Part B | 17.3
410 Part B Autonomous Ocean Vehicles, Subsystems and Control
During the manipulation, the end-effector or tool
of the manipulator could be in contact with the environment, and forces and moments would occur at the
end-effector. In this situation, (17.3) must be modified
by adding the reaction torque to the end-effector force
on the right side of (17.2).
r D J
T
m .q/f e ;
(17.4)
where is J m 2 R
6n the manipulator Jacobian matrix,
f e 2 R
61 a vector of forces and torques at the endeffector.
The dynamics of the underwater vehicle manipulator is quite complicated as shown in (17.2), highly nonlinear, and involves coupled equations with unknown
parameters and disturbances. It is almost impossible to
accurately model the dynamic equations and to operate
the manipulator at a nominal speed using a conventional
controller. However, when the manipulator moves very
slowly in the friendly environment of no or minimal
current or wave, the dynamics of the manipulator could
be represented by the simplified linear version of (17.3)
since the effect of velocity-dependent terms in (17.3)
becomes negligible. In fact, during the actual operation
in the field, most commercial underwater manipulators
move very slowly (much less than 1 rad s
1 ) for safety
and other reasons, and they use conventional joint controllers for the master–slave teleoperation.
17.3 Teleoperation of Underwater Vehicle Manipulators
A typical setup for teleoperation involves the operator
sitting in the control room located on the mother ship
and holding a master arm that is a miniature of the
actual underwater vehicle manipulator (slave arm) attached to the ROV. The operator controls the slave arm
by moving the master arm whose motion is followed
by the slave arm. During the operation, the operator relies on the visual information of the work site in the
form of a series of 2-D video images that are captured
by cameras on the vehicle’s main body and the slave
arm and then transmitted to the control room monitors by an underwater cable. Not only the view of the
work site is limited but also the visual information especially for the deep-sea operation is often delayed. If the
vehicle is not holding or sitting on an underwater structure, the ROV operator must try to control the vehicle
for station-keeping during the manipulation. However,
even if there is no current or wave during the stationkeeping, the vehicle moves like free-floating within the
accuracy of the vehicle position sensors in the order of
a meter or so. Therefore, it is difficult to achieve the
accuracy at the end-effector since underwater vehicle
manipulators are attached to the vehicles that are constantly moving, unlike industrial manipulators whose
bases are fixed on the factory floor. As Hildebrandt
et al. [17.23] state, many simple tasks on the ground
such as plugging become very difficult to perform by
the underwater vehicle manipulator. Underwater plugging tasks, done on a trial-error basis since the required
precision could not be achieved by the ROV manipulator system, take hours to complete. Therefore, operator
fatigue often becomes a critical issue in teleoperating
the underwater vehicle manipulator.
Several researchers have addressed the issues mentioned above and have improved the accuracy of
the manipulator in teleoperation. Among them include Sayers et al. [17.24], LaPierre et al. [17.25],
Zhang et al. [17.26], Bowen et al. [17.27], Hildebrandt
et al. [17.23], Hildebrandt et al. [17.28], and Shim
et al. [17.29]. Zhang et al. [17.26] developed a virtual
model-based teleoperating system and experimentally
showed a good performance of moving target tracking
with up to a two second time delay. They used on-line
parameter estimation and showed that the manipulator
could be continuously controlled with a low bandwidth
network. Hildebrandt et al. [17.23] implemented precise manipulator position control by compensating for
errors due to the ROV movement. The overall scheme
was divided into:
1. Prediction of vehicle movement
2. Calculation of manipulator compensation movement
3. Compensating motion with the manipulator.
The prediction of vehicle movement relied on a position sensor in the vehicle and a mathematical model
of the vehicle. The manipulator compensation movement was calculated by the difference between the
current position and the predicted position. The compensation position in the Cartesian space was then
transformed into a joint space position to be given to
the manipulator. They showed that the ROV movement compensation algorithm considerably increased
the position accuracy of the end-effector of the commercial hydraulic underwater manipulator, Schilling Orion
7P (Fig. 17.1). Notably, Hildebrandt et al. [17.28] and
Shim et al. [17.29] have developed additional computer
control systems to the top of a commercial underwater vehicle manipulator system, Schilling’s Orion
410 Part B Autonomous Ocean Vehicles, Subsystems and Control
During the manipulation, the end-effector or tool
of the manipulator could be in contact with the environment, and forces and moments would occur at the
end-effector. In this situation, (17.3) must be modified
by adding the reaction torque to the end-effector force
on the right side of (17.2).
r D J
T
m .q/f e ;
(17.4)
where is J m 2 R
6n the manipulator Jacobian matrix,
f e 2 R
61 a vector of forces and torques at the endeffector.
