Skip to content

Vehicle G-force telemetry — source candidate, not a playable release

Player feature

LateBrake can now estimate signed longitudinal and lateral chassis G from consecutive real world-space Chaos velocity samples. This creates useful racing feedback without adding driver assists or changing vehicle handling. A positive longitudinal reading means forward acceleration, negative means deceleration; positive/negative lateral readings indicate acceleration toward the car's right/left axis.

On sufficiently spacious desktop HUDs, the optional CHASSIS G (EST) line shows smoothed signed longitudinal and lateral G. After a valid finish the paged Results menu can display the run's peak braking G and peak cornering G. Both are advisory session-only estimates. The game does not turn them into race rewards, records, rankings, garage credits, saved progress or driving instructions. Very small screens prioritize authentic speed, clock, gates and driving controls.

Physics and continuity policy

  • The player pawn must pass the existing real start, circuit, restart placement, physics readiness and focus/menu state gates before taking samples; a stopped red-light frame is not telemetry.
  • Observations use the actual chassis world XY linear velocity in cm/s. Differentiated velocity is projected onto the vehicle's observed yaw for forward/right acceleration. Local speed differences alone would miss real constant-speed cornering acceleration.
  • Unreal Tick samples preceding frame-to-frame velocities. A first valid sample after startup or any interruption arms a baseline but earns no acceleration result.
  • Samples outside 1–250 ms, with nonfinite/malformed velocity or yaw, with speed above 360 km/h, or with resultant acceleration above 6g are rejected; continuity and live displays clear. A collision or teleport spike must not become the session's best cornering reading.
  • Live G readings use bounded first-order smoothing (~120 ms). Peak statistics are from accepted raw derivatives rather than smoothed HUD motion; they are not independently calibrated to sensors.
  • A peak braking value requires an observed negative forward-axis acceleration while actual BRAKE input is held, and while moving forward at least 2.8 m/s. Passive engine drag and reversing do not count. Cornering peaks require at least 5 m/s of planar speed.
  • Pause, OS focus, blocked physics and paid recovery interrupt samples; an accepted race restart resets all readings and peaks. A recovery does not erase earlier legitimate peaks from that run.

Source and tests

  • Source/Latebrake/Public/Racing/VehicleGForceMeter.h — portable bounded C++20 differentiator, filtered readouts and session peaks.
  • Source/Latebrake/Private/Vehicles/LatebrakeCarPawn.cpp and its header — sample the actual player chassis only on valid physics frames; reset and suspend around interruption/recovery.
  • Source/Latebrake/Private/UI/LatebrakeRaceHUD.cpp — optional expanded HUD telemetry and honest finished-results page.
  • Tests/Standalone/VehicleGForceMeterTests.cpp — yaw/world-axis correctness, braking conditions, low-speed suppression, NaN/Inf, dropped frames, teleport-style impulses, resets and 20k sample bounded-state stress.
  • Tests/Editor/VehicleGForceBindingTests.py — source-only Unreal attachment and gameplay-authority boundary checks.

Run on a developer system with a C++20 compiler:

cmake -S Tests/Standalone -B Tests/Standalone/build -DCMAKE_BUILD_TYPE=Release
cmake --build Tests/Standalone/build --config Release
ctest --test-dir Tests/Standalone/build -C Release -R '^VehicleGForceMeter$' --output-on-failure
python3 -m unittest discover -s Tests/Editor -p 'VehicleGForceBindingTests.py' -v

No source-only CI test compiles Unreal Engine 5.8/UHT, cooks the real circuit, runs the Win64 game or certifies hardware G force. Actual vehicle suspension and tire physics are still prototype-level and need careful validation against real Chaos world velocities. Measure HUD layout at desktop sizes, all handling presets, pause/restart/focus, rotation, brake/reverse, collisions, teleports and long frame hitches. Keep raw and observed game logs and link the installed candidate's SHA256 to playable-release gate #148.

Do not label this an automotive safety system, precise accelerometer, human driver rating, or demonstrated gameplay until engine/device testing has actually happened.