[Martin Taylor 2018.07.15.09.16]
[From Rupert Young (2018.07.15 12.30)]
Here's a short description of control within robotics which, I
think, typifies the conventional approach to control,
On the surface it seems to express many of the characteristics
of PCT; goal-based, closed loop, feedback, error reduction.
Why, do you think, it is not perceptual control?
I have read many of the articles on the web site, but I haven't been
able to find out a couple of key points, the most salient of which
is how the levels of control interact with each other. Is the action
output of a high level subsystem accomplished by sending references
down through the levels, for example. Some of the write-up seems to
suggest that the levels work autonomously, and the action output is
independently programmed. I get the impression, but it may be a
biased impression, that the method is functionally equivalent to the
inverse kinematics approach to producing desired movements, rather
than the PCT approach exemplified in Powers’s Little Man and Arm2
demos, of letting lower-level systems evolve do the work effectively
and speedily. It’s hard to tell from the articles I read whether
this is a fair comment, but I would judge that the gestalt robotics
approach probably is not PCT.
Now, as to whether it is or is not perceptual control, that's a
different matter. Clearly it is, because everything is designed to
bring a perception, such as of a cup having been correctly grasped
by its handle, to a prescribed reference value.
Martin
···
https://www.gestalt-robotics.com/advanced-control.html