Actual chassis ground-contact traction — source and installed-game QA

SOURCE-ONLY Unreal Engine 5.8 C++ integration. NOT a playable, qualified UE5.8 build. The existing vehicle is a pitch/roll-locked single simulated Chaos cube chassis, not a genuine per-wheel spring/damper, articulated suspension or Chaos wheeled vehicle.

Why this changes actual vehicle gameplay

Previously the physical player/rival pawn used only planar chassis velocity and a rough circuit XY road mask to apply engine, braking, lateral tire-like force and yaw torque. The source could therefore produce strong in-air turning, acceleration and tire grip despite having no actual supporting physical road beneath the vehicle.

This change samples four real Unreal line traces from the current physical chassis transform (approximately front-left, front-right, rear-left and rear-right) down onto blocking WorldStatic surfaces. The car's own actor is ignored. Only well-formed trace hits below the chassis within 115 cm and with upward-facing surface normal Z ≥ 0.68 count as ground support.

The same source-driven traction scale is used by both human and autonomous physical pawns:

Qualified supporting probes	Contact-based drive/grip scale
0 or 1	0.00 — airborne/insufficient support
2	0.45
3	0.75
4	1.00

This scales physical baseline/nitro/drafting propulsion, opposite throttle braking, service/handbrake resistance, lateral tire-like force and yaw torque without overwriting actual Chaos velocity, applying artificial suspension springs or modifying gravity. Raw free-flight motion, air resistance, chassis gravity/collisions and race-clock authority stay managed by their existing systems.

Airborne frames do not earn extra synthetic tire wear, new brake-pad heat from unsupported braking or consume a fresh nitro burst. The existing pit-service eligibility also requires contact.

A four-point trace is NOT a wheel contact manifold. No force is applied at wheel location, pitch/roll remain locked, normal load is not integrated, and the current visual cylinder wheels are still non-colliding decorations. A future real suspension effort needs separate UE wheel components, contact tuning and real builds.

Cross-platform source qualification

• ChassisGroundContactTractionTests.cpp verifies all 0–4 support counts, near-ground full driving, airborne contact removal, steep wall rejection, two-probe road-edge response, invalid/paused conditions and 80,000 varied measured road-contact frames.
• UnrealChassisGroundContactBindingTests.py checks the real C++ pawn calls four Unreal WorldStatic line traces, rejects its own chassis, verifies hit normals/depths and gates all actual force paths before the motor/turn controller.

These are deterministic portable C++ and source-text tests. None of them uses licensed Unreal SDK compilation.

Real UE 5.8 hardware acceptance tests — mandatory

1. Compile the real project on an authorized UE5.8 Windows workstation. Cook the original project map; verify four asphalt slabs/ground are actually queryable ECC_WorldStatic and physics-blocking.
2. At GO and after each legitimate recovery, confirm gravity settles the chassis near asphalt, all four traces qualify, and the human/rival can still start, steer, brake and finish valid gates.
3. Drive slowly across the shoulder and road edge; verify partial support reduces actual force without stalling a car that remains reliably supported. Check inside a pit stop and restart.
4. Deliberately leave a raised surface using a safe test fixture; verify no midair motor/nitro/steering impulses and recovery of physical grip after landing. Check 30, 60, 120 FPS and physics substepping. An unbuilt source test cannot prove this.
5. Verify wall/side collisions do not become false grounding contacts, visually elevated cars are not wrongly accelerated, and slopes behave sensibly. Tune query ranges/normal thresholds only against real traces and actual chassis collision dimensions.
6. Test all six car visuals and both handling modes with true rendered material/collision geometry, plus brake heat, damage, AI corner braking and lane-overtake interaction.
7. Confirm no save, gate, reward or race session authority changed.

The authentic installed-game release gate is still issue #148.

Observational live ground-contact readout (PR #292)

The expanded desktop in-race HUD now includes a ground-contact readout: GROUND ROAD 4/4 | TRACTION 100% when four real traces accept WorldStatic support, GROUND PARTIAL 2/4 | TRACTION 45% or 3/4 | TRACTION 75% for partial contact, and GROUND AIRBORNE 0/4 | TRACTION 0% (or 1/4) with insufficient road support. The percentages exactly match the current accepted traction multiplier, not a new estimate of wheel load, tire friction or gravitational normal force. The trace count is not an actual physical wheel-support count.

This readout is available only during an accepted live pawn physics frame on a sufficiently large expanded HUD. Invalid frames, focus loss, player restart, unpossession and resets clear cached values. It is hidden on compact/touch layouts and does not claim race position, apply forces, adjust a lap, persist user data or bypass the previous authoritative clock. Its HUD formatting is verified by static Unreal bindings and the pure ChassisContactReadoutTests.cpp target (25,000 consistent accepted states plus malformed input rejection). Running those tests is not proof of displayed on-screen pixels in a licensed UE5.8 editor or installed Windows game.