> -----Original Message-----
> From: David VanHorn [SMTP:dvanhorn@CEDAR.NET]
> Sent: Thursday, February 07, 2002 1:52 AM
> To:   PICLIST@MITVMA.MIT.EDU
> Subject:      Re: [PIC]: controller
>
> >
> >I'm thinking along the lines of a vision based system.
>
> You may not need full vision, if you can sense the line at multiple
> points.
> The far out ones would tell you what's coming up, and the nearest would
> tell you where you are. You should be able to feed that into a PID sort of
> system pretty easily.
>
I initialy thought this was one way to go, but if you consider the speeds
that the car is going to achieve, the control system is going to have to
anticipate the corners well before it gets to them in order to slow down.
At 110KPH you are travelling at over 30 m/s.  Even some huge 10 meter
extension in front of the car is only going to give you 0.3 seconds advance
information which somehow dosen't seem enough (IMO).  Perhaps this could
work if none of the corners were fairly shallow?

If the system has to be completely self contained within the car, a vision
system is probably the way to go.  With a high contrast line, a cheap B/W
CCD camera would likely be adequate.  The processing is way beyond the
average 8 bit micro though, I would be looking at embedded PC's at this
stage, but of course cost will be an issue.  If it's ruggedised enough then
perhaps it could be made re-usable?

OTOH you could have beacons around the course indicating the approach to a
bend.  A coded RF data stream could tell the car what speed it should be
hitting in X meters and the onboard control system would have to sort out
the braking/deceleration.  A similar beacon would be used to accelerate the
car after exiting the corner.

The requirement for one sytem to work in all cars is going to require some
kind of adaptive control system, perhaps a self tuning PID controller?

Regards

Mike

--
http://www.piclist.com hint: The list server can filter out subtopics
(like ads or off topics) for you. See http://www.piclist.com/#topics