem with a standard objective (no adversarial term) and checking whether the same three methods still pass validation while the GL2-RK6 inversion disappears.

Figures

Figures reproduced from arXiv: 2605.08420 by Beh\c{c}et A\c{c}{\i}kme\c{s}e, Franck Djeumou, Jonathan Y.M. Goh, Prayag Sharma.

Figure 1
Figure 1. Figure 1: Feasible landing trajectories obtained from [PITH_FULL_IMAGE:figures/full_fig_p004_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Quaternion norm drift ∥q(t)∥2 − 1 under open-loop propagation on the adversarial trajectory. GL2 exhibits near-zero drift around 10−12 , while RK5 and Lobatto IIIA drift on the order of 10−4 . Quaternion normalization and its failure modes: To iso￾late the role of this invariant drift on the optimizer outcome, [PITH_FULL_IMAGE:figures/full_fig_p006_2.png] view at source ↗
Figure 4
Figure 4. Figure 4: Isolated per-step local truncation errors and principal [PITH_FULL_IMAGE:figures/full_fig_p007_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Representative divert maneuvers (normalized coordinates) with [PITH_FULL_IMAGE:figures/full_fig_p007_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Average wall-clock time (function evaluation + remaining solver [PITH_FULL_IMAGE:figures/full_fig_p008_6.png] view at source ↗
read the original abstract

Solving optimal control problems via large-scale NLP solvers depends on discretizing continuous dynamics. Yet, this transcription step hides critical vulnerabilities-most notably truncation error and invariant drift-that can drive solvers toward dynamically infeasible or suboptimal trajectories. To expose these hidden failures, we introduce a problem- and transcription-agnostic adversarial objective that leverages the structure of local truncation-error bounds to aggressively amplify such defects. When applied to a 6-DOF rocket-landing problem, we reveal a stark reliability gap: of fourteen transcription methods tested, only three satisfy rigorous validation criteria. These results also expose a striking performance inversion: even in the absence of classical stiffness, a fourth-order implicit scheme (GL2) matches the fidelity of a sixth-order explicit method (RK6). Using B-series expansions and symplectic Runge-Kutta theorems, we isolate the specific truncation errors and quaternion-invariant drift responsible for these failures. Crucially, these theoretical vulnerabilities dictate operational performance: in practical lateral-divert scenarios, the implicit GL2 consistently outperforms the explicit RK6 in both end-to-end solve speed and robustness.

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 paper introduces a problem- and transcription-agnostic adversarial objective that amplifies local truncation-error bounds to expose vulnerabilities (truncation error and quaternion-invariant drift) in direct transcription methods for optimal control. Applied to a 6-DOF rocket-landing problem, it reports that only three of fourteen tested methods satisfy rigorous validation criteria, identifies a performance inversion in which a fourth-order implicit Gauss-Legendre scheme (GL2) matches the fidelity of a sixth-order explicit Runge-Kutta method (RK6), and shows GL2 outperforming RK6 in lateral-divert scenarios. B-series expansions and symplectic Runge-Kutta theorems are used to isolate the responsible truncation errors and drift.

Significance. If the adversarial objective is shown to expose operationally relevant defects rather than artifacts, the work would usefully highlight transcription-induced reliability issues in aerospace trajectory optimization and provide concrete guidance on method selection. The explicit use of B-series analysis and symplectic-integrator theorems to connect theoretical error sources to observed NLP behavior is a strength, as is the empirical testing of fourteen distinct transcription schemes under a common validation framework.

major comments (2)
  1. [Abstract / adversarial-objective definition] The central reliability-gap and performance-inversion claims rest on the adversarial objective (described in the abstract) exposing genuine, generalizable failure modes. No comparison is provided to conventional objectives such as minimum-fuel or minimum-time; therefore it remains possible that the three-method subset and the GL2/RK6 inversion are specific to the constructed adversarial formulation rather than intrinsic properties of the transcription methods.
  2. [Validation criteria and results section] The validation criteria used to declare only three methods 'satisfactory' are not shown to be independent of the adversarial objective itself. If the criteria incorporate the same truncation-error amplification that the objective exploits, the selection of the three methods risks circularity and does not yet demonstrate dynamic feasibility under standard problem statements.
minor comments (2)
  1. [Notation and method enumeration] Notation for the fourteen transcription methods (e.g., GL2, RK6) should be defined once in a table or dedicated subsection rather than introduced piecemeal in the abstract and results.
  2. [Validation criteria] The manuscript would benefit from an explicit statement of the precise validation thresholds (e.g., maximum quaternion drift, integrated residual tolerance) used to classify methods as passing or failing.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for their constructive feedback. We address the major comments point by point below.

read point-by-point responses

  1. Referee: The central reliability-gap and performance-inversion claims rest on the adversarial objective (described in the abstract) exposing genuine, generalizable failure modes. No comparison is provided to conventional objectives such as minimum-fuel or minimum-time; therefore it remains possible that the three-method subset and the GL2/RK6 inversion are specific to the constructed adversarial formulation rather than intrinsic properties of the transcription methods.

