Skip to content

Bare DirectTrajOptProblem + TerminalObjective: Ipopt restoration stall (inf_du spike ~1e3 at iter 1) — dynamics-violating returns, warp-independent #153

Description

@aarontrowbridge

Found during the Piccolo #321 end-to-end smoke. Bare DirectTrajOptProblem construction (no du/ddu spline machinery) + a TerminalObjective-class objective stalls Ipopt into restoration ('Error in step computation', inf_du ~1e3 at iteration 1) and returns dynamics-violating points. Reproduces with NO warp on pinned DTO 0.10.1 — not a warp regression.

Evaluator verified CORRECT: a runtime-patched MOI.eval_objective probe returns the full objective (~99.9) at every solver probe with properly packed z — the defect is downstream in the Ipopt step computation (Ipopt's displayed 0 is scaled output, not an objective drop).

Repro + probe: the smoke/debug scripts are preserved in the piccolo-stack-program audit trail (session-20260829, L12); construction shape = bare problem + TerminalObjective on a warp-free 1q unitary. Suspected seam: bare-problem objective scaling/regularization defaults vs the SmoothPulseProblem battle-tested path (whose solves converge with the same objective class).

Activity

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

No one assigned

    Labels

    hitlNeeds human decision/review before merge

    Type

    No type

    Projects

    No projects

      Milestone

      No milestone

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions