[ODE] nonholonomic vehicle path following

[STenyaK (Bruno González)]

On 6/11/07, Ke <kxubb at yahoo.com> wrote:
> After tuning up the parameters like steering,linear
> velocity, our car can reach one target position w/
> less than 5% error. But that set of parameters may not
> give us good error control for other testing cases.But
> we need to find a good set of parameters good for
> general cases. Looks we have to tune the control
> parameters case by case.
>
> > You should not copy the steering values from one car
> > to the other
> > verbatim, but calculate the difference to the
> > desired path and apply
> > that to the car. This way you will have a
> > self-correcting steering. I
> > hope this is stable, which depends on how big the
> > errors are from
> > step to step.

Wheels will always slide a bit, so even when the front wheels point to
the correct point, there will probably exist some understeer that will
prevent the car from steering enough to reach the target.

If that's the case, try to feed it wrong data (make the car think it
has to go further to the left/right than it really has). Then maybe
apply some dampening to the steering wheel, so that it doesn't swerve
left-right like mad when it's actually following the desired path
correctly.

With some luck, that'll work for most situations, not needing to
fine-tune each case. Hope it helps :)

-- 
Saludos,
     STenyaK

_______________________________________________
Msn/Jabber: stenyak AT gmail.com
ICQ: 153709484
Webs: http://1ksurvivor.homeip.net 1kSurvivor
http://motorsport-sim.org Motorsport
http://emuletutorial.info EmuleTutorial