physics, weight transfert

[ghiboz]

hi all!
i've solved the problem of the physics feeling before (i've configured the car correctly, i've tested also the car inside vdrift and is ok), but now I've another little question.... the car seems that doesn't affected to the weight transfert when I try to make a scandinavian flick (http://en.wikipedia.org/wiki/Scandinavian_flick) the car doesn't do this or better, the car is affected always by a understeer, and never by a oversteer..

regards

[NaN]

You could try to reduce the suspension bounce, rebound values. Play a bit with the stiffness, antiroll setting.

Btw do you mind to share your config? Can you name the track or upload a pic of the curve you are testing on?

Some detail info:

The suspension model is completely kinematic(no chance to have something like anti-dive geometry).
Tire contact forces are applied at wheel center. I've tried to fix this by moving the application point to the contact, assuming a very rigid suspension. I called it the Elchtest(moose test) mod at that time. It broke about half of the cars, let them roll over the side due their mass center. But there has been visible, feelable mass transfer hehe.
There are also issues due to rigid body and spring/damper model. You have 1500kg falling down each time step and being pushed up by the spring/damper where the damping forces are working against the movement. You can see it while running "vdrift -debug". In the real world the damping values should be zero for a car standing still. I've got some experimental code to deal with this, but it will most probably break car handling too.

Maybe it is time for another special mod, we need a better model for the karts anyway.

[ghiboz]

thanks nan!
i'm trying with g4 and xs at estoril, first turn...
another question, have you got an idea to calculate the longitudinal and lateral g-force?
thanks

[ghiboz]

another thing about the weights: example, the CO car have this

Code:

[particle.00]
mass = 36.0
position = 0.0, -1.25, -0.43

[particle.01]
mass = 155.0
position = 0.5, -1.3, -0.13

[particle.02]
mass = 155.0
position = -0.5, -1.3, -0.13

[particle.03]
mass = 155.0
position = 0.5, 1.4, -0.13

[particle.04]
mass = 155.0
position = -0.5, 1.4, -0.13

but the function

Code:

    // load children bodies
    for (PTree::const_iterator it = cfg.begin(); it != cfg.end(); ++it)
    {
        if (!loadBody(it->second, error_output)) return false;
    }

doesn't load these mass... is right? these values doesn't affect nothing? thanks in advance!


mmm... the release seems that this is considered 1100kg instead of 475...

[NaN]

have you got an idea to calculate the longitudinal and lateral g-force?

In cardynamics updateAction() before updating linear_velocity = body->getLinearVelocity() calculate acceleration = velocity_delta / dt; and transform it into car space.

All mass values are accumulated. If you are running the latest code make sure you are using the latest data as there might be not backward compatible changes.

[ghiboz]

have you got an idea to calculate the longitudinal and lateral g-force?

In cardynamics updateAction() before updating linear_velocity = body->getLinearVelocity() calculate acceleration = velocity_delta / dt; and transform it into car space.

All mass values are accumulated. If you are running the latest code make sure you are using the latest data as there might be not backward compatible changes.

ok! thank you very much!, and where I can found the latest data?

edit: http://vdrift.svn.sourceforge.net/viewvc.../CO/CO.car
found it!
but, just for information, the github repository is only for the code, and sourceforge for the data?

[NaN]

You will want to check out the whole data, can use tortoisesvn on windows.

There have been proposals to move data to github too.

[ghiboz]

have you got an idea to calculate the longitudinal and lateral g-force?

In cardynamics updateAction() before updating linear_velocity = body->getLinearVelocity() calculate acceleration = velocity_delta / dt; and transform it into car space.

All mass values are accumulated. If you are running the latest code make sure you are using the latest data as there might be not backward compatible changes.

just a thing, the fuel mass is calculated or not? the consumption reduce the total mass of the car? :wink:

[NaN]

Not since the switch to bullet. Patches are welcome.

If you are interested in implementing it start with simple mass reduction first(assuming fuel tank at center of mass). Then try to take fuel tank position into account(you will have to recalculate center of mass and inertia tensor).

[ghiboz]

before updating linear_velocity = body->getLinearVelocity() calculate acceleration = velocity_delta / dt; and transform it into car space.

sorry, i'm looking now, but i'm not too familiar with the btRigidBody.... how can I transform getLinearVelocity into car space? I need to multiply with GetOrientation()?

thanks

[NaN]

The velocity is in world space. You will want to use the inverse rotation. If you want to use quaternions it is the negative. If you want to go with bullets transform matrix use the transpose of the rotation matrix.

transform.getBasis().transpose()

or

body->getCenterOfMassTransform().getBasis().transpose()

[ghiboz]

thanks, so tontransform the linear velocity from world space i use this piece.. i'll try!!

thanks

[vdriftfan]

Now I don't know if I should start a new thread for this but anyhow...
While the engine , transmission etc. simulation is great, I find the suspension behavior on a lot of cars a bit unrealistic. Most of them are very stiff. When the wheel losses contact with the surface, the rebound is instantaneous. I mean too fast. Also when driving on the track La Conca you'll find some tiles(like those in the margin of sidewalks) that you should climb over. Suspension behavior is different between the stable release and the nightly.
In the first that car passes over like it should, the wheels climb then fall, but the car doesn't bounce at all when coming down not even on low speeds. The wheels hit the road lightning fast.
In the nightly it barely goes over even if it's obvious it should pass over, and the car gets stuck over it or it rebounds as it had hit a wall.

Here is a link to Speed Dreams simulation engines code which behaves really good, somewhat like Race 07. Check it out.

http://speed-dreams.svn.sourceforge.net/...ules/simu/