Re: Barcodes & error detection

In lugnet.robotics, Geoffrey Hyde wrote:

> how much development did you do making the area where
> the light sensors themselves are mounted as black and
> light-proofed as possible... If you have some fairly
> clear close-up photos of the sensor and lighting
> assembly, perhaps you could post them somewhere.

  I'll have to take some small photos and post them to my Brickshelf account,
and I should really include seom response curves from the various lightsensors
so you can "see" what the sensor is seeing during a barcode read.

> Have you thought about making some track correction
> routines that can use the light readings so as to
> correct it's own errors while driving around -
> I'd have thought that any programmable bot ought to
> be able to spot when an error is coming up to
> unacceptable tolerances and try a course correction
> routine to negate or minimise the error. 

   Hmmm. I'm not sure I understand here. Generally, the forklift follows the
line quite well placing the "outrigger" lightsensor that's reading the barcode
firmly in the middle of the barcode. But if for some reason it doesn't how can
the RCX tell? How can it tell a "misreading" of a barcode from just a different
barcode? This is why I brought up the question in the first place in terms of
error detection & (perhaps) correction.

> In fact most good line follower bots should be able
> to navigate any strip and be accurate all the time,

   Define "accurate". The forklift never looses the line, and generally can not
be seen to even "hunt" for the line (watch the SMART Crate Contraption
forklifts; Gus is significantly smoother in linefollowing than them).
Arbitrarily high levels of precision (narrow tolerance) don't translate into
high accuracy either (a bright area applies a bias, shifting the forklift
slightly, on the order of less than 1 stud, perpendicular to the line edge).
   Can I make it more accurate? Perhaps, but only at the cost of being slower.
Line following is easier the slower you go... but watching a forklift creep
along isn't much fun either ;-).

   Perhaps one thing that's spawning confusion: there are *two* light sensors,
one doing the line following, and a 2nd dedicated to looking for barcode
commands. I've tried reading barcodes along the very edge of the line using one
light sensor for both barcode reading & line following (the SMART group does
something like this), but it has its own issues, specificly knowing when to
"dead-reckon" forward reading a barcode, and when to use the sensor information
to keep the forklift on the line edge.
   also, it makes "deployable barcodes" (barcodes that can be swung into the
forklift barcode readings FOV dynamicly by other mechanisms "requesting service"
from the forklift) tougher as well.

-- 
Brian Davis