Autonomous Underwater Vehicle Docking 16.7 Conclusions and Future Prospects 403
Part B | 16.7
servatory are likely to have a very high bandwidth
connection with the shore. This provides the potential
for operator monitoring and control of dock, and to the
extent that the vehicle is in communication, monitoring
and control of the vehicle as well.
16.6.4 Detecting and Recovering
from Failures
Failures are inevitable with autonomous systems, particularly when deployed for extended periods in a hostile and unpredictable environment. The robustness of
a system is thus a function of both its inherent ability to avoid failures and its ability to recover from
failures when they do occur. Typically the first is
achieved through overdesign, which is expensive, and
the second remains a research topic. For most complex engineered systems in use today, robustness is
achieved largely through experience; by having processes in place to learn from failures and to implement
the lessons learned. The few docking systems that have
been tested in the ocean, have been in use only briefly.
Consequently, it seems likely that only the most common failure modes have been experienced.
The large number of discrete control functions involved with docking, and the sensitivity of many of
these activities to environmental conditions or disturbances, makes failures difficult to predict. Consequently, the overall docking process must be robust to
failure of individual steps. For example, a missed approach, where the vehicle attempts to home on the dock
but misses, is a particular class of failure that is likely
to occur many times. This might be caused by currents, poor sensing conditions, turbulence, or anomalies
in vehicle performance. The appropriate response for
the AUV may be quite different, depending on the
root cause of the missed approach. This leads to both
a classification problem and a contingency management
problem. It is not enough to know that an approach
has failed, because the reason for failure is important as well. However, to the degree that contingency
responses are created for different failure modes, mission management becomes more complex. In effect,
each contingency response creates the potential for unintended consequences that must be anticipated and
mitigated. At present there is no good way to do this.
Handling of some simple failure modes is discussed
by both Singh et al. [16.5] and McEwen et al. [16.2].
The state transition diagram in Fig. 16.6 describes the
vehicle response to some simple failures. For example,
the missed dock transition leads to a state of which the
goal is repositioning the vehicle for another approach.
If the vehicle misses the approach three times, it aborts
the docking attempt and goes to the surface.
16.7 Conclusions and Future Prospects
Of the activities attempted with AUVs, docking is
among the most challenging. The literature reflects this
difficulty, with most papers testing some element of
docking, and only a few reporting results from a deployed system in the ocean. Most tests focus on homing. Those that deploy and exercise a full docking suite,
including battery recharge, data dump, and initiation of
new vehicle missions, report results for comparatively
brief deployments. Consequently, one must conclude
that docking has not made the transition from demonstration to operations. This prompts the questions: why
has docking not become an operational capability, and
what are the conditions that might allow this event to
occur?
In the early 1990s, when the AOSN paper first proposed AUV docking as an integral part of an ocean
observing system, little attention was given the infrastructure required to support and power a docking
system. However, docking does imply the existence of
an infrastructure. For example, if the objective is to support ten AUV full missions from the dock, then the
dock must be able to supply power levels equal to ten
times the battery capability of the vehicle (assuming
the vehicle missions uses the full battery capacity each
mission). Early attempts to use docking to support field
programs thus had to invest in their own infrastructure,
such as moorings with satellite communications and
large quantities of batteries [16.1]. This created a significant barrier to the adoption of AUV docking. Not
only does a team have to master the most difficult aspect
of AUV operations, it must also be capable of building
from scratch the dock and the infrastructure supporting
the dock.
The deployment of cabled observatories, which
adapt subsea telecommunications technology to provide high bandwidth, low-latency communications and
high power to nodes in the ocean interior, provide
an ideal infrastructure to support docked AUVs. Early
scientific cabled observatories include a system deployed off Japan in the late 1970s [16.61] and the
LEO-15 system, installed a few kilometers off the coast
of New Jersey in the 1990s [16.62]. Since then, a wide
number of cabled systems have been installed, supporting plug and play instruments on the deep seafloor
Part B | 16.7
servatory are likely to have a very high bandwidth
connection with the shore. This provides the potential
for operator monitoring and control of dock, and to the
extent that the vehicle is in communication, monitoring
and control of the vehicle as well.
