|
k
F
REF
F
SENS
k
f
F
k
x
X
X
F
Ȧ
k
mx
dx
cx
+
+
..
.
Figure 2.5: Classification of interaction control schemes in the robot task space according
to Surdilovic & Vukobratovi´c (2002)
2.3.2.1 Passive control
Passive control schemes allow the robot to react to forces due to inherent compliance
whether structural or controller based (Khalil & Dombre 2002, P. 378-380). Duly the
19
2 Literature Review
sti
ffness of the robot is intentionally lowered by a multitude of methods to achieve this.
One of the more popular methods widely deployed in industrial practice is the remote
center compliance (RCC) device (Whitney 2004, P. 260). Another method is building
su
fficient compliance in the robot’s mechanical structure i.e. increasing the joint and arm
elasticity (Pfeiffer 1992)(Ang Jr & Andeen 1995) to accommodate forces on the TCP.
The latter are termed ’fixed methods’ due to their constant behavior, since they depend on
components that are not interchangeable or adaptable to the needs of the task. If however,
RCC devices were adjustable (Chun 1992) or structural compliance was tunable through
motor servo-gains they, would then fall under the adaptive category.
2.3.2.2 Active control
Active control di
ffers from passive control in that it requires the measurement and feedback
of forces on the robot’s TCP to achieve a specific robot behavior during interaction. In the
force control scheme, the forces and position are simultaneously controlled to track the
nominal positions and forces (Eppinger & Seering 1987)(Bigras et al. 2007a). While in
the hybrid control scheme both forces and positions are controlled in orthogonal subspaces.
Therefore the directions in which the robot interacts with the environment (constrained)
are force controlled while the directions in which the robot is allowed to move freely
(unconstrained) are position controlled (Raibert & Craig 1981)(Craig 2005, P. 373-378).
On the other hand impedance control advocates controlling the relationship between the
robot and the environment thereby acquiring a target impedance (Hogan 1985)(Bigras
et al
. 2007b). Implementing latter control schemes is usually executed in either the joint
space or in the task space. This depends on two major factors; the first is the availability
of a near-realistic dynamic model of the robot and the second is the interfaces available to
control the robot i.e. whether access to motor control is permitted (Zhang et al. 2004).
The latter factor also plays a role in re-structuring the control schemes around the position
controller of the robot where the controller’s output is defined in terms of a correctional
position signal (Pelletier & Doyon 1994)(Lange & Hirzinger 2005)(Winkler & Suchy
2007b).
2.3.3 Remote center compliance
Given the wide spread application of these types of devices in manufacturing processes
(involving interaction), they will later be separately discussed. RCC devices were first
developed in The Charles Stark Draper Laboratory (Drake & Simunovic 1979). Initial
investigations resulted in mathematical models to facilitate the fitting of parts with non-
compliant mechanisms i.e. robots (Holzbock 1996, P. 91). Based on these models, the
first devices for mechanical compliance were developed and patented (Drake & Simunovic
1979)(De Fazio & Whitney 1984)(Whitney 1985). Although their development continues
to present day, the principles upon which the devices are built are unaltered (Holzbock
1996, P. 92). As seen in Figure 2.6 an RCC consists of two parts; a translational part
allowing movement in the lateral direction and a rotational part allowing movement in
20
2.3 Interaction control
the angular direction. The ’remote center’ represents the optimum location at which
the insertion process may take place. In order to avoid jamming or wedging, specific
conditions regarding clearances and friction coe
fficients have to be met. Additionally,
at least one of the parts in the assembly process has to possess chamfered edges. The
mechanical embodiment of this principle consists of two plates connected by several
columns of elastic pads interspersed with metal plates (Hoshizaki & Bopp 1990, P. 238).
This geometrical formation o
ffers much higher stiffness in compression than in shear,
subsequently causing the lower plate to move in a lateral and
/or torsional direction w.r.t.
the upper plate (Whitney 2004, P. 260). By changing the type of elastic material and the
geometrical stacking shape, the compliance of the device could be altered to correspond
to specific application requirements. Furthermore, commercial devices are built in a
manner allowing them to be easily mounted between the manipulator and the gripper using
standard adapters. In practice, RCC devices are selected and deployed according to the
following criteria (Hoshizaki & Bopp 1990, P. 238-240):
1. Nominal values of misalignment correction could be achieved at the remote center
point. Any change in the distance between the mounting face of the RCC and the
remote center point adversely a
ffects the insertion process.
2. The repeatability of the RCC must be considered in the context of the whole robotic
system. Hence it is imperative to select the RCC conforming to the tolerances
expected from a system.
3. Given the inherent compliance of an RCC, oscillation of the work-piece during
movement of the robot is inevitable. Hence the natural frequency of the device
is selected so as to reduce the damping time of the oscillation based on the given
weight class.
Although the vast majority of RCC devices exhibit constant behavior, tunable devices have
also been developed (Chun 1992). Additionally, researchers investigated the classical RCC
principle to achieve more accurate mathematical models (Ciblak & Lipkin 1996)(Ciblak
& Lipkin 2003).
2.3.4 Impedance control
On the contrary to direct force control, impedance represents a general framework for
controlling the dynamic interaction of robots with environments (Vukobratovi´c et al. 2003,
P. 4) which was introduced by Hogan in his seminal work in 1985 (Hogan 1985)(Hogan
1989). The naming convention owes itself to the electrical engineering discipline from
which the original idea stemmed. Instead of defining and controlling a certain force at the
interaction’s interface i.e. TCP, it defines the shape of the dynamic interaction. Hence,
the controlled variable is neither the positions nor the forces but rather the relationship of
interaction (Nagchaudhuri & Garg 2001). Such control techniques are beneficial when
the magnitude and direction of the forces are not vital to accomplishing the task at hand.
21
2 Literature Review
a. No loading
Translational
Do'stlaringiz bilan baham: |
