Wheel Settings mLongitudinalFriction

[Ono-Sendai]

Hi,
It seems that the wheel friction behaviour has changed in a recent Jolt version (am updating), so am looking at mLongitudinalFriction again. it's confusing to me though - the commented behaviour is

"///< On the Y-axis: friction in the forward direction of the tire. Friction is normally between 0 (no friction) and 1 (full friction) although friction can be a little bit higher than 1 because of the profile of a tire. On the X-axis: the slip ratio (fraction) defined as (omega_wheel * r_wheel - v_longitudinal) / |v_longitudinal|. You can see slip ratio as the amount the wheel is spinning relative to the floor: 0 means the wheel has full traction and is rolling perfectly in sync with the ground, 1 is for example when the wheel is locked and sliding over the ground."

But shouldn't friction be highest when there is no slipping? e.g. when the slip ratio is zero?
As it is the default curve goes through the origin, e.g. friction is zero when slip ratio is zero:

mLongitudinalFriction.Reserve(3);
mLongitudinalFriction.AddPoint(0.0f, 0.0f);
mLongitudinalFriction.AddPoint(0.06f, 1.2f);
mLongitudinalFriction.AddPoint(0.2f, 1.0f);

Or am I misunderstanding something?

[Ono-Sendai]

Tested with a flat curve for mLongitudinalFriction and it seems to work better, no wheel skidding when accelerating. So maybe this is a straight-up bug?

[jrouwe]

There was a bug in that system before, see:

JoltPhysics/Docs/APIChanges.md

Line 36 in 21ba054

* 20240203 - Longitudinal friction impulse for wheeled/tracked vehicles could become much higher than the calculated max because each iteration it was clamped to the max friction impulse which meant the total friction impulse could be PhysicsSettings::mNumVelocitySteps times too high. In case this breaks your vehicle, the new max tire impulse callback can be used to restore the old behavior, see [the vehicle constraint test](https://github.com/jrouwe/JoltPhysics/blob/a456b244aa2ad2ce0a8124d27823377ed0b1c4b4/Samples/Tests/Vehicle/VehicleConstraintTest.cpp#L156-L164). (a456b244aa2ad2ce0a8124d27823377ed0b1c4b4)

There's a workaround to get the old behavior back:

https://github.com/jrouwe/JoltPhysics/blob/a456b244aa2ad2ce0a8124d27823377ed0b1c4b4/Samples/Tests/Vehicle/VehicleConstraintTest.cpp#L156-L164).%20(a456b244aa2ad2ce0a8124d27823377ed0b1c4b4

I agree that the wheels are skidding much more than they used to do, but that's explainable as the friction was 'num solver iterations' times higher than what was configured. I also thought the new behavior felt buggy, so I configured a vehicle in PhysX and compared it with Jolt and the results looked very similar. The setup was to have a vehicle in 1st gear and to floor the gas pedal for 5 seconds:

I don't remember the meaning of all these stats anymore (forgot to write that down), but I think Ffront/Fback is the force of the wheels on the ground, RPM is engine speed, Front/Back are probably wheel rotation speeds, Vel is vehicle velocity, Slip is the slip ratio of the driven wheels.

[jrouwe]

And to answer your question:

But shouldn't friction be highest when there is no slipping? e.g. when the slip ratio is zero?

When slip ratio is zero, the wheel is rotating to match the ground speed perfectly. Friction can only exist if they move at different speeds (otherwise the rubber of the tire does not experience any longitudinal forces from the road), so the curve should be zero for a slip ratio of zero.

[Ono-Sendai]

Ah ok I was thinking it was the friction coefficient, instead of the friction force.

[jrouwe]

It is a friction coefficient, but the friction coefficient is dependent on the relative motion between the rubber and the road (i.e. non-linear friction).