[Avery Andrews 940803.1149]
(Bill Powers (940802.0630 MDT))
As I see it, the general thrust of the posting is that the right way
to control `elbow stickout' in nu14 would be to vary the reference
level for the elbow-roll/shoulder-roll control system, rather than
just clobber it. Haven't done it yet, but hopefully I've managed
to set the system up so that it will be easy to try. This should make
the whole thing look more like Rick's spreadsheet model, which would
be good.
At the second level of control there would then be 3 df's of location
for the wrist (azimuth, elevation and distance, in terms of
shoulder-centered polar coordinates), three of orientation for
the hand (roll, pitch and yaw), plus the roll ratio, giving seven
higher level df's connected to seven joint angles, which is basically
what we want.
Elbow-stickout would be controlled at a third level, but to make things
look interesting we need some more third-level systems. Whether or not
hand pitch and yaw is being controlled w.r.t. the wrist or w.r.t.
the shoulder or lab space would be an obvious candidate, but I'm
not quite sure how the details would go.
Another possible area to play with is movement of the shoulder itself:
people don't normally have to do things with their shoulders pinned
into some position, so I think a realistic model of ordinary arm
activities needs to include shoulder movements, at least three and maybe
five dfs (we can assume that the person is a Tai Chi student, eager to keep
their spine vertical). This raises the question of at what level
cartesian lab space gets introduced, and also whether there are
any other relevant coordinate systems we'd want to consider, like
cylindrical ones around the trunk).
As far as path-generation (note the careful avoidance of the word
`planning': paths can be generated as executed as well as perhaps
planned before execution), there actually is an empirical issue as
to whether the path is specified as a time-varying positions
(`position contour') on the one hand or a time-varying velocity on
the other. If for example one applies a disturbance to push the
hand ahead of where it's supposed to be on the path, it ought to
bounce back to where it's supposed to be on the first story, but
not on the second. Bizzi did some experiments with fast, highly
practiced movements which seemed to support the positional view,
but of course that doesn't mean it's true for all movements.
As for changing lo-level references slowly: in nu14 I imagine that there
is an implicit first level of control systems in charge of the joint
angles, so that there is a choice between summing up the error-signals
from the second order systems and then slowing that (what nu14 actually)
does, or first slowing and then summing. Any thoughts?
Avery.Andrews@anu.edu.au