5.5 Control structures
will actually maintain its reference position and also that of the work-piece
8
. However, the
trajectory of the work-piece will still be lost, but could be easily referenced to that of the
master manipulator. The latter scheme can be termed a dynamically coupled master-slave
approach analogous to the terminology mentioned in section 2.4.1. This is illustrated in
Figure 5.9. Where the second robot is position controlled and hence represents the master
manipulator while the first is the slave manipulator since the compliance
/impedance
controller is active. The interaction control loop of the second robot is grayed out to
indicate that it is available if needed. According to preference, the roles could simply be
switched to reference the work-piece trajectory to other robots.
Robot 1
commercial robot controller
Robot 1
Trajectory
Generator
Robot 2
Robot 2 Trajectory
Generator
commercial robot controller
FTS 1
WP Trajectory
Generator
WP
Trajectory
Comp/Imp
Controller 1
Gravity
Compensation 1
FTS 2
Comp/Imp
Controller 2
Gravity
Compensation 2
Environment
WP Gravity
Compensation
Geometrical
Constraints
Figure 5.9: Control structure facilitating adaptive control for manipulator
/
work-piece
interaction
5.5.3.2 Work-piece
/environment interaction
This structure addresses the case when interaction between the work-piece and the envi-
ronment takes place. Compared to the latter structure, this is fully centralized in nature.
The forces on the respective TCP are fed into a work-piece force observer to determine
the total forces on the work-piece. Consequently these forces could be controlled and
a correction trajectory for the work-piece is generated, which subsequently augments
the original reference trajectory specified by the operator as shown in Figure 5.10. It
is apparent that the outcome here is a compliant work-piece motion, whether it be a
compliance or impedance controller. This structure is of utmost importance in operations
where the 6
th
phase (refer to section 4.2) comprises the main objective of the operation
8
Based on the assumption that the contact between them is rigid as mentioned in section 5.2.3.2
75
5 Control Architecture
i.e. an assembly task. Not only is it possible to control the forces but by dynamically
changing the controller parameters it is also possible to execute autonomous tasks with
the work-piece by controlling the shape of the interaction in a variable fashion.
Robot 1
commercial robot controller
Robot 1 Trajectory
Generator
Robot 2
Robot 2 Trajectory
Generator
commercial robot controller
FTS 1
WP Trajectory
Generator
WP
Impedance
Controller
FTS 2
Environment
WP Force
Observer
WP Gravity
Compensation
WP
Trajectory
Geometrical
Constraints
Figure 5.10: Control structure facilitating adaptive control for work-piece
/
environment
interaction
5.6 Architecture
In the latter section the di
fferent control structures pertaining to all functionalities required
from the control module were discussed. These structures are based on the interaction
control principles presented in section 5.3. Along with the assistance functions, the latter
components establish the core of the control architecture as a practical implementation
of the control module in the framework. The architecture in its entirety is illustrated in
Figure 5.11. Referring to the definitions and terminology in section 5.1.1, the architecture
is based on three main layers.
The first layer, termed the low-level control layer, consists of what requires real-time
execution and hence lies on the higher end of the bandwidth scale. It represents the
integration of the components from the aforementioned control structures in a manner
which facilitates an exclusive triggering of any structure at any given time without in-
terfering with the others. A rearrangement of the blocks reveals three major loops: two
decentralized inner loops representing the local control around each robot, a
centralized
outer loop representing the global control on the work-piece. Since the work-piece in
itself is a non-drivable object, its movement is based on the robots’ motion and hence the
centralized loop is based on the decentralized loops. Each loop has its own fully parame-
terized interaction control law (ICL) making it capable of enforcing the triggered structure.
The second layer, termed the high-level control layer, contains two main components
executed in a way similar to programmable logic controllers (PLC). The operation modes
trigger either the robot mode which mainly invokes the decentralized control loops or the
work-piece mode which triggers the centralized control loop.
76