🚀 go-pugleaf

In article <1993Jan27.140858.29306@news.media.mit.edu>, fredm@media.mit.edu (Fred G Martin) writes:
|> In article <1993Jan27.043305.7387@sbcs.sunysb.edu> shane@cs.sunysb.edu
|> (Shane Bouslough) writes: 
|> 
|> >I've been thinking a lot about legs recently (don't we all? :-).
|> >If you're going to do a nice leg, say Brooksian, it seems like
|> >you'd need about one miniboard per. If you use 6 legs on your
|> >beastie, that's quite a few miniboards. Is that a good way to go?
|> >Or is something a bit more powerful a better idea?
|> 
|> In fact, this is just about what is done with Rod Brooks' robots:
|> there is a dedicated CPU per leg.  But, they use very small DIP CPU's
|> (maybe a 24-pin model), and surface mount motor controllers, so the
|> leg processor board is pretty small.  Also, it's kind of shaped like a
|> leg, long and skinny.


Actually, each leg on Attila and Hannibal has a dedicted 8751 on it that runs
three pwm servo loops, does sensor i/o, and communicates with the master
68000.  The only reason we really use this system is to save wiring.  With
20odd sensors per leg,  plus motors and power lines, we'd have needed
something like 40 wires per leg to do everything.  With the local micros, we
just have six (motor power/gnd, digital power/gnd, two wire serial
communications, the Signetics I2C in fact).  The 8751 boards have tiny FET
drivers for the motors--there are two boards piggy-backed, about 1.5 x 2
inches in size.


Were we to do it again, we'd definitely go with the Motorola parts--6811/16's
nand 683xx's, using the SCI bus.  The 6811's have much more flexibility, and
the communications bus is much faster.  Right now every bit is being used in
the I2C bus, and the 8751's are running a full load continuously (plus they
need a bunch of peripheral stuff for all the i/o).

6811 boards can be made pretty small too, at least as small as the 8751
boards we have (using surface mount parts).  Anita Flynn's "Squirt" robot has
a 6811 board on it that is just barely larger than the 6811.