Open the experiment
In Studio, choose Test with the Full Car context, expand Simulation Model & Advanced Analysis and select 14-DOF Coupled Transient — Experimental. The normal Full Car option continues to mean 7 DOF.
Uses a generic, independently qualified 14-DOF reference vehicle. Current applied Vehicle assembly is not yet mapped to this model. This is an unmeasured educational reference without experimental vehicle correlation.
Physical model and scenarios
The 28-scalar state contains world position, right-handed ZYX body attitude, four physical suspension jounces, four spin phases and their coordinate derivatives. World axes are +X forward, +Y left and +Z up. Physical body angular velocity is exported in body axes separately from Euler coordinate rates.
The unchanged ML-59 kernel supplies coupled body and wheel inertia, nonlinear inertial bias, gravity, installed springs and dampers, physical linkage geometry and internal drive/brake reactions. Four compliant vertical supports are recomputed at every stage. No slip, traction, steering actuator, imported road or applied-assembly coupling is supplied.
A preserves the solved flat equilibrium. B applies a vertical force at the sprung CG. C adds a vertical FL contact force separately from baseline support. D applies signed internal FL rotor/carrier torque, beginning with +10 rad/s FL spin. Pulses follow sin⁴ between 20% and 80% of duration, with zero forcing outside.
Qualified operating envelope
Use 0.04–0.4 s and fixed RK4 steps of 1, 2 or 4 ms. At most 400 steps and 401 samples are allowed. Output every 1, 2 or 4 steps must exactly divide the duration. Signed amplitude bounds are ±1000 N body, ±250 N FL contact and ±25 Nm wheel torque. Static amplitude is zero.
Four supports must stay loaded; contact switches are rejected. The model checks travel, geometry, chart, finite values, mass conditioning and algebraic residual at each stage. Small-motion limits are 0.15 rad attitude, 0.5 m/s body velocity, 1 rad/s attitude rates, 1 m/s jounce rates and 20 rad/s spin rates. Failed or budget-limited runs do not become completed evidence.
Numerical evidence
The separate Cartesian/Newton–Euler reference uses extrapolated modified midpoint integration, independently checked at sixth and eighth order. Actual state histories are compared on a common grid. Fixed 4/2/1 ms refinement demonstrates approximately fourth-order behavior for the smooth forced cases. Static preservation, signed excitation work, damping loss, energy balance, support potential consistency and envelope extrema are checked separately.
Download the canonical numerical verification evidence. These checks establish the stated numerical cases and envelope, not experimental correlation or accuracy for arbitrary vehicle configurations.
Solver Diagnostics v1
Results retain the completed physical plots and metrics first. Diagnostics report simulated duration, measured integration wall duration, real-time factor (wall / simulated), actual dynamics evaluation count and mean evaluation cost. RTF above one is slower than real time. A fixed RK4 run uses exactly four dynamics evaluations per step.
Compilation, solved-equilibrium initialization and result assembly have separate timings. The integration wall interval includes support and validation, dynamics, RK4 updates, output sampling, progress and cooperative yields. Per-evaluation timing includes stage support and validation. Browser transfer and rendering are outside these Worker measurements. A low algebraic residual is solve health; it is not a trajectory convergence claim.
Ownership, cancellation and exports
A dedicated Worker owns each run. Cancel, leaving the experiment and project replacement terminate active work; late messages cannot publish results. An input edit marks earlier completed evidence stale. Cancelled or failed reruns retain that earlier record with its original run identity.
Configuration and results are session only and excluded from workspace-project.v2. Save/Open continues to describe the applied Vehicle, Ride and Handling. Export the experimental numerical CSV, Solver Diagnostics JSON and exact model/input provenance separately. Exports never rerun the solver. There is no experimental 7-DOF animation adapter or new PDF/video format.
Performance baseline
The unoptimized mechanics can be slow and allocation heavy. Versioned static/body/corner/torque benchmark cases preserve numerical identities, exact settings, raw timings, source fingerprints and runtime/memory observations for ML-61. Timing is descriptive and is not a scientific pass/fail target. Optimization and adaptive integrators are deferred.