Arcade driving force composition — source regression and real-engine QA

Status: development source-regression coverage only. The authoritative published version must be checked against GitHub Releases. This regression is not an installed Unreal 5.8 test and does not constitute an EXE/APK, cooked circuit or hardware qualification.

Why this integration test matters

Several independent engine-neutral policies already influence the same Chaos rigid-body car: throttle corner assist, tire heat/wear, asphalt-edge traction, service brake fade, physical chassis collision damage, engine heat, gear-aware motor taper, nitro and qualified rival drafting. Their individually bounded outputs must compose in a safe order.

The new ArcadeDriveComposition standalone C++20 test exercises real production policy headers together and mirrors the three AddForce motor paths without changing engine physics, gameplay, economy or saved data.

For a valid, forward-driving frame, the intended shared motor budget is:

applied force / healthy base force
  = qualified throttle × valid surface motor factor
    × retained impact motor factor × qualified engine heat factor
    × gear-aware motor taper × (1 + allocated nitro + draft)

Here allocated nitro + draft is at most 0.40. The existing allocator prioritizes finite nitro charge. Confirmed dirt reduces the entire motor budget to 78%; impact damage reduces it to no less than 79%. The original corner assist never increases positive throttle. Engine heat only derates motor power after qualified propulsion samples (down to 82%); a legitimate coast can recover it. Gear-aware taper smoothly removes additional propulsion near the established forward speed cap. Both scales apply equally to baseline, nitro and drafting, and neither adds throttle or makes official lap timing dependent on advisory telemetry. The real pawn must not drive after an invalid/blocked chassis or surface sample. Reverse and braking keep their existing separate guards.

The existing lateral resistance scale uses real tire condition, surface traction and damage multipliers, all bounded at or below one. Service brake heat only affects service braking and never grants motor power.

Automated checks

• The portable C++20 test covers baseline road power, combined nitro and draft, horizontal collision damage, confirmed dirt, shared engine heat fade and recovery, gear-aware half-power taper, combined grip, invalid physics/surface/thermal inputs, pause impact suspension, race reset, and 200,000 deterministic combined policy samples.
• The Python binding suite checks production chassis validation, ordered source observation, all three shared motor scales (including thermal and gear taper), brake fade branch, and reset/suspension lifecycle.
• No runtime force values or source physics are changed.

From the repository root:

cmake -S Tests/Standalone -B Tests/Standalone/build -DCMAKE_BUILD_TYPE=Release
cmake --build Tests/Standalone/build --parallel
ctest --test-dir Tests/Standalone/build --output-on-failure -R ArcadeDriveComposition
python3 -m unittest discover -s Tests/Editor -p '*Tests.py'

Required real-device acceptance

1. Compile licensed Unreal Engine 5.8 and UHT, cook the owned race map and package actual Win64 and Android candidates.
2. Verify all three motor paths in actual forward driving with nitro, real rival drafting, confirmed dirt and chassis damage simultaneously.
3. Check service brake heat, parking-brake floor, coasting, and recovery across pause, focus interruption, race restart and failed physics.
4. Repeat with Android touch, keyboard/gamepad and rival AI; compare actual 30/60/120-fps installed driving and physics traces.
5. Record engine build, source/map/package hashes and startup/QA logs.

This does not close the playable-game qualification gate.