Part B | 17.4
412 Part B Autonomous Ocean Vehicles, Subsystems and Control
basic concept is to keep the joint angles of both master and slave arms the same so that the operator feels
the contact force from any changes in (master/slave)
arm positions. To achieve a 1 W 1 direct mapping of both
arms’ angular positions, a motor-driven active master
arm is needed for duplicating the slave arm’s position.
Kraft Telerobotics [17.6] has commercialized a bilateral
force control-based deep-sea underwater manipulator
for ROVs such as WHOI’s Hybrid ROV Nereus.
Back-EMF feedback is another way of (measuring/assessing/mapping) the force on the underwater
manipulator [17.34]. It uses back-EMF voltage of electric motor in the manipulator to calculate reaction force.
Since the back-EMF voltage is proportional to the motor speed and has relatively low noise [17.38], it can
be used to measure the motor speed and position without a direct sensor. An obstruction of the robotic arm
or an object inside the gripper could be detected by calculating a feedback force on each motor from the command voltage and back-EMF voltage. This method can
be implemented only on electric motor-driven manipulators. Kerdels et al. [17.34] showed the performance
of the back-EMF-based force control of the gripper by
grabbing a light bulb with a 3 DOF Sub-Atlantic 123
DC-manipulator (rated for a depth of 300 m).
An FT sensor-based force control scheme for autonomous underwater manipulation was developed by
Marani et al. [17.35]. Actual contact forces and torques
on the manipulator are directly measured by the underwater FT sensor and used to control the manipulator.
However, as mentioned above, accuracy and precision
of most FT sensors in the market are heavily influenced
by the operating pressure. Developing FT sensors for
deepwater applications would be useful.
Visual feedback control schemes use a noncontact
sensor like a low-light charge-coupled device (CCD)
camera. However, the performance of visual feedback
control depends on environmental conditions such as
light and visibility. One good example of using a vision
sensor for robot manipulator control is visual servoing. It has been used to control a robot manipulator
with visual information, mostly with relative pose (position and orientation) between a camera and an object
in a hostile environment [17.32, 33, 39–42]. With the
visual servoing concept, the target object can be kept
within the camera field of view (FOV), so that a manipulator maintains its relative pose with respect to a static
or moving object.
It is well known that visual servoing has two types
of control methods: position-based and image-based
ones [17.43]. The position-based visual servoing is used
to calculate the relative pose of an object with respect
to the camera, which is used to control the manipulator.
In this method, the camera calibration is very critical
to get accurate relative pose. The image-based method
uses image features such as size or orientation of the
object in the camera space to calculate control commands making the current features match desired ones
that were precalculated. While it is free from the camera calibration, it requires control commands mapping
from image features to real world control space since it
does not use the position information.
One commercial position-based visual servoing for
underwater manipulator control was recently developed by FMC Technologies Schilling Robotics for
autonomous hot stab insertion [17.44]. The hot stab
insertion task is an underwater version of traditional
peg-in-hole task with a manipulator, and is more difficult than the traditional one due to very tight (almost
zero) mechanical clearance between a hot stab plug and
a receptacle (port) and the unstable base of the manipulator. In addition, the main body of the vehicle that is
the base of the manipulator is continuously moving due
to underwater current and buoyancy unless the vehicle
is sitting on the ocean floor or fixed to an underwater structure. Due to the aforementioned difficulties, it
usually takes 30 min to 1 h for well-experienced operators to complete the task. Furthermore, failure occurs
often, causing serious damage to the plug or the port.
The visual servoing helps making the operation much
faster than the human operator, completing the task
before it fails to keep the target object within the camera FOV. The visual servoing is more efficient than
the human operator in completing the operation. To
overcome drawbacks in using an optical camera in
turbid water, a high-frequency acoustic imaging sonar
camera has been developed [17.45, 46]. Acoustic imaging is a valid alternative for a human operator to get
information about the object, but it requires further
research to make a computer automatically recognize
the object [17.47–49]. A sensor fusion scheme for autonomous underwater manipulation was developed for
the SAUVIM project, using various sensors such as
an imaging sonar, a dual-frequency identification sonar
(DIDSON), optical cameras, and an underwater FT sensor [17.50, 51]. Details are described in Sect. 17.6.
