CHAPTER 2. State of the art
tion on the normal component of the measured force, on the basis of some fuzzy
interferences of the measured component of the tangential force.
Another method based on the linearization of the system is proposed in [25].
The dynamics of the system is derived in joint space so as to avoid the kinematic
inversion and have a set of differential equations with algebraic constraints. Us-
ing some tools present in [50], a state-space input/output linearization is achieved
in order to decouple position and force control problems. Moreover, via a pole-
placement method, it is possible to give to the system the desired behavior. Fur-
ther, it has been shown that the controller exists in all the space, except in the
singularity points of the fingers. This approach requires the use of force sensors.
The approaches proposed in [28, 31] can be also enumerated in this class of
controllers. In those works, the problem of stable grasping and manipulation is
faced using a finger pairs covered with a soft compressible layer material. The
controller output is a linear combination of signals, each of one addressing a par-
ticular subtask, namely: internal forces, external torques balance, desired position
of the object's center of mass, etc. Moreover, a high level controller is assumed to
manage some grasp quality measures.
Finally, starting from the consideration that if the control of the hand is cen-
tralized and the object changes, the controller should change consequently: in [83],
a hierarchical controller is implemented, and it achieves a decentralization in or-
der to control each finger in an independent way through a hybrid position/force
control scheme.
2.3.4
Hierarchical control
A hierarchical control method can be, in general, included in the previous sec-
tions about hybrid position/force control and computed torque, since one of the
objectives of a hierarchical controller is to decouple force and position spaces. But
some works about hierarchical control have something more: in particular, they
take inspiration from human motion control.
In [41], it is pointed out the question of how human brain could control a
system like the human body with so many different degrees of freedom interacting
in such a complex fashion. Such a complexity is also present in the robotic hands.
Starting from these studies, in [67], a hierarchical control scheme for robot
manipulation is built, reflecting a possible hierarchical control scheme for a human
finger. The robotic hand is modeled as a set of entities called robots, which are
coupled to each other and to an object through a set of constraints. The controller
has to modify the properties of these entities with the desired attributes. Attaching
all these modules, a hierarchical control structure grows up.
Instead, the work in [49] is inspired by two classes of human motions, namely
reflexes and voluntary movements: these last suppress a reflex when needed. In
this way, a control scheme made up of three modules has been implemented: the
14