    Authors: The adversarial objective is constructed to be problem- and transcription-agnostic by amplifying bounds on local truncation error that are intrinsic to each discretization method. These truncation errors and invariant drifts (e.g., quaternion drift) are properties of the integrator and affect dynamic feasibility independently of the cost function. B-series expansions in the manuscript isolate the responsible error terms without reference to any particular objective. Although the current manuscript does not contain side-by-side numerical results for minimum-fuel or minimum-time formulations, the validation step re-integrates each candidate solution against the original continuous dynamics using an independent high-order integrator; this check is objective-agnostic. We will add a clarifying paragraph in the discussion section. revision: partial

  2. Referee: The validation criteria used to declare only three methods 'satisfactory' are not shown to be independent of the adversarial objective itself. If the criteria incorporate the same truncation-error amplification that the objective exploits, the selection of the three methods risks circularity and does not yet demonstrate dynamic feasibility under standard problem statements.

    Authors: The validation criteria are post-optimization feasibility checks performed after the adversarial optimization has concluded. Each candidate trajectory is re-integrated with a high-order adaptive-step explicit Runge-Kutta method; the pointwise deviation from the transcribed solution is measured, and boundary conditions plus path constraints are verified to within numerical tolerances. These checks contain no reference to the adversarial objective or its amplification mechanism. We will revise the results section to state this separation explicitly and to include a brief description or pseudocode of the validation procedure. revision: yes

Circularity Check

0 steps flagged

No significant circularity; empirical validation and standard theory are self-contained

full rationale

The paper introduces an adversarial objective to amplify truncation and drift defects in transcription methods, then empirically tests fourteen methods on a 6-DOF rocket-landing problem against explicit validation criteria. It invokes B-series expansions and symplectic Runge-Kutta theorems (standard external tools) to isolate specific error sources. No load-bearing step reduces a claimed prediction, uniqueness result, or performance inversion to a fitted parameter, self-definition, or self-citation chain by construction. The reliability gap and GL2/RK6 inversion are presented as outcomes of the tests and analysis, not as tautological consequences of the inputs.

Axiom & Free-Parameter Ledger

0 free parameters · 2 axioms · 1 invented entities

The central claim rests on the validity of the adversarial objective construction and the assumption that the 6-DOF model plus chosen validation criteria capture real operational defects. No free parameters are described in the abstract.

axioms (2)
  • standard math B-series expansions accurately capture local truncation errors of the tested Runge-Kutta methods
    Invoked to isolate specific truncation errors responsible for failures
  • standard math Symplectic Runge-Kutta theorems correctly predict quaternion-invariant drift behavior
    Used to explain why certain methods preserve or violate invariants
invented entities (1)
  • adversarial objective no independent evidence
    purpose: to aggressively amplify truncation-error and invariant-drift defects in transcription methods
    Newly introduced to expose hidden vulnerabilities that standard objectives miss

pith-pipeline@v0.9.0 · 5512 in / 1498 out tokens · 68218 ms · 2026-05-12T01:17:56.438749+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

20 extracted references · 20 canonical work pages

  1. [1]

    CasADi – A software framework for nonlinear optimization and optimal control,

    J. A. E. Andersson, J. Gillis, G. Horn, J. B. Rawlings, and M. Diehl, “CasADi – A software framework for nonlinear optimization and optimal control,”Mathematical Programming Computation, vol. 11, no. 1, pp. 1–36, 2019

    work page 2019

  2. [2]

    A multiple shooting algorithm for direct solution of optimal control problems,

    H. G. Bock and K.-J. Plitt, “A multiple shooting algorithm for direct solution of optimal control problems,”IFAC Proceedings Volumes, vol. 17, no. 2, pp. 1603–1608, 1984

    work page 1984

  3. [3]

    Convex optimization for trajectory generation: A tutorial on generating dynamically feasible trajectories reliably and efficiently,

    D. Malyuta, T. P. Reynolds, M. Szmuk, T. Lew, R. Bonalli, M. Pavone, and B. Ac ¸ıkmes ¸e, “Convex optimization for trajectory generation: A tutorial on generating dynamically feasible trajectories reliably and efficiently,”IEEE Control Systems Magazine, vol. 42, no. 5, pp. 40– 113, 2022

    work page 2022

  4. [4]