The dynamics of the underwater vehicle manipulator is quite complicated as shown in (17.2), highly nonlinear, and involves coupled equations with unknown
parameters and disturbances. It is almost impossible to
accurately model the dynamic equations and to operate
the manipulator at a nominal speed using a conventional
controller. However, when the manipulator moves very
slowly in the friendly environment of no or minimal
current or wave, the dynamics of the manipulator could
be represented by the simplified linear version of (17.3)
since the effect of velocity-dependent terms in (17.3)
becomes negligible. In fact, during the actual operation
in the field, most commercial underwater manipulators
move very slowly (much less than 1 rad s
1 ) for safety
and other reasons, and they use conventional joint controllers for the master–slave teleoperation.
17.3 Teleoperation of Underwater Vehicle Manipulators
A typical setup for teleoperation involves the operator
sitting in the control room located on the mother ship
and holding a master arm that is a miniature of the
actual underwater vehicle manipulator (slave arm) attached to the ROV. The operator controls the slave arm
by moving the master arm whose motion is followed
by the slave arm. During the operation, the operator relies on the visual information of the work site in the
form of a series of 2-D video images that are captured
by cameras on the vehicle’s main body and the slave
arm and then transmitted to the control room monitors by an underwater cable. Not only the view of the
work site is limited but also the visual information especially for the deep-sea operation is often delayed. If the
vehicle is not holding or sitting on an underwater structure, the ROV operator must try to control the vehicle
for station-keeping during the manipulation. However,
even if there is no current or wave during the stationkeeping, the vehicle moves like free-floating within the
accuracy of the vehicle position sensors in the order of
a meter or so. Therefore, it is difficult to achieve the
accuracy at the end-effector since underwater vehicle
manipulators are attached to the vehicles that are constantly moving, unlike industrial manipulators whose
bases are fixed on the factory floor. As Hildebrandt
et al. [17.23] state, many simple tasks on the ground
such as plugging become very difficult to perform by
the underwater vehicle manipulator. Underwater plugging tasks, done on a trial-error basis since the required
precision could not be achieved by the ROV manipulator system, take hours to complete. Therefore, operator
fatigue often becomes a critical issue in teleoperating
the underwater vehicle manipulator.
Several researchers have addressed the issues mentioned above and have improved the accuracy of
the manipulator in teleoperation. Among them include Sayers et al. [17.24], LaPierre et al. [17.25],
Zhang et al. [17.26], Bowen et al. [17.27], Hildebrandt
et al. [17.23], Hildebrandt et al. [17.28], and Shim
et al. [17.29]. Zhang et al. [17.26] developed a virtual
model-based teleoperating system and experimentally
showed a good performance of moving target tracking
with up to a two second time delay. They used on-line
parameter estimation and showed that the manipulator
could be continuously controlled with a low bandwidth
network. Hildebrandt et al. [17.23] implemented precise manipulator position control by compensating for
errors due to the ROV movement. The overall scheme
was divided into:
1. Prediction of vehicle movement
2. Calculation of manipulator compensation movement
3. Compensating motion with the manipulator.
The prediction of vehicle movement relied on a position sensor in the vehicle and a mathematical model
of the vehicle. The manipulator compensation movement was calculated by the difference between the
current position and the predicted position. The compensation position in the Cartesian space was then
transformed into a joint space position to be given to
the manipulator. They showed that the ROV movement compensation algorithm considerably increased
the position accuracy of the end-effector of the commercial hydraulic underwater manipulator, Schilling Orion
7P (Fig. 17.1). Notably, Hildebrandt et al. [17.28] and
Shim et al. [17.29] have developed additional computer
control systems to the top of a commercial underwater vehicle manipulator system, Schilling’s Orion