>
>
It's just my feeling about it, but I don't think you can ever be
certain what you're getting is reliable, when using multiple echos
from wide-field measurements. Maybe, if the unit is in an environment
with very very sparse distribution of objects. Not the typical indoor
room. But, have fun, in any case.
>
> The point with multiple readings (I never got 17, and only the
first five or
> six were ever meaningful) was that I could process multiple hits at
once.
> With two eyes, had the accuracy been ok, then the distance was
sufficient to
> allow a positive fix in 2D space by using a bit of geometry (keep
in mind, I
> had two computers receiving and processing sensor data, so
processing power
> wasn't an issue). This would take care of the wide cone issue,
unless both
> sensors got echoes off of different things that were at about the
same
> distance, but off to different sides. This scenario, too, could be
handled,
> as all echoes would be written to a DB, and another process would be
> studying the echoes and putting them into objects. Under that
scenario,
> three to five objects would provide a pretty good fix, even in
transit.
>
> A narrow cone on a pivot is a viable alternative, but a whole
different
> concept, so I'll have to think about that.
>
>
> dan michaels wrote:
> >
> >
> > Actually, I don't think I would try to do this with SRF0x units.
They
> > have wide beams, and how will you ever know that any 2 echos are
from
> > the same object?
> >
> > I think you'd do better with some kind of narrow beam device,
which
> > you could pan on a servo, in order to localize objects. Eg, the
Sharp
> > sensors have very narrow beams, and you could correlate the
readings
> > with the sonar distance readings. Also, some of the newer
Maxsonars
> > are supposed to have narrower beams.
> >
> > I also don't see the point of having 17 echo returns, as IIRC,
they
> > are all from objects at different distances with the SRF0x, and
you
> > seem to want to ID a specific object.
> >
> > What I would do is mount a couple of different types of sensor,
> > narrow- and wide-field, on the servo, and pan the array, and then
> > cross-correlate the results for consistency. I might even use a
CdS
> > cell with a tube on the front to narrow the FOV, as most objects
will
> > have a different reflectance than the background or nearby
objects.
> >
> > If you read Joe Jones' book, Robot Programming, he spends a lot
of
> > time talking about how sensor readings are generally unreliable,
and
> > you need a lot of cross-comparison capability.
> >
> >
> >
>
> --
> View this message in context: http://www.nabble.com/New-range-
finder-needed-tp15923461p15945933.html
> Sent from the OOPic mailing list archive at Nabble.com.
>
Yahoo! Groups Links
<*> To visit your group on the web, go to:
http://groups.yahoo.com/group/oopic/
<*> Your email settings:
Individual Email | Traditional
<*> To change settings online go to:
http://groups.yahoo.com/group/oopic/join
(Yahoo! ID required)
<*> To change settings via email:
mailto:oopic-digest@yahoogroups.com
mailto:oopic-fullfeatured@yahoogroups.com
<*> To unsubscribe from this group, send an email to:
oopic-unsubscribe@yahoogroups.com
<*> Your use of Yahoo! Groups is subject to:
No comments:
Post a Comment