Just to clear up some confusion here. The Maxelbot localization module uses
"trilateration", not "triangulation". With triangulation you compute with angles.
With trilateration you compute with distances. Trilateration is easy to visualize
as the intersection of 3 circles of different (probably) radii.
So, each Maxelbot has 3 acoustic transducers. All three can listen. The center
one can also emit. Each Maxelbot takes turns "pinging" acoustically. When it
pings it also emits an RF pulse. When the other Maxelbots hear the RF pulse,
they start counting. When they hear the acoustic ping at each transducer,
that gives three different times-of-flight. These are converted to distances because
one knows the speed of sound. Then trilateration equations use the three distances
to give the range and bearing to the pinging Maxelbot. Parabolic cones focus the
acoustic energy into the horizontal plane.
Hence, for each pinging Maxelbot, all other Maxelbots that can hear it will know where
it is.
A token is passed on the RF, so that each Maxelbot will take turns pinging.
Furthermore, the RF can be modulated to also send information. Hence, localization
is coupled with data exchange. This can be very powerful for various applications.
For example, we recently have had 5 Maxelbots self-organize into a chain formation
in an L-shaped environment. The Maxelbots then send information back down the
chain, on the RF link. The first Maxelbot sends all this information to a laptop, that
displays an estimation of the shape of the environment, based on the relative positions
of the robots with respect to each other.
There are variations that can be done. It could all be done acoustically. It can also
be done all in RF (this option is being built in Florida).
Why are we doing it this way? First, cameras are difficult to deal with and require
light sources, more CPU power, and potentially omni-directional mirrors (which crowd
the objects towards the edge of the image). LEDs work pretty well, and can be modulated
to provide for information exchange also. For our tasks, we needed quite a bit more accuracy.
Judging distance based on received LED intensity is problematic, especially in outdoor conditions
with lots of dust, rain and snow. This is not to say it can't be done, but we decided not to
adopt the LED approach. We especially did not want light sources, which are not stealthy.
We also did not use IR because it would require a ring of them (as mentioned) and our
calculations showed it would take more power and provide less accuracy.
As mentioned by others, not all parts are available off the shelf. This is why it took 2 years of
research to get it working well. However, at this point, if you have our schematics, you could
have someone like PCB Express make the boards for you. The components are off the shelf for
that. The only specialty items are the parabolic cones, which are machined. If you have a lathe you
could make a wooden cone that works pretty well. You could even make the inverse, and then
pour liquid plastic to make the cones.
If people are interested, we are willing to try to assemble instructions for the hobbyists. Perhaps
the UW machine shop could even make cones if people are interested (they have the program
to do it, and it is easy for them to do).
I hope this helps. Please send email to me if you have more questions or would like more
information. Thanks for your interest!
[email protected]