> This sounds like an interesting possibility. Rather than sensing colour
> per se, perhaps a whole suite of ground-looking sensors (including IR
> and ultrasound) would give us enough variables to reliably identify the
> type terrain.
Multi sensor systems usually perform better. Ultrasonic is very good for
determining hardness (as evidenced by my old boat sonar). If we can
fix the minimum distance problem we might be able to use it.
I did build a sonar system last year that had a range of about 3 metres,
but it also had a minimum of about 2 inches. I might do a few
experiments
with that later on. I used the transducer pairs that you can buy cheaply
at
most electronics stores. The hardware consisted of an LF353 configured
for about X200 with bandpass filtering around 40khz. The rest of the
circuit ran in software on an ST6 micro.
> What are we likely to encounter? Grass, obviously, (but
> different varieties and lengths will probably exhibit somewhat different
> properties), earth, tree root, paving stones, tarmac, brick, concrete.
> Most of hte man-made surfaces are hard and flat and so should give a
> good ultrasound echo.
That's what I think. My boat sensor could detect sand and rock, so I'm
sure
we could detect grass and concrete.
>
> btw Dave, why do your posts have such erratic word-wrap?
Let me know how this one came out. I use the email part of Netscape for
all
my email. I've changed it to variable width fonts, hope this helps.
Dave Everett.