412 Part B Autonomous Ocean Vehicles, Subsystems and Control
basic concept is to keep the joint angles of both master and slave arms the same so that the operator feels
the contact force from any changes in (master/slave)
arm positions. To achieve a 1 W 1 direct mapping of both
arms’ angular positions, a motor-driven active master
arm is needed for duplicating the slave arm’s position.
Kraft Telerobotics [17.6] has commercialized a bilateral
force control-based deep-sea underwater manipulator
for ROVs such as WHOI’s Hybrid ROV Nereus.
Back-EMF feedback is another way of (measuring/assessing/mapping) the force on the underwater
manipulator [17.34]. It uses back-EMF voltage of electric motor in the manipulator to calculate reaction force.
Since the back-EMF voltage is proportional to the motor speed and has relatively low noise [17.38], it can
be used to measure the motor speed and position without a direct sensor. An obstruction of the robotic arm
or an object inside the gripper could be detected by calculating a feedback force on each motor from the command voltage and back-EMF voltage. This method can
be implemented only on electric motor-driven manipulators. Kerdels et al. [17.34] showed the performance
of the back-EMF-based force control of the gripper by
grabbing a light bulb with a 3 DOF Sub-Atlantic 123
DC-manipulator (rated for a depth of 300 m).
An FT sensor-based force control scheme for autonomous underwater manipulation was developed by
Marani et al. [17.35]. Actual contact forces and torques
on the manipulator are directly measured by the underwater FT sensor and used to control the manipulator.
However, as mentioned above, accuracy and precision
of most FT sensors in the market are heavily influenced
by the operating pressure. Developing FT sensors for
deepwater applications would be useful.
Visual feedback control schemes use a noncontact
sensor like a low-light charge-coupled device (CCD)
camera. However, the performance of visual feedback
control depends on environmental conditions such as
light and visibility. One good example of using a vision
sensor for robot manipulator control is visual servoing. It has been used to control a robot manipulator
with visual information, mostly with relative pose (position and orientation) between a camera and an object
in a hostile environment [17.32, 33, 39–42]. With the
visual servoing concept, the target object can be kept
within the camera field of view (FOV), so that a manipulator maintains its relative pose with respect to a static
or moving object.
It is well known that visual servoing has two types
of control methods: position-based and image-based
ones [17.43]. The position-based visual servoing is used
to calculate the relative pose of an object with respect
to the camera, which is used to control the manipulator.
In this method, the camera calibration is very critical
to get accurate relative pose. The image-based method
uses image features such as size or orientation of the
object in the camera space to calculate control commands making the current features match desired ones
that were precalculated. While it is free from the camera calibration, it requires control commands mapping
from image features to real world control space since it
does not use the position information.
One commercial position-based visual servoing for
underwater manipulator control was recently developed by FMC Technologies Schilling Robotics for
autonomous hot stab insertion [17.44]. The hot stab
insertion task is an underwater version of traditional
peg-in-hole task with a manipulator, and is more difficult than the traditional one due to very tight (almost
zero) mechanical clearance between a hot stab plug and
a receptacle (port) and the unstable base of the manipulator. In addition, the main body of the vehicle that is
the base of the manipulator is continuously moving due
to underwater current and buoyancy unless the vehicle
is sitting on the ocean floor or fixed to an underwater structure. Due to the aforementioned difficulties, it
usually takes 30 min to 1 h for well-experienced operators to complete the task. Furthermore, failure occurs
often, causing serious damage to the plug or the port.
The visual servoing helps making the operation much
faster than the human operator, completing the task
before it fails to keep the target object within the camera FOV. The visual servoing is more efficient than
the human operator in completing the operation. To
overcome drawbacks in using an optical camera in
turbid water, a high-frequency acoustic imaging sonar
camera has been developed [17.45, 46]. Acoustic imaging is a valid alternative for a human operator to get
information about the object, but it requires further
research to make a computer automatically recognize
the object [17.47–49]. A sensor fusion scheme for autonomous underwater manipulation was developed for
the SAUVIM project, using various sensors such as
an imaging sonar, a dual-frequency identification sonar
(DIDSON), optical cameras, and an underwater FT sensor [17.50, 51]. Details are described in Sect. 17.6.