On 24 Sep 2006 18:14:14 -0700, "G_dir" wrote:

I was thinking of having the robot o forward (with no control for
that), and using the beacon to correct my course.

Hi OM,

Still very simple. A single, broad beam of light. A dual sensor that
is aimed "bore sight" to the direction of the robot's forward motion.
Even with RF at 5 GHz, the ability to discriminate is going to be very
poor. When both sensors see the light, equal displacement to the
steering (they cancel). When one sensor sees the light, then a
correction to steering until both see the light (they add).

If the robot thrashes (zig-zag) down a practice course, then reduce
the amount of steering the sensors add. If the robot loses track of
the beam, then turn up the amount of steering the sensors add. If the
correction makes slow looping curves, put horse blinders on the
sensors so they can only see a 10 degree field of view and turn up the
gain.

The best solution will always have some error in it (correction
circuits have to have something to work against). This means the
optimal design given your constraints will have some weave. It will
travel 22 to 24 meters to finish that 21 meter course.

73's
Richard Clark, KB7QHC