The control of adequate contact forces between the skin and an object
(grasp stability) is examined for two classes of prehensile actions th
at employ a precision grip: lifting objects that are ''passive'' (subj
ect only to inertial forces and gravity) and preventing ''active'' obj
ects from moving. For manipulating either passive or active objects th
e relevant fingertip forces are determined by at least two control pro
cesses. ''Anticipatory parameter control'' is a feedforward controller
that specifies the values for motor command parameters on the basis o
f predictions of critical characteristics, such as abject weight and s
kin-object friction, and initial condition information. Through vision
, for instance, common objects can be identified in terms of the finge
rtip forces necessary for a successful lift according to previous expe
riences. After contact with the object, sensory information representi
ng discrete mechanical events at the fingertips can (i) automatically
modify the motor commands, (ii) update sensorimotor memories supportin
g the anticipatory parameter control policy, (iii) inform the central
nervous system about completion of the goal for each action phase, and
(iv) trigger commands for the task's sequential phases. Hence, the ce
ntral nervous system monitors specific, more or less expected peripher
al sensory events to produce control signals that are appropriate for
the task at its current phase. The control is based on neural modellin
g of the entire dynamics of the control process that predicts the appr
opriate output for several steps ahead. This ''discrete-event, sensor-
driven control'' is distinguished from feedback or other continuous re
gulation. Using these two control processes, slips are avoided at each
digit by independent control mechanisms that specify commands and pro
cess sensory information on a local, digit-specific basis. This scheme
obviates explicit coordination of the digits and is employed when ind
ependent nervous systems lift objects. The force coordination across d
igits is an emergent property of the local control mechanisms operatin
g over the same time span.