[Avery Andrews 920101.2030]
(a little discussion which it seems appropriate to put on the net. I
asked Bill Powers whether there was any evidence as to whether
pointing/reaching behavior used slowed proportional integrators
in the output, or direct control of velocity via differentiators in
the perceptual circuits. said that it makes no difference, since
in either case, control of velocity is achieved; I reply ...)
Yes, the gross behavior would be the same, but (a) there might be subtle
differences (b) the actual circuitry would be different. With
differentiators for example, an error in, say joint angle would set
a reference level for joint angular velocity, so the diagram would
look like this:
joint position ref.
>
v
------> comparator---->sigmoid(?) output unit
> >
> v
>--->differentiator----->comparator-->---
> >
> v
joint possie alpha-gamma effer.
sensor
Which is different from what appears in Arm, I think.
So I still think there's an empirical issue to be settled here.
Avery.Andrews@anu.edu.au