    Advances in trajectory optimization for space vehicle control,

    D. Malyuta, Y . Yu, P. Elango, and B. Ac ¸ıkmes ¸e, “Advances in trajectory optimization for space vehicle control,”Annual Reviews in Control, vol. 52, pp. 282–315, 2021

    work page 2021

  5. [5]

    Knitro: An integrated pack- age for nonlinear optimization,

    R. H. Byrd, J. Nocedal, and R. A. Waltz, “Knitro: An integrated pack- age for nonlinear optimization,” inLarge-scale nonlinear optimization. Springer, 2006, pp. 35–59

    work page 2006

  6. [6]

    On the implementation of an interior- point filter line-search algorithm for large-scale nonlinear program- ming,

    A. W ¨achter and L. T. Biegler, “On the implementation of an interior- point filter line-search algorithm for large-scale nonlinear program- ming,”Mathematical programming, vol. 106, no. 1, pp. 25–57, 2006

    work page 2006

  7. [7]

    Ampl-nlp benchmark,

    H. Mittelmann, “Ampl-nlp benchmark,” Jun. 2025, updated June 28,

    work page 2025

  8. [8]

    [Online]

    Benchmarks for Optimization Software. [Online]. Available: https://plato.asu.edu/ftp/ampl-nlp.html

    work page

  9. [9]

    trajopt-util: Utilities for trajectory optimization,

    P. Elango, “trajopt-util: Utilities for trajectory optimization,” https://github.com/purnanandelango/trajopt-util, 2023, gitHub reposi- tory

    work page 2023

  10. [10]

    Successive convexification for trajectory optimization with continuous-time constraint satisfaction,

    P. Elango, D. Luo, A. G. Kamath, S. Uzun, T. Kim, and B. Ac ¸ıkmes ¸e, “Successive convexification for trajectory optimization with continuous-time constraint satisfaction,” 2024

    work page 2024

  11. [11]

    Onboard dual quaternion guidance for rocket landing,

    A. G. Kamath, J. A. Doll, P. Elango, T. Kim, S. Mceowen, Y . Yu, T. P. Reynolds, G. F. Mendeck, J. M. Carson III, M. Mesbahiet al., “Onboard dual quaternion guidance for rocket landing,”arXiv preprint arXiv:2508.10439, 2025

    work page arXiv 2025

  12. [12]

    J. C. Butcher,Numerical Methods for Ordinary Differential Equations, 2nd ed. Chichester, UK: John Wiley & Sons, Ltd, 2008

    work page 2008

  13. [13]

    Hairer, S

    E. Hairer, S. Nørsett, and G. Wanner,Solving Ordinary Differential Equations I: Nonstiff Problems, ser. Springer Series in Computational Mathematics. Springer Berlin Heidelberg, 2008

    work page 2008

  14. [14]

    Hairer and G

    E. Hairer and G. Wanner,Solving Ordinary Differential Equations II: Stiff and Differential-Algebraic Problems, ser. Springer Series in Computational Mathematics. Springer Berlin Heidelberg, 1996

    work page 1996

  15. [15]

    Implementation and experimental demonstration of onboard powered- descent guidance,

    D. P. Scharf, B. Ac ¸ıkmes ¸e, D. Dueri, J. Benito, and J. Casoliva, “Implementation and experimental demonstration of onboard powered- descent guidance,”Journal of Guidance, Control, and Dynamics, vol. 40, no. 2, pp. 213–229, 2017

    work page 2017

  16. [16]

    Successive convexification for mars 6-dof powered descent landing guidance,

    M. Szmuk, U. Eren, and B. Acikmese, “Successive convexification for mars 6-dof powered descent landing guidance,” inAIAA Guidance, Navigation, and Control Conference, 2017, p. 1500

    work page 2017

  17. [17]

    Kirk,Optimal Control Theory: An Introduction, ser

    D. Kirk,Optimal Control Theory: An Introduction, ser. Dover Books on Electrical Engineering Series. Dover Publications, 2004

    work page 2004

  18. [18]

    Atkinson and W

    K. Atkinson and W. Han,Theoretical Numerical Analysis: A Func- tional Analysis Framework, ser. Texts in Applied Mathematics. Springer New York, 2001

    work page 2001

  19. [19]

    Geometric numerical integration,

    E. Hairer, M. Hochbruck, A. Iserles, and C. Lubich, “Geometric numerical integration,”Oberwolfach Reports, vol. 8, no. 1, pp. 825– 900, 2011

    work page 2011

  20. [20]

    Butcher,Numerical Methods for Ordinary Differential Equations

    J. Butcher,Numerical Methods for Ordinary Differential Equations. Wiley, 2008

    work page 2008