[From Bill Powers (920503.1300)]
(to Chris Love, cdlove@ben.dciem.dnd.ca, who works with Martin Taylor)
Hello Chris,
Welcome to CSGnet. That's all there is to it. To sign on, you send a
message to this address (from the logon you're going to use):
listserv@vmd.cso.uiuc.edu
The message should look like this:
subscribe csg-l lastname, firstname, location
[total of <= 78 characters]
When you send this message, the listserver will automatically pick up your
return address from the electronic header and put you on the subscriber
list. You don't have to ask anyone's permission -- just do it. You'll get
some introductory messages.
I'm posting this to CSGnet as well as sending it to you; there are other
modelers on the net who will be interested in your reply, so we can conduct
this conversation in public (if you don't mind). Perhaps your first post to
CSG net could be your message to me, followed by your reply to this.
Details: Our custom is to insert by hand the name of the sender and the
date/time sent, as in the first line of this message. We use European-style
dating: year, month, day. When quoting replies, I copy them into the text
and put ">" before each line by hand. Some mail systems do that for you
automatically, but I write on a word-processor off-line, and do it by hand.
You can do it any way you want.
···
-------------------------------------------------------------
Only today though have I been able to make my "little baby" follow the
target about. He does this very slowly though, but at least he moves >his
finger in the right direction!
Congratulations, proud papa. There's no thrill like seeing your own brain-
child start to work properly.
I've gone a long way down this road; my current version simulates muscles
with stretch and tendon reflex, a higher level of kinesthetic control, and
a visual system that detects target and finger position in three dimensions
by optical ray tracing (of a very simple sort, but legitimate) and uses the
kinethetic systems to place the fingertip on the target or follow it around
(or even draw circles around it). At present the model is running at about
1/5 of real time on my 10 MHz AT-compatible. I'm using a full kinematic
model for the arm, with two degrees of freedom at the shoulder and one at
the elbow. At this very moment I'm sweating out a system for learning the
correspondence between the shoulder-based kinesthetic control actions and
the eye-based visual perception of positions (turns out to be a lot harder
to do than to describe).
My model, like yours, has the Little Man pinned to a stationary position,
with only the arm moving. The head can actually also move to keep the
target centered visually, but the head movements turn out to cause
trajectories of arm movement that deviate unacceptably (to me and Greg
Williams) from published trajectories of free movements in real arms. I
could put in a visual compensation system, and will eventually do so (or
Joe Lubin and students at Princeton will), but for now I've frozen the head
and solved the problem that way.
Martin told me before he left on his trip that you have found
an effective way to move the arm quickly without running into the
oscillation problems due to overshoot. Do I have my information
correct, and if so, is it difficult to explain?
Before I start giving you a lot of unnecessary advice, perhaps you'd better
fill me in on the organization of your model. I wouldn't want to waste your
time telling you a lot of things you already know. Yes, my model is quite
stable and can move rapidly (in simulated time); a one-radian movement to a
settled position can occur in 0.1 simulated second (dt in the model is 0.01
second). That's at the kinesthetic level. Under visual guidance, the
movement takes about 0.3 seconds min. Of course it can move more slowly.
With all the parameters trimmed properly, it's a critically-damped
movement, with no overshoot. Of course real people do show some overshoot
when they move fast, so in simulating real arm movements we wouldn't use
the optimum parameters.
Glad to have you aboard.
Best,
Bill Powers