Part B | 24.3
532 Part B Autonomous Ocean Vehicles, Subsystems and Control
and
G 4 D 1 .1 P5/ .1 P6/ .1 P7/ ;
(24.2)
G 5 D 1 .1 P8/ .1 P9/ :
(24.3)
The probability of failure for each basic event can be
estimated using statistical modeling discussed in the
previous subsection in which a probability function
is fitted to observed data. Alternatively, if operational
data is nonexistent, the probability of failure for the
base events can be estimated using expert subjective
judgment. In Sect. 24.5, we provide more details on
the formal methods that can be used to elicit expert
judgment.
24.3 Risk of Collision
The risks of collision for AUVs are very real. However,
as unregulated vehicles, no figures for the number of
losses due to collisions can be given; there is no requirement to record losses for these unmanned vehicles.
Through anecdote, the incidence of collision for underwater or surface vehicles is sufficiently high that any
group operating several vehicles is likely to encounter
this risk regularly. Sensitivity of many operators to the
details of collisions also means that quantitative, or attributable, information is rare. The examples given here
are based on our knowledge of real incidents.
24.3.1 Risks of Collision
on or Near the Surface
For surface vehicles, and when underwater vehicles operate on or near the surface, the main collision risks are:
The shore, or water shallower than the minimum
operating depth of the vehicle: At least one autonomous vehicle has run ashore because parts of
the chart from which waypoints were taken were
surveyed in the 19th century, and the position error for the shoreline was substantial. Operating near
a coast with very gentle beach gradients without full
knowledge of, and accounting for, the tides can give
rise to grounding.
Manned surface craft: Collision with manned surface craft has probably been responsible for many
autonomous vehicle losses. In some cases, the
manned craft concerned has been the craft deploying the autonomous vehicle. Mistakes and miscommunication during launch and recovery, especially,
can lead to collision with the support vessel. Collision with the sides of the vessel may inflict little
damage, but collision with propellers has on more
than one occasion led to damage to the pressure vessels of underwater vehicles and immediate loss.
Where statistics of vessel traffic are available, probabilistic models may be used to assist in estimating the risk of collision. Merckelbach has developed such a model, specifically for undersea gliders [24.21]. In principle, his approach is applicable
to other autonomous vehicles. The model considers
the vessel and vehicle speeds, the vessel traffic density, the statistics of vessel length, draft and breadth,
the water depth, the operating parameters of the vehicle, for example, time on surface and within the
draft of vessels, and a bow-wave factor. This last
factor accounts for the pushing aside that can happen as collision is imminent; in effect the effective
breadth of the ship is reduced by this factor.
Mission planning to take account of surface vessels
can make use of the automatic identification system required on internationally voyaging ships of
300 tonnes or more, all passenger ships, and, from
2014, the entire European Union (EU) fishing fleet
of vessels over 15 m in length. Websites with realtime maps enable planning ahead, and also near-real
time collision avoidance.
For very intensive shipping areas, for example,
Bosporus Strait with typically six large ships per
hour, autonomous vehicles should avoid the surface.
Large flotsam and jetsam, such as logs and baulks
of timber, lost freight containers: These hazards are
less amenable to quantifiable modeling than ship
traffic. Local knowledge is likely to be important,
for example, on the known areas where felled trees
from boreal forests may be prevalent.
Floating nets and fisheries-related hazards: Despite being banned by international agreement since
1992 in international waters, drift nets with surface floats that are meant to catch pelagic fish are
still in use within the exclusive economic zones of
many countries, for example, the USA. Other nations stipulate maximum lengths (2:5 km for the
EU). These nets are a hazard for AOVs on the
surface. Local knowledge may suggest where such
fishing happens. Avoidance using telepresence from
the vehicle to the command center of video or
radar imagery may assist in spotting the surface
floats.
Offshore structures, for example, oil and gas installations, wind farms: The positions of these surface
532 Part B Autonomous Ocean Vehicles, Subsystems and Control
and
G 4 D 1 .1 P5/ .1 P6/ .1 P7/ ;
(24.2)
G 5 D 1 .1 P8/ .1 P9/ :
(24.3)
The probability of failure for each basic event can be
estimated using statistical modeling discussed in the
previous subsection in which a probability function
is fitted to observed data. Alternatively, if operational
data is nonexistent, the probability of failure for the
base events can be estimated using expert subjective
judgment. In Sect. 24.5, we provide more details on
the formal methods that can be used to elicit expert
judgment.
24.3 Risk of Collision
The risks of collision for AUVs are very real. However,
as unregulated vehicles, no figures for the number of
losses due to collisions can be given; there is no requirement to record losses for these unmanned vehicles.
Through anecdote, the incidence of collision for underwater or surface vehicles is sufficiently high that any
group operating several vehicles is likely to encounter
this risk regularly. Sensitivity of many operators to the
details of collisions also means that quantitative, or attributable, information is rare. The examples given here
are based on our knowledge of real incidents.
24.3.1 Risks of Collision
on or Near the Surface
For surface vehicles, and when underwater vehicles operate on or near the surface, the main collision risks are:
The shore, or water shallower than the minimum
operating depth of the vehicle: At least one autonomous vehicle has run ashore because parts of
the chart from which waypoints were taken were
surveyed in the 19th century, and the position error for the shoreline was substantial. Operating near
a coast with very gentle beach gradients without full
knowledge of, and accounting for, the tides can give
rise to grounding.
Manned surface craft: Collision with manned surface craft has probably been responsible for many
autonomous vehicle losses. In some cases, the
manned craft concerned has been the craft deploying the autonomous vehicle. Mistakes and miscommunication during launch and recovery, especially,
can lead to collision with the support vessel. Collision with the sides of the vessel may inflict little
damage, but collision with propellers has on more
than one occasion led to damage to the pressure vessels of underwater vehicles and immediate loss.
Where statistics of vessel traffic are available, probabilistic models may be used to assist in estimating the risk of collision. Merckelbach has developed such a model, specifically for undersea gliders [24.21]. In principle, his approach is applicable
to other autonomous vehicles. The model considers
the vessel and vehicle speeds, the vessel traffic density, the statistics of vessel length, draft and breadth,
the water depth, the operating parameters of the vehicle, for example, time on surface and within the
draft of vessels, and a bow-wave factor. This last
factor accounts for the pushing aside that can happen as collision is imminent; in effect the effective
breadth of the ship is reduced by this factor.
Mission planning to take account of surface vessels
can make use of the automatic identification system required on internationally voyaging ships of
300 tonnes or more, all passenger ships, and, from
2014, the entire European Union (EU) fishing fleet
of vessels over 15 m in length. Websites with realtime maps enable planning ahead, and also near-real
time collision avoidance.
For very intensive shipping areas, for example,
Bosporus Strait with typically six large ships per
hour, autonomous vehicles should avoid the surface.
Large flotsam and jetsam, such as logs and baulks
of timber, lost freight containers: These hazards are
less amenable to quantifiable modeling than ship
traffic. Local knowledge is likely to be important,
for example, on the known areas where felled trees
from boreal forests may be prevalent.
Floating nets and fisheries-related hazards: Despite being banned by international agreement since
1992 in international waters, drift nets with surface floats that are meant to catch pelagic fish are
still in use within the exclusive economic zones of
many countries, for example, the USA. Other nations stipulate maximum lengths (2:5 km for the
EU). These nets are a hazard for AOVs on the
surface. Local knowledge may suggest where such
fishing happens. Avoidance using telepresence from
the vehicle to the command center of video or
radar imagery may assist in spotting the surface
floats.
Offshore structures, for example, oil and gas installations, wind farms: The positions of these surface