[ODE] wheels attached at the wrong spot

[Ian McMeans]

I actually am using spheres, I'm just drawing them as cubes for now (it's
easier to see their orientation).

I've posted my source code recently, here is the relevant part for setting
up each wheel:
(LRfactor is -1 for wheels on the left, 1 for wheels on the right)

// If I change the arguments below to x,y,z, ODE blows up :/
// Why can't the wheel position be right on the anchor? I've seen example
code doing that.
dBodySetPosition(wheels[wheel], x + 0.5 * LRfactor, y, z);

shocks[wheel] = dJointCreateHinge2(world, 0);
dJointAttach(shocks[wheel], body, wheels[wheel]); // does the order of args
here matter?
dJointSetHinge2Anchor(shocks[wheel], x, y, z);

dJointSetHinge2Axis1(shocks[wheel], 0, 1, 0);
dJointSetHinge2Axis2(shocks[wheel], 1, 0, 0);


What really bothers me is that it seems the wheels aren't being constrained
in the way hinge2's are supposed to constrain wheels! The wheels should only
be able to move along axis 1 (the axis with suspension), and rotate about
axis 2, however in the pictures I showed you, the wheels are moving in the X
and Z axes, and if I enable gravity then they lean sideways and no longer
spin around their correct axis.

Here's what I'm talking about:
http://www.imageshack.us/my.php?loc=img1&image=broken1.JPG
How can that possibly be a hinge2? The wheels don't look constrained at all,
and there aren't even any forces acting on them because the car is floating
(no gravity) above the ground plane. And the wheels aren't oriented along
the right axis.

Car::Car(dWorldID world, dSpaceID world_space, Vector3d initialPos)
{
    Vector3d carSize(0.5, 0.2, 1);
    Vector3d wheelOffset(0.25, -0.5, 0.5);    // wheel offsets from the car
centre

    car_space = dSimpleSpaceCreate(world_space);

    // create the masses that we attach to the body and wheels
    dMass bodyMass, wheelMass;
    dMassSetBoxTotal(&bodyMass, 50, carSize.X(), carSize.Y(), carSize.Z());
    dMassSetSphereTotal(&wheelMass, 10, 0.2);

    // create the car body
    body = dBodyCreate(world);
    dBodySetMass(body, &bodyMass);
    dBodySetPosition(body, initialPos.X(), initialPos.Y(), initialPos.Z());
    dBodySetLinearVel(body, 0.05, 0.05, 0.05);

    // the car geometry
    box = dCreateBox(car_space, carSize.X(), carSize.Y(), carSize.Z());
    dGeomSetBody(box, body);

    // set up the wheels and their joints/geoms
    for (int wheel = 0; wheel < 4; wheel++)
    {
        wheels[wheel] = dBodyCreate(world);
        dBodySetMass(wheels[wheel], &wheelMass);

        wheelGeoms[wheel] = dCreateSphere(car_space, wheelOffset.X()/2.0);
// change to cylinders?
        dGeomSetBody(wheelGeoms[wheel], wheels[wheel]);

        // factors for knowing if the wheel is on the left or right
        float LRfactor = ((wheel == 0 || wheel == 3)?1:-1);
        float FBfactor = ((wheel == 0 || wheel == 1)?1:-1);

        dReal x = initialPos.X() + LRfactor * wheelOffset.X();    // wheel 0
and 3 on the right
        dReal y = initialPos.Y() + wheelOffset.Y();
        dReal z = initialPos.Z() + FBfactor * wheelOffset.Z();  // wheel 0
and 1 at the front

        // should we have the wheel right on the hinge2 anchor?
        // If I change the arguments to x,y,z, ODE blows up :/
        // Why can't the wheel position be right on the anchor? I've seen
example code doing that.
        dBodySetPosition(wheels[wheel], x + 0.5 * LRfactor, y, z);

        shocks[wheel] = dJointCreateHinge2(world, 0);
        dJointAttach(shocks[wheel], body, wheels[wheel]);    // does the
order of args here matter?
        dJointSetHinge2Anchor(shocks[wheel], x, y, z);

        dJointSetHinge2Axis1(shocks[wheel], 0, 1, 0);
        dJointSetHinge2Axis2(shocks[wheel], 1, 0, 0);

//        dBodySetFiniteRotationMode(wheels[wheel], 1);    // it apparently
makes wheels turn better

        if (wheel <= 1)        // front wheels can turn left and right
        {
            dJointSetHinge2Param(shocks[wheel], dParamHiStop, M_PI/6.0);
            dJointSetHinge2Param(shocks[wheel], dParamLoStop, -M_PI/6.0);
        }
        else
        {
            dJointSetHinge2Param(shocks[wheel], dParamHiStop, 0);    // no
back wheel steering
            dJointSetHinge2Param(shocks[wheel], dParamLoStop, 0);

            // the finite rotation axis needs to be constantly updated for
the front wheels :/
            dBodySetFiniteRotationAxis(wheels[wheel], 1, 0, 0);
        }

        // is there a way to make the suspension "bottom out"? It seems not,
        // unless we add an additional slider joint at each wheel.

        dJointSetHinge2Param(shocks[wheel], dParamFMax, 5);    // force to
steer
        dJointSetHinge2Param(shocks[wheel], dParamFMax2, 200);    // force
to accelerate

        dJointSetHinge2Param(shocks[wheel], dParamVel2, 2);

        // feel free to fiddle with the below two numbers - increasing one
means you should decrease the other, though
        dJointSetHinge2Param(shocks[wheel], dParamSuspensionERP, 0.9);
// "spring strength"
        dJointSetHinge2Param(shocks[wheel], dParamSuspensionCFM, 0.004);
// simulation "squishyness"
    }

}


----- Original Message ----- 
From: "akio" <cullum_a at bellsouth.net>
To: "Ian McMeans" <imcmeans at telus.net>; <ode at q12.org>
Sent: Saturday, April 03, 2004 10:49 AM
Subject: Re: [ODE] wheels attached at the wrong spot


> Those pics are great. Love the wheels, hehe.  Why don't you use spheres
> instead.
>
> You've not copied and pasted any configuration, so it's difficult to
analyze
> the problem but I will make some suggestions.
> One or two possible invalid configurations:
> 1) your body is configured as hinge2 - it'll make the body spin as if it
was
> a wheel (like the pic).
> 2) hinge2 axis alignment possibly could be off - all wheels must be the
same
> otherwise you'll see weird body movements.
>
> ----- Original Message ----- 
> From: "Ian McMeans" <imcmeans at telus.net>
> To: <ode at q12.org>
> Sent: Saturday, April 03, 2004 1:49 AM
> Subject: [ODE] wheels attached at the wrong spot
>
>
> > Here are two pictures that I think illustrate my problem:
> > http://www.imageshack.us/my.php?loc=img2&image=normal1.JPG
> >
> > This is how my car looks with no gravity. Once I press forward and the
> > wheels start spinning around their centres (IE in the normal way wheels
> > spin), my car looks like this:
> > http://www.imageshack.us/my.php?loc=img2&image=after1.JPG
> >
> > It appears the pairs of wheels are revolving around an axis at the
centre
> of
> > the car body, and the car body is spinning in the opposite direction (I
> > guess it's preserving linear momentum). Has anyone run into this weird
> issue
> > (or understand why it's happening ot me)? I'm using hinge2's, and the
each
> > wheel's joint's anchor is position at the centre of the wheel.
> >
> > _______________________________________________
> > ODE mailing list
> > ODE at q12.org
> > http://q12.org/mailman/listinfo/ode
>