bullet dynamics

[joevenzon]

I fixed the tire squeal sound bug in R2726.

[alex25]

I fixed the tire squeal sound bug in R2726.

thanks joe.

--alex--

[NaN]

We've got proper suspension simulation now taking wheel mass and tire stiffness into account(r2736)! The stiffness and mass are hardcoded atm. This will change in the near future.

[joevenzon]

Wow, nice job! Seems to work just fine.

[stan.distortion]

Noticed something with the speed effect on steering, it seems to effect the linearity not the max angle. Dropping the max-angle value in the .car file is making slides a lot more controllable for me.

[portets]

Noticed something with the speed effect on steering, it seems to effect the linearity not the max angle. Dropping the max-angle value in the .car file is making slides a lot more controllable for me.

i notice this too. i think i started noticing it a couple of months back

[NaN]

So it was not an issue before? I think there have been no changes to the steering code. It might be related to my cardynamics rewrite.

[stan.distortion]

Never noticed it to this degree before, dont see anything different in the code but the cars seem nearly impossible to slide controllably.
It had me banging my head on the wall trying to get tire params to feel right, as soon as the rear steps out it either spins or snaps the other way if the slide is caught.
Will try and hard code something into carcontrolmap_local.cpp tonight and if it works try and add an extra slider for max angle sensitivity under the speed sensitivity slider in the GUI.

[joevenzon]

The speed sensitive steering changes were from this thread:
http://vdrift.net/Forum/viewtopic.php?t=1173

Feel free to tweak with the algorithm (carcontrolmap_local.cpp around line 594) and post results to the above thread.

[stan.distortion]

Something I probably should have mentioned earlier with the tyre's, bullets btSoftbody would be very well suited to the job. Not dug down deep into the model yet but what I have seen so far suggests the model is sophisticated enough to take the place of the pacejka calculations.
NaN, you mentioned the bullet suspension code. Are you working on it at the mo or did you mean it's on the cards? I was planning on going back to it but would like to look into softbody tyres too.

[NaN]

I am working on the suspension, still in the experimentation phase. The idea is to use four custom constraints between car and track(world) to simulate the suspension/wheels.

I strongly suggest to grab bullet source. The demos are great to try things out:

Code:

svn checkout http://bullet.googlecode.com/svn/trunk/

Softbody tires!? I don't share your optimism towards btSoftBody. But I'd love to be taught otherwise, even if it doesn't work in real-time.

[NaN]

Have got the constraints working. The only problem now is that bullets damping is not viscous. It is scaling the spring force or something(damping=0.0 => infinite damping damping=1.0 => no damping). Don't have an idea how to solve this. Wouldn't want to write my own constraint solver, lol.

Edit:
The damping is scaling the target velocity of the constraint which is calculated from spring force due to displacement.

[NaN]

I checked in normal force dependent rolling resistance code(r2788).

Our rolling resistance model approximates tire rolling resistance by a quadratic function(rolling-resistance = 1.3e-2, 6.5e-6):


Real world rolling resistance depends on tire temperature due to heat generated by rolling resistance(tire/surface deformation):


We are somewhat off, but i think it's OK.

Due to this changes the rolling resistance factor(coefficient) of some tracks might need adjustments in surfaces.txt(or objects/list.txt). I've also interpreted/implemented rolling drag as viscous wheel drag(sand, mud and stuff) proportional to wheel velocity.

[NaN]

Had a silly bug in rolling resistance calculation. The track parameters are OK, no need to change them.

[NaN]

Have been working on the suspension recently, rewrote bullet-vdrift synchronization code and hit a bug in the suspension implementation. The force(blue) applied to car is suspension force(red) projected onto surface normal(green). This is wrong as the force created along the suspension hinge is missing(yellow). The force onto car decreases with the car rolling.