At 10:53 PM 10/6/97 -0700, Karl F Bohringer wrote:
>Hi Christopher,
>
>we have been working on sorting strategies with smart surfaces for some
>time now.  Check out our '96 and '97 ICRA papers, or the January-March
>issue of "IEEE Computer Science and Engineering", there we show some
>strategies on how to separate parts (without sensor feedback) based on
>part size.  This is also explained in our web site at
>http://www.cs.cornell.edu/home/karl/MicroManipulation

I've seen some sorting stuff that sounds similar, but I'll check it out.
But I'm not so much interested in sorting, as in avoidance of emergent
properties like traffic jams.  The application is nanotech-based, with
trillions of one-micron parts over several square meters of area.  The parts
are all the same size, and would be moved by smart "cilia" robot arms that
can sense the part type and move in an arbitrary direction.  I'm trying to
figure out communication and computation requirements to move the parts from
place to place on the surface.  The parts are actually containers that get
filled and drained at various special points on the surface.  But product A
may need to be shipped from (0, 0) to (100000, 100000), while B has to go
from (0, 1000000) to (100000, 0).

>Motion planning with multiple objects has been investigated by several
>robotics researchers.  Latombe's book on robot motion planning has a good
>overview on this area.

I took a robot motion class from Latombe about six years ago.  I don't
remember anything that would help with this class of problem.  Certainly any
centralized strategy is doomed.  A hierarchical strategy might work, or a
local strategy--but I don't really trust local ones, and my coauthor doesn't
like hierarchical.

>       title = " What Programmable Vector Fields Can (and Cannot) Do:
>                Vector Field Algorithms for {MEMS} Arrays and
>                Vibratory Parts Feeders",

This one sounds interesting.  Does it talk about avoiding gridlock?
Crossing streams?

Thanks,
Chris
--
Chris Phoenix   cphoenix@best.com
Work (Reading Research Council): 650-692-8990;    Home/voicemail: 415-860-1536