rec.autos.simulators

I've been thinking about suspension geometry lately and the path that
it prescribes on the hub centre. Assuming that we don't want any
forward / backward motion of the hub centre with up and down
movements, the path of the hub centre with relation to the body is a
2d path. Assuming that this is correct, rather than waste all of that
CPU power on constraining the links of the suspension in order to
remove the 5 DOFs necessary, why not just move the wheel / hub through
this path?

Just food for thought, I know that Gregor is calculating all of the
links realtime, and from the research that I have done into rigid body
dynamics, closed-loop constraint systems are quite expensive CPU wise.

Is there something that I am blatantely missing?

Colin

  Not missing anything at all as far as I can tell.  I'm doing it just as
you've described.  The only possible drawback that I can see is that
acceleration laterally of the wheel in relation to the body from lateral force
at the contact patch (it can arc towards/away from the body since it's a 2-D
path now) isn't accurately calculated as it would be with methods like Gregor
uses.  My method isn't as perfectly correct as it could be, but since
compliance isn't modelled too often, even in really high end simulation, it's
negligable enough to forget about in comparison imo.  The only time I can see
it making a difference is on suspension systems that are so wildly out of
typical design that the wheels can travel further along the track width axis
than they can vertically.  Nobody does this in real cars anyway, so I plan to
follow my current course.

Todd Wasson
---
Performance Simulations
Drag Racing and Top Speed Prediction
Software
http://PerformanceSimulations.Com

-Gregor

Well, seeing as I'm going for a 'game' simulation I don't think I
would really have to worry about the lateral forces mentioned above.
But if I do it right, I'll be able to have realtime editable
suspension geometry with camber / caster / WPI / toe etc all
calculated automatically :-)

Colin

  Yes, the force effect is negligable.  Even the SAE stuff online rarely tries
to take it into account, and this is stuffed aimed at real life vehicle design
for the most part.

  To get caster changes, you'll need to do forward and rearward motion of the
wheel too.  I've got camber and track width change with wheel travel, but toe
change (bump steer) will require tie rod ends to be calculated too.  Not
something I've done just yet.

  Be warned, allowing the players to set up the car gets much more complicated
to deal with when you go this route.  Also, a lot of what people have learned
about front/rear roll stiffness can go out the window when dynamic camber
starts changing.

Todd Wasson
---
Performance Simulations
Drag Racing and Top Speed Prediction
Software
http://PerformanceSimulations.Com

My reason is that I don't only want to use it for suspension, but if I just
develop a method of generally linking and constraining rigid bodies together
(either with rigid links or springs) then it can also be used for other
things (eg towing a trailer, or towing another car using tow rope). It also
means it's not just limited to vehicle simulation, but any physics system
involving linked rigid bodies.

Just thought I'd add another perspective onto things.

Mike.

You can still have fore/aft movement of the hub due to bushing compliance under
acceleration and deceleration. I wouldn't model it in a Formula Ford sim,
because of the low power and spherical bearings in the suspension, but a Gran
Turismoish game with hopped up street cars would be more realistic if it was
included.

Bushing compliance is why static toe-out is so great on the front end of FWD
street cars - they toe-in under power and toe-out under braking.

-Scott Johnson
 "There is nothing, I think, more unfortunate than
  to have soft, chubby, fat-looking children who go
  to watch their school play basketball every Saturday
  and regard that as their week's exercise."
                       - John F. Kennedy, 1962

If you don't mind me asking Gregor, what method do you use for the
constraints, I've looked into generalized coords and constraint forces
but neither handle closed systems very well (plus I'm not very good
with large matrices :-(). I've read some stuff on Lagrangian
multipliers but that went straight over my head :-}

I think I'll stick with the 2d method for the while until I find a
system that satisfies my requirements.

The best 'feeling' suspension that I have done so far was a system of
two REALLY stiff springs for the two lower links with a macpherson
strut connecting the wheel and body. 'Twas absolutely lovely, until
you hit a wall or steep slope  and the wheels exploded because the
springs became unstable. (:-( deep sigh)

I also had a system using the Dynamo dynamics package which looked
really realistic but for some reason when specifying the inertia
tensor using the correct method it would blow up and bodies of about
1000kg would make it totally upset :-(

Colin

P.s. Todd, when are we gonna be seeing some graphical shots of your
sim, I'm really interested :-)

I therefore do as if I were calculating static loads on the suspension
(a matrix inversion), and after I get the load on each of the links, the
load on the spring-damper link gives the rate of change of the length of
that link. I also take into account the rate of change on the steering
link, and then I need to (iteratively, three iterations suffice) solve a
nonlinear set of equations to find the velocity and angular velocity of
the wheel mount. After that, I update the position and orientation of
the wheel mount based on these velocities, but again need to solve a
nonlinear system of equations to take into account any errors due to
initial linear movement of the wheel mount. I repeat the procedure then.

By not taking into account the mass of the wheel I'm not actually
loosing much, and certainly nothing about car response. The only thing I
am missing here is the high-frequency wheel hopping mode, as this
one is eliminated. I could introduce accelerations of the wheel along
the instantaneous travel axis, and as such take wheel hopping into
account, but that would be somewhat faking it as not all forces would be
properly taken into account should the link forces interest you (but
would be almost perfectly well for almost any other application). I'll
probably do a full Lagrangian picture in the future, it is definitely on
my to-do list, but not anytime soon. I need to learn how to program more
efficiently first :).

Cheers,

-Gregor

Out of curiosity, since I'm not into coding stuff, how much are you
guys worrying about roll centre location and movement?  Though that
too could be pre-mapped in 2D...

Jonny

  Funny you should ask.  I just put this up in this thread.  Don't get e***d,
it's pretty ugly :-)  I'll put up a couple more from outside one of the cars
sometime soon.

http://www.racesimcentral.net/
Todd Wasson
---
Performance Simulations
Drag Racing and Top Speed Prediction
Software
http://www.racesimcentral.net/

  I'm very concerned with it.  After I get a few bugs worked out with what I
have here and get some other stuff done, roll centers are going to be
calculated on the fly in real time.
Todd Wasson
---
Performance Simulations
Drag Racing and Top Speed Prediction
Software
http://PerformanceSimulations.Com

The other thing that I was thinking about last night is that because 5
dofs are removed from a wheel that means I can effectively model the
wheels with weighted 'fat' particles taking into account only the
vertical velocity of the wheel, once the position has been decided,
calculate the orientation.

I really wish I had more time to work on this stuff, I am in crunch
mode right now to finish my 'other' game for the GameCube launch here
in Japan. Still I manage to get about an hour a day (whilst eating
lunch) so that's not so bad.

Colin

Indeed. If you manage to find a path for the vehicle, and the
orientation on the path, you solved the problem completely. The
Lagrangian multilpliers are only used when you cannot solve for the
explicit path and orientation, but if you can do that, you are set to
rock :). It may be hard to do with the general suspensions, but with
some compromises I'm sure it's possible.

Be aware, though, that if you want to do things really properly, you
must also take the changes of orientation and subsequent changes of dhe
direction wheel angular momentum into account, and that is probably the
hardest part (I do handle that in my sim, but I neglect the actuall mass
of the wheels, which is indeed far less relevant). If you neglect this,
though, things are much easier.

Good luck with your game! Man, you've gone Japanese, haven't you,
cramming all free time activities (such as eating etc. :) ) into one
hour ;).

Cheers,

-Gregor

-Gregor