Newsgroups: comp.robotics
Path: brunix!uunet!microsoft!wingnut!petesk
From: petesk@microsoft.com (Pete Skelly)
Subject: Re: Need info on four legged robots
Message-ID: <1992Jul18.194109.14908@microsoft.com>
Date: 18 Jul 92 19:41:09 GMT
Organization: Microsoft Corporation
References: <19077@ector.cs.purdue.edu> <1992Jul16.212647.11722@eagle.lerc.nasa.gov> <19089@ector.cs.purdue.edu>
Distribution: usa
Lines: 61

In article <19089@ector.cs.purdue.edu> bouma@cs.purdue.EDU (William J. Bouma) writes:
>
>In article <1992Jul16.212647.11722@eagle.lerc.nasa.gov>, smjeff@lerc05.lerc.nasa.gov (Jeff Miller) writes:
>|> In article <19077@ector.cs.purdue.edu> bouma@cs.purdue.EDU (William J. Bouma) writes:
>|> >  read, but I object to the deceptive (too general) title.  He should have
>|> >  called it "Legged Robots that Hop"!
>|> 
>|> What you are describing is the difference between statically stable
>|> legged locomotion and dynamically stable legged locomotion.
>
>    It seems to me there is more to dynamically stable walking than just
>    bouncing around on pogo-stick limbs.

Yes, there is a bit more.  In most dynamically stable walking systems, you
end up with multiple legs on the ground at once, and therefore have to deal
with closed loop kenimatic chains.  However, Raiberts works are "classics"
when it comes to robotic walking (esp dynamic).  Don't blow them off as
being just a bunch of robots bouncing around on pogo-stick limbs.
>
>|> If anything balance has more to do with the dynamic case than the
>|> static case.
>
>    How did you figure that out?  I think it has everything to do with
>    both cases.

Well, I think it was implied that controlled balance had more to do with
the dynamic case.  In static walking, the robots balance in much the same
way your kitchen table balances.  There is no computation going on in
balancing besides path planning to make sure the robot doesn't need to try
to balance.  For info on dynamic balancing, look at some of the articles
on inverted pendulums (I don't have any references here with me now, but
maybe someone else could post them).
>
>|> You don't have a valid complaint. (You don't have a leg to stand on.)
>
>    All I am saying is that the book only covers mechanical running with
>    a rather simple leg model.  It does not cover walking or even standing,
>    both of which require balance. 

Both of which require computed balance in the dynamic case.  In static
walking, No CPU cycles are wasted on balancing.  The only rule to follow
is: Keep the center of mass within the area formed by the contact points
with the gound.

As for saying that the book "Only" covers mechanical running, what other kind
of running is there.  It's rather like saying an atom bomb will "Only" blow
up a small portion of a city.  Running is much more difficult in many cases,
especially compared to standing.  Balancing software is pretty easy to write
(see inverted pendulum).  Walking is difficult, but in running, you lose
contact with the ground.  After that, you have NO control for some time, and
then when you do get control, it's only for a short period of time.

(alright, maybe you could shift your weight around or something to get some
control).

>-- 
>Bill <bouma@cs.purdue.edu> 

petesk@microsoft.com
My Opinions