16.6.4 Detecting and Recovering
from Failures
Failures are inevitable with autonomous systems, particularly when deployed for extended periods in a hostile and unpredictable environment. The robustness of
a system is thus a function of both its inherent ability to avoid failures and its ability to recover from
failures when they do occur. Typically the first is
achieved through overdesign, which is expensive, and
the second remains a research topic. For most complex engineered systems in use today, robustness is
achieved largely through experience; by having processes in place to learn from failures and to implement
the lessons learned. The few docking systems that have
been tested in the ocean, have been in use only briefly.
Consequently, it seems likely that only the most common failure modes have been experienced.
The large number of discrete control functions involved with docking, and the sensitivity of many of
these activities to environmental conditions or disturbances, makes failures difficult to predict. Consequently, the overall docking process must be robust to
failure of individual steps. For example, a missed approach, where the vehicle attempts to home on the dock
but misses, is a particular class of failure that is likely
to occur many times. This might be caused by currents, poor sensing conditions, turbulence, or anomalies
in vehicle performance. The appropriate response for
the AUV may be quite different, depending on the
root cause of the missed approach. This leads to both
a classification problem and a contingency management
problem. It is not enough to know that an approach
has failed, because the reason for failure is important as well. However, to the degree that contingency
responses are created for different failure modes, mission management becomes more complex. In effect,
each contingency response creates the potential for unintended consequences that must be anticipated and
mitigated. At present there is no good way to do this.
Handling of some simple failure modes is discussed
by both Singh et al. [16.5] and McEwen et al. [16.2].
The state transition diagram in Fig. 16.6 describes the
vehicle response to some simple failures. For example,
the missed dock transition leads to a state of which the
goal is repositioning the vehicle for another approach.
If the vehicle misses the approach three times, it aborts
the docking attempt and goes to the surface.
16.7 Conclusions and Future Prospects
Of the activities attempted with AUVs, docking is
among the most challenging. The literature reflects this
difficulty, with most papers testing some element of
docking, and only a few reporting results from a deployed system in the ocean. Most tests focus on homing. Those that deploy and exercise a full docking suite,
including battery recharge, data dump, and initiation of
new vehicle missions, report results for comparatively
brief deployments. Consequently, one must conclude
that docking has not made the transition from demonstration to operations. This prompts the questions: why
has docking not become an operational capability, and
what are the conditions that might allow this event to
occur?
In the early 1990s, when the AOSN paper first proposed AUV docking as an integral part of an ocean
observing system, little attention was given the infrastructure required to support and power a docking
system. However, docking does imply the existence of
an infrastructure. For example, if the objective is to support ten AUV full missions from the dock, then the
dock must be able to supply power levels equal to ten
times the battery capability of the vehicle (assuming
the vehicle missions uses the full battery capacity each
mission). Early attempts to use docking to support field
programs thus had to invest in their own infrastructure,
such as moorings with satellite communications and
large quantities of batteries [16.1]. This created a significant barrier to the adoption of AUV docking. Not
only does a team have to master the most difficult aspect
of AUV operations, it must also be capable of building
from scratch the dock and the infrastructure supporting
the dock.
The deployment of cabled observatories, which
adapt subsea telecommunications technology to provide high bandwidth, low-latency communications and
high power to nodes in the ocean interior, provide
an ideal infrastructure to support docked AUVs. Early
scientific cabled observatories include a system deployed off Japan in the late 1970s [16.61] and the
LEO-15 system, installed a few kilometers off the coast
of New Jersey in the 1990s [16.62]. Since then, a wide
number of cabled systems have been installed, supporting plug and play instruments on the deep seafloor