Chapter 2. Prehension

Chapter 5 - Movement Before Contact 187
5.5 Schemas for Setting up an Opposition Space
Modelling the reaching and grasping schemas in the CCP can be
done by focusing on what is the problem each schema is trying to
solve and what type of data are perceptual schemas providing. An im-
portant point identified in the previous section is that preshaping the
hand seems to be fundamentally different than enclosing the hand
around the object. A secondary point, that necessarily follows from
this, is that the arm slows down in order to effect this enclosing pro-
cess. In this section, various models are explored that might account
for this phenomenon. In addition, an analysis of preshaping vs en-
closing is presented, with terminology from the robotics community
used to introduce the enclosing process as a guarded motion.
5.5.1 Getting in the ballpark: Movefast and Moveslow
schemas
In all the studies of grasping kinematics, one consistent finding is
that the movement has two distinct phases, a fast phase getting the
hand near the object, and a slower, final adjustment phase. If there
are indeed two phases of movement (some argue more: e.g. Crossman
& Goodeve, 1983; Milner & Ijaz, 1990; some argue less: e.g., Hoff
& Arbib, in press), what might be the reason? It is not necessarily for
visual error-correcting, since Jeannerod (1984) and many other
researchers showed that the second phase occurs with and without
visual feedback. Arbib, Iberall, & Lyons (1985) suggested that the
initial phase of reaching could be the result of a plan to first put the
arm into the ‘ballpark’ of the final goal and then the second phase to
ensure contact. Suppose, in a slightly different context, one must
place a mug on a table (see Figure 5.29a). The mug is brought
towards a target location A on the table during a fast movement, with
the goal of the movement actually slightly above A; that is, A+&. As
the mug is lowered onto the table during a slower movement, the goal
is actually inside the table; that is, A-&. The environment stops this
from occurring, of course, thus stopping the mug from actually going
through the table. This second movement must be relatively slower
than the first, otherwise, something (maybe fingers!) is in danger of
breaking. For control, a generalization using two interacting schemas
is seen in Figure 5.29b. The first schema generates a movement to a
location nearby the target, the second generates a movement through
the target. Arbib and colleagues note the advantage of assuming a
small buffer factor fd for movements involving contact with objects.