[From: Chris Love (920527.1000)]
[To: Bill Powers (920526)]
Thanks Bill for "clearing up" the model details. And yes, I was
using a basic engineering control theory text, "Modern Control
Systems", Richard C. Dorf. I got a little confused from what I was
taught and what you are doing, but your explanation was very
clear and I feel my understanding has improved a lot.
I see your point about the random reorganization method also. I
never thought to consider that,
raising gain may lower error up to a point, and *then* make it
increase again.
I will also give the random dot experiment a try. Should be easy
enough to implement in Prograph.
I wish I could attend your PCT conference because I know I
would really enjoy it. I will try to plan for it next year. I didn't
know about it soon enough.
···
------------------------------------------------
[To : Tom Bourbon (920527)]
Hi Tom,
I donot yet have my Little baby up and running. What I do have
though is an early Beta version of him. The intent of this version
was to get the environment up and running and to check the angle
relationships in the arms, etc.. I suppose it was, more or less, a
debugging version. The entire graphical display is functional with
all the buttons and stuff. The baby does not use any PCT in this
version. What it does use is three *rules* to change the arm angles.
The first rule says that if the target is to the left of the finger
tip then swing the arm (at the shoulder) left (and vice versa).
The second rule says if the target is above the fingertip then
move the arm (at the shoulder) up (and vice versa).
The last rule says if the target is further out (depth) than the
fingertip then bend the elbow outwards (and vice versa).
As you might imagine, these rules are hardly sufficient to
correctly map the space but it does work well in *nice* regions,
i.e., don't put the target above his head or at his toes!
This version has been compiled and is executable. I'm sorry that
my PCT version is not running yet. Hopefully soon!
Since this Beta version does not use PCT (but is interesting all
the same) I will leave it up to you whether you would like to show
it. I suppose it could be used to show the idea of what I'm doing
and give an impression of what it will look like. What I could do
though is also provide a few 8.5X11 page pictures of the
*proposed* structure of the current model???
Let me know what you think of this. If you feel it would not be
warranted to discuss, no problem (no offence taken - it is kind of
early in the development cycle still).
Thanks for the offer,
Chris.