rec.autos.simulators

As you might know, GPL uses some FF prediction time (I believe it
defaults at 0.018) to make up for the fact that most FF devices have a
bit of latency that is noticable.

I'm trying to think of ways to get something similar. Mainly because
hopefully this year I'll connect Racer to a high-end steering wheel
(one which is powerful enough to rip your arm off, hehe) and I think
this will be one of the bigger problems.

I believe most racing games have had to deal with this; small back &
forth motions once you pull the wheel a little and then let it go. It
may spring back & forth when the actual force is not in sync with the
simulation itself.

The only thing I could come up with sofar is a prediction based on the
derivate of Mz (aligning moment). So:
- keep last_Mz
- calculate new Mz
- Mz'=(last_Mz-Mz)/dt
- predict FF as: Mz_wheel=Mz+Mz'*prediction_time

In other words, from the last Mz change you predict ahead (with a
user-configurable 'prediction_time') where Mz is going to be once it
actually reaches the wheel.

I don't know how GPL does it; you might think of really putting the
car ahead of time and calculating its tire forces there, giving a
future Mz, and then retreating the car back to the realtime position,
but that seems very costly.

Anybody have any cunning ideas? :)

Ruud van Gaal
Free car sim: http://www.racesimcentral.net/
Pencil art  : http://www.racesimcentral.net/

That's how i'd do it, but I havent really thought about it yet :)

Give it a go you have nearly written it already :P

All the best,
Ash

IMHO, the right solution (from Hard Drivin' & Race Drivin') is to integrate
the FF servo-control (PID loop?) into the vehicle model.  Then the FF
device is fully under your control.  After all, that is where the steering
is in a real car...

Yes this will be a pita, but the results are so much better than any
"consumer" FF wheel that I've tried, that you owe it to yourself to try it
once this way.

-- Doug Milliken
   www.millikenresearch.com/olleyfl.html  <-- review of new book

Well, you are actual trying to predict a timeseries: so you may use a Neural
Network, to try something fancy ;-), just an idea...  This shouldn't be too
costly if you're not training it constantly.

Cheers,
Tom

So that involves UDP as well, and is indeed really high-end.
We'll see what happens. :)

What's a PID loop?

Ruud van Gaal
Free car sim: http://www.racer.nl/
Pencil art  : http://www.marketgraph.nl/gallery/

It would be interesting to have some REAL figures about how much
latency a standard controller has.

If we talk about 0.018 seconds...hmmm... that cannot be a real
problem.. if we get into 0.1 seconds it will become a problem, in both
ways, u'll feel it on the wheel and it will be very difficult to
predict.

I tell you what...
I pass... I'll stick in real time blaming wheels makers for their poor
response time ;-)..

Stefano Casillo
driving.multiplayer.it/netkar

Problem is, even regularly, that for stiff cars the wheel might
oscillate by itself when not damped by the driver. Try a F1-type car,
drive a bit fast, then pull the wheel one side and let go. Chances are
it will go and oscillate.

Good choice. :)

Might be USB drivers which are slow, or the wheels themselves. Anyway,
once you get to high-end wheels (and high-end companies/products) it
might be something that needs more than just a big bunch of damping.
;-)

Ruud van Gaal
Free car sim: http://www.racer.nl/
Pencil art  : http://www.marketgraph.nl/gallery/

Someone else suggested that latency of 0.018 seconds should not be a
problem, but I disagree.  It typically requires a factor of 10 in servo
loops to really get good control over the next stage (much worse than the
"times-two" factor suggested by theory -- Nyquist Rule).  Kind of like
this:

  Performance of human driver -- 2-5Hz
    Senna was recorded at 5Hz back when he drove for Lotus F1
  Electro-mechanical steering wheel needs x 10 = 20-50 Hz response
    to be acceptable to the driver
  Digital control of analog steering wheel motor x 10 = 200-500 Hz updates,
    this dictates the minimum speed of the physics model

I'd be happy if anyone with a real control theory background would
jump in at this point...

Proportional-Integral-Derivative -- common type of servo feedback
controller, different gains set on each of the three terms.  I don't
actually remember what kind of controller was used in Hard Drivin'.

-- Doug Milliken
   www.millikenresearch.com