14 degree arm, stage II

[Avery Andrews 931012.1753]

Although I don't consider the low level control issues for the 14 dof
arm to be all sorted out, I've started to think about how to get it
to do something, such as pick up a coffee cup. Overall, coffee-drinking
seems to me to involve 4 phases:

  I. Pick up cup

  II. Move cup to lips

  III. Sip from cup

  IV. Put cup back.

Thinking about I, I divide it into two phases:

  A. Pre-Positioning
        move arm from where it is to a good position from which to grip
         the cup, being careful not to knock anything over.

  B. Gripping
        move hand to where grasp can close on cup, and close it.

For gripping, I think I have a preferred yaw angle for the forearm to
be, about 45 degrees (projection of forearm into the horizontal plane),
which is constant as the cup moves around. There is also a preferred
`wrist roll angle', again about 45 degrees (I grip cup with thumb and
index finger, the rest of the hand sloping away from the cup).

Note that if there is a reference level for the forearm yaw angle and
for the position of the hand, the four dof's of the arm itself (as
opposed to the wrist and fingers) are used up.

At stage A, the main concern is not to knock anything over. Grownups
have a sort of `alerting system' whereby, when the arm has a significant
motion component in the horizontal plane, care is taken that things not
be knocked over. This works less well in children. Therefore I
conjecture that it is actually a hard problem. I don't have any very
clear ideas about how to model it (it maybe be similar to getting
a more intelligent model of obstacle avoidance built into the CROWD
people, for example, one that would permit them to go straight6t to a gap
in a wall, rather than going all the way to the wall and then moving
along it until they find a gap.)