[ODE] Car stability with lowered centre of mass leads to problemswhen rolling

[James Bamford]

------- Forwarded message -------
From: "Jon Watte" <hplus at mindcontrol.org>
To: "James Bamford" <fromweb at jimtreats.com>
Subject: RE: [ODE] Car stability with lowered centre of mass leads to  
problemswhen rolling
Date: Mon, 4 Oct 2004 09:14:33 -0700


Because you sent this privately, I responded privately, but I'd appreciate
it if you forwarded this to the ODE list itself.

> steering amount at higher speeds slightly.. a problem i was getting was
> that at really fast speeds (well around 100mph perhaps) the turning  
> circle
> of the car was very poor.. i realise that a good part of this has to

Just like in real life! There's a reason they need 200 m(*) radius turns
on the NASCAR ovals -- when they add slope to the turns! If you want
to model reality, you have to accept the constraints of reality; one
such constraint is that you can't turn a lot when you go fast.

(*)I have no idea what the actual radius of the turns is, but it's pretty
big compared to a regular four-lane street that you can U-turn across.

> but it just fell far too exagerated... 90 degree turn taking many seconds
> to perform as the wheels are forced to be more or less straight.. i tried

Go to a racing track. Measure speeds, turn radii, and the time it
actually takes to take turns.

> increasing the force the joint uses to steer but it still didn't really
> have any effect.. this is definately something else i'm going to have to
> look at, its also something not present as much in some of the demos i've
> seen posted on the list so clearly you guys are sorting this issue out.

I just added active sway bars with really strong force, and very low
center of gravity, and limit the speed to about 85 km/h so that my
jeep won't tip over. If I go up even another 5, it will, at maximum
excursion :-) I use arrow keys and non-proportional steering, though,
for that arcade-like feel.

> The centre of gravity being outside of the visible geometry body is what
> makes it.. I may try raising this to being inline with the visible
> geometry and seeing if it improves further.. at the moment its not very
> good, especially when at high speeds your roll rate is very fast so the
> translation effect is magnified.

No, no, you don't want that. The center of gravity could be perhaps
10-15 centimeters above the undercarriage of the car. Also not that
most cars don't have all that much clearance to the ground; the
wheels are usually mounted at or above the center of gravity.

> I will have a go at this, i'd thought of doing something similar before..
> to be honest my geoms dont match up with the visible geometry all that
> well sometimes, i'm already using quite a low height geom as the main  
> body
> (0.5m-1m perhaps) trying to reduce this further though to being a plate
> seems sensible.. presumably its the equivelant mass thats created that
> will offer any improvements, can i not perhaps just tweak the mass  
> created
> without resizing the geometry to match?

Geom and Body are very different. In my car, I create a body at
position (0,0,0) and then create a geom that's a box that's size
(2,1.2,4) (width,height,length) and that's translated up by about 0.5
 from the body. I set the body mass to a box of the same size as the
geom, but centered on the body. The wheel centers attach at or above
the Y=0 line. You'll see that the geom bottom is at Y=-0.1, the geom
top is at Y=1.1, the wheel centers are at Y=0.0, and with a wheel
radius of 0.5 m, the clearance with no suspension compression is 0.15m.

> I definately want to improve the slip settings that i'm using.. i've
> looked over the carworld example source and just can't really understand
> what its doing to set the slip directions.. swizzling the axis etc.. all

Here's the problem: the wheel is rolling. I want a good fdir2, which
is implicitly derived from fdir1 and the contact normal. Because the
wheel is rolling, "front" of the wheel may be closely parallel to the
contact normal at any one time, so if that's the case, I use the "up"
of the wheel orientation instead. The end result is that "front" or
"up" is crossed with fdir1, to generate fdir2, which is sideways for
the wheel (unless I get a contact on the left/right side of the wheel).

> I've tried some of it.. and would definately like to try more.. I've seen
> something about anti-roll bars somewhere.. can these help!?

There are "active" anti-sway bars in the carworld jeep, but they add
a certain ponderousness to really drastic rolls. In fact, if you
don't like the roll beahvior of the jeep, the #1 improvement (to make
the rolls smoother) is to reduce the sway bar force, or just turn them
off.

> themselves when the car is rolling seem to corrupt anyways.. ie. the back
> wheels will temporarily go perpendicular to where they should be locked..

Did you turn on finite rotation axis, and make your suspension parameters
very stiff in the turn axis?


Last: I'm now playing with a car (tank, actually :-) based on a box
with four rays. It is MUCH harder to flip. In fact, it feels very
arcade-ey and good. It doesn't roll at all when turning, so I have
to detect and add that manually... Some later release of CarWorld
will allow you to choose vehicle to try them both out.


Cheers,

			/ h+



-- 
Using Opera's revolutionary e-mail client: http://www.opera.com/m2/