[transfix]

Follow-up to #137. transfix/libcvc#851 and transfix/libcvc#852 are still open, so their rows cite their branch commits (6e7ab557 and 1a0acc77); a later docs PR can swap in the squash commits once they merge.

Changes to docs/vehicle-ph-plant.tex (+ PDF, 28 pages)

• Appendix A, reproducing the tables:
  • the exact setup behind every measured value: libcvc c1f104c0, MSVC 19.44, CMake 3.31.7, Ninja 1.12.1, cvcpkg 2.5.6, torch 2.8.0+cpu;
  • the Windows commands that were run, with the extra CMake flags and why each is there;
  • the Linux/macOS equivalents. These were not run verbatim (no Linux machine was available); libcvc's CI runs the suites under CTest on all three platforms for every PR;
  • each test's printed lines;
  • GRL-SNAM's pytest line, with its expected 53 passed, 2 skipped.
• §6.6, the learned plant's steering:
  • in the reference form, steering supplies power: 13.4% of |y·u| on the golden fit, and 5.6% on a fit to the analytic plant, which itself books none;
  • the opt-in modulated form from ph: steering as a modulation of D and G, opt-in, so the drive alone is the port transfix/libcvc#851: D(v, δ) and G(v, δ), with the drive alone as the port;
  • a new validation table for #851's tests.
• Corrected provenance: public CVC-Lab/LieSPHGP has no SE(2) vehicle. The port and its golden follow an SE(2) package built on it at 65db994, which is not public. LieSPHGP's own trainers use the EKF ELBO.
• Review fixes:
  • the dynamic plant shipped in 3.7.0 (#714) and the learned plant in 3.7.1 (#758);
  • ψ for the kinematic heading, since θ is the grade;
  • a "mixed" parity tier;
  • the sharp check's wording, and its value given once as 3.5e-5;
  • the health-check range;
  • test totals, the CTest regex and NavPotential.DriveForce*;
  • the GRL-SNAM paper's full title.

The wheel-frame front slip (149af06, 0020f2a)

The slip-switch sections, for transfix/libcvc#852 and #138:

• the abstract, §3's slips, Lemma 1 and its proof, Prop 1, and §4.2, §4.4 and §4.5: the floor term f_ε is now round-off at any steering angle;
• §4.7: the exact front tangent, and the test that holds it to the tire's derivative;
• §5.2: braking books −358 J at a 4e-6 residual;
• §8 and §10, the validation tables, Appendix A.3's printed lines, and the versions table;
• the Torch twin of the single-track pH form and its books; dyn_bike's front slip in its wheel's frame #138 rows cite its squash commit 2f76ab0, where the dynamic twins' parity tests pass (57 passed, 6 skipped for pycvc).

The navigator's point-mass books (86fabe2)

• Table 9 (the twins) adds Torch twin of libcvc's per-step energy books: book_step, the bound, the sharp check #139's torch twin of the point-mass books: book_step, energy_residual_bound, material_energy and integrate_surrogate_material's books hook, held to libcvc's NavPotential.BooksGolden.
• §8.3 records the point-mass books as twinned. The tick telemetry belongs to a running world, so it has no twin; pycvc: sim_world's drive telemetry, with the tick's books against U_nav transfix/libcvc#857 binds it in pycvc.
• §8.4 adds "Book the navigator's energy": how to request the books, the measured parity, and the curved-patch caveat.
• §10's "Python access" and the equation-to-code map are updated to match.

Elsewhere

• README.md's documentation line links the PDF.
• The guide's "See also" (§11) has a new bullet for the paper. Only the markdown changed: docs/cvc-nav-and-grl-snam-guide.pdf is not regenerated, because pandoc isn't available where this was built.

Open questions for Joe

• Naming the vehicle preset. The tests print the preset's name (libcvc's vehicle_desc::humvee()). The appendix summarizes those lines instead of quoting them; whether to name the preset is your call.
• Citing the SE(2) package. Where should it be cited? The note to the LieSPHGP authors asks the same.

[transfix]

I added 86fabe2 on this branch. It records two things in the paper: the navigator's point-mass books now have a torch twin (#139, merged), and pycvc can read a running world's tick telemetry (transfix/libcvc#857, on auto-merge).

• Table 1 (the twins): a new row for book_step, energy_residual_bound, material_energy and integrate_surrogate_material's books, held to libcvc's NavPotential.BooksGolden (nav: a golden for the per-step energy books, for GRL-SNAM's torch twin transfix/libcvc#849).
  • book_step on libcvc's arguments is held to 1e-12 (double tier).
  • The books are held to 1e-5 of the step's bound, and 2e-5 on the clamped rollout (float tier).
• §8.3, "What is not twinned yet": the point-mass books are twinned now. The tick telemetry (potential_end, projection_work, context_work) belongs to a running world, so it has no twin. #857 binds it in pycvc instead: nav_sim_world_set_capture_drive_telemetry and nav_sim_world_drive_telemetry.
• §8.4, a new paragraph, "Book the navigator's energy":
  • how to ask for the books;
  • the measured parity: 1.1e-6 and 8.3e-8 of the bound;
  • a caveat: on a patch whose gradient channels are not its values' gradients, each step books the gap between the force and -grad U. On NavPotential.Golden's curved patch that gap fills 99.65% of the bound at h = 0.002;
  • where a live world's context work comes from.
• §10, "Python access", and the equation-to-code map: updated to match.

The PDF was rebuilt with TeX Live pdflatex, three passes: 28 pages, no warnings. The citation of #857 assumes it merges before this PR does.