Road → Traversal → Excitation → Full-Car
VehicleLab 1.6.0 adds a sampled-grid source to the verified procedural Road/Traversal engine. Road/Traversal milestone 1.5.0-alpha.2 and derivation vehiclelab.road-to-corner-excitation v0.1.0 remain unchanged. A straight forward constant-speed path locates nominal tire contacts; absolute road elevations become datum-relative inputs to the geometry-coupled passive Full-Car model.
The reference cases use independently coded formulas and geometry. Every sampled elevation equals its reference exactly (maximum absolute error 0 m). RD002 records a 0.75 s axle delay for a 3 m wheelbase at 4 m/s. These are numerical verification results, not external vehicle or road-test correlation.
| Case | Scenario | Independent acceptance |
|---|---|---|
| RD001 | Flat road | Absolute elevation at all four contacts |
| RD002 | Full-width raised cosine | Independent waveform and wheelbase/speed axle delay |
| RD003 | Left-only feature | Positive road t supports the left wheel tracks |
| RD004 | Lateral offset | Shifted traversal changes support membership |
| RD005 | Speed scaling | Doubled speed halves encounter duration with unchanged amplitude |
| RD006 | Signed negative feature | Negative amplitude and exact baseline return |
| RD007 | Periodic sine | Independent elevation formula at every contact |
| RD008 | Boundaries and holds | Exact feature endpoints, start/end and endpoint holds |
| RD009 | Unequal geometry | Independent front/rear axle-arm and track offsets |
| RD010 | Determinism | Repeated Float64 bytes, provenance and corner ordering |
OpenCRG import verification: CRG001–CRG016
The independently versioned sampled-grid Road variant extends road-definition.v1. Importer vehiclelab.opencrg-importer v0.1.0 and supported subset v1 normalize one complete straight-reference grid into Float64 metre elevations. Bilinear interpolation uses the full grid; strict inclusive bounds reject outside queries. Source SHA-256 identifies exact bytes; normalized Road SHA-256 identifies physical grid and interpolation semantics. Road/Traversal, Road-to-corner derivation, passive Full-Car equations and RK4 identities are unchanged.
| Case | Acceptance |
|---|---|
| CRG001–003 | ASCII LRFI/LDFI, big-endian binary KRBI/KDBI, dimensions, row/column ordering and wrapped records. |
| CRG004–007 | Exact nodes including signed zero; independent longitudinal, lateral and bilinear equations (1e-15 m bound). |
| CRG008–009 | u forward, v left, z up; explicit m/cm/mm conversion; deterministic source-byte and engineering identities with filename/comments excluded from physical meaning. |
| CRG010–011 | All four nominal contact paths inside domain, exact boundaries and rejection of truncated, overlong, missing and nonfinite payloads before simulation. |
| CRG012–013 | Curvature, nonzero slope/bank and unsupported modifiers/options rejected; diagnostic-only and zero-valued strict options explicitly classified. |
| CRG014–015 | Equivalent LDFI/KDBI surfaces produce exact complete Full-Car channels and metrics; external references, map projection and unknown meaningful records rejected. |
| CRG016 | 224 independent queries against unmodified official ASAM 2.0.1 C evaluation, across four formats, bilinear/rough fixtures and two unchanged external samples: maximum observed error 8.5e-17 m; predeclared absolute bound 1e-7 m. |
CRG016 evaluates the loaded source frame without C API default repositioning or modifiers. Its bound accounts for the reference C API storing elevation channels as binary32 even for double source formats. The source is the official v2.0.1 release, commit 2530b4e711989acbcfa520e2b99034976afc589c. No ASAM implementation or third-party sample is vendored. Original VehicleLab synthetic fixtures provide reproducible local regression.
An unchanged binary engineering road from Project Chrono, detrended_rms_course_1in.crg, passed the production file-picker, Run, Drive over Road and all four engineering exports. Its header identifies detrended KRC data. The 122,836-byte KRBI source contains 10,096 × 3 nodes over s = 0–504.75 m and t = −3–3 m, with elevations −0.077184–0.123819 m. The reviewed run starts at s = 110 m, travels at 4 m/s for 5 s, and uses the default Generic Full-Car with zero lateral offset. ASAM's measured Belgian-block sample was also selected in the production UI and explicitly rejected because its reference line curves. No curved road is flattened to pass this subset.
Browser checks cover file-selection and Worker-response races, invalid-file retention, source switching, exact freshness restoration, all 41 imported-road result/grid buffers, six plots, 16 metrics, five cameras, four viewport widths, WebGL failure and exports. Procedural merged-main records retain exact bytes in separate Node and Chromium baselines. Imported roads remain unsaved session evidence; provenance export requires the original source for reconstruction. Numerical agreement and file interoperability do not establish ASAM certification or external road/vehicle physical validation.
Full-Car bridge and evidence transport
Flat road at its own datum gives exact zero response from zero state. Symmetric full-width cases give exact zero roll; a left feature gives the verified positive left-side-up initial roll. All 34 complete result channels agree with equivalent prescribed temporal inputs, including the non-divisible final step. The representation-only acceptance bound remains 1e-12; the recorded equivalence case has zero error.
The excitation schedule covers every literal RK4 start, midpoint and end-stage request, including the final short step, without nearest-sample matching. Direct and browser Worker runs preserve all Float64 arrays, 16 metrics and provenance exactly. All 39 distinct result/excitation buffers transfer once and detach at the sender.
Display and physical evidence boundaries
A stationary, capped display mesh evaluates the same selected road but never supplies solver input. Completed response interpolation and traversal use one playback clock. Cameras, display gain and body artwork cannot change numerical evidence.
Flat, raised-cosine and sine surfaces, with full-width or finite raised-cosine lateral support, and the documented OpenCRG subset are available. No OpenDRIVE/OpenSCENARIO ingestion, stochastic standards, steering, yaw, lateral or longitudinal dynamics, calibrated tire radius/spin, tire contact/enveloping/slip, nonlinear suspension or active control is claimed. Physical correlation has not been performed. Completed Spatial evidence can publish local Drive-over-Road engineering video. Spatial project persistence, Rig Full-Car video, Compare, Explore and Frequency remain unavailable.
Configure Full-Car sources and Drive over Road Motion · Full-Car numerical evidence · Evidence boundaries