Autonomy: R
527
Part B | 24
24. Autonomy: Risk Assessment
Mario P. Brito, Gwyn Griffiths
Oceanography and ocean observation in general are ever trending toward both automated
in situ observation and working in extreme environments. These goals can only be met by
de-risking the technology and deployment practices to acceptable levels of risks. A number of
industries have standardized risk management
processes to support the design and development
of their systems. The lack of formal risk assessment of autonomous ocean vehicles has hindered
the potential for true autonomy, which is required for exploring unstructured and unexplored
environments. When discussing risks, different
stakeholders may have different consequences
foremost in mind. For example, the vehicle owner
may be interested in risk of loss, whereas the user
is interested in risk of vehicle unavailability. Other
risks, such as legal risks and risk of collision, affect
all stakeholders. This chapter presents a risk management process using several methods tailored
24.1 Risk Management Process
for Autonomous Ocean Vehicles ............ 528
24.2 Risk of Failure ..................................... 529
24.2.1 Reliability Estimation................. 529
24.2.2 Reliability Modelling.................. 530
24.3 Risk of Collision ................................... 532
24.3.1 Risks of Collision
on or Near the Surface ............... 532
24.3.2 Risks of Collision Underwater ..... 533
24.4 Risk of Unavailability ........................... 534
24.5 Risk of Loss ......................................... 535
24.5.1 Expert Judgment Elicitation
Processes .................................. 536
24.5.2 Survival Prediction .................... 540
24.6 Legal Risks .......................................... 541
References................................................... 542
to autonomous ocean vehicles in which risk assessment is a key component.
Our human curiosity to understand unexplored and
hostile environments has led us to develop state of
the art automated technology capable of meeting our
measurement needs [24.1–3]. Despite having identified
the requirements for greater autonomous ocean vehicles (AOV) intelligence and autonomy, to date, physics
prohibits the deployment of the truly intelligent autonomous underwater vehicle (AUV) [24.4]. Over the
years, for practical implementations, the community
has addressed risk by making the vehicles as simple as
possible, the missions as modest as possible, and the
level of supervision as high as possible [24.4–6].
Nevertheless, over the years there have been a number of vehicle losses. Among the most high-profile
losses are the loss of Autosub2 under the Fimbulisen
ice-shelf, Antarctica on 16 February 2005, during mission 383 [24.7] and more recently the loss of the
Autonomous Benthic Explorer (ABE), during dive 222,
off the coast of Chile, on 5 March 2010. Both losses are
thought to have been caused by technical failure. In the
case of Autosub2, a formal independent inquiry concluded that an abort command or a loss of power were
equally likely to have caused the vehicle loss. There was
no formal independent inquiry into the loss of ABE;
following the accident, the design and operation team
concluded that ABE suffered a catastrophic implosion
of a glass sphere used for providing buoyancy causing
instant destruction of on-board systems [24.8]. Anecdotal evidence exists of losses of smaller vehicles such
as of Remus 100s and undersea gliders but these have
never been formally reported.
Prior to the work presented in this chapter, the risk
of AOV loss may have been estimated by the principal engineer during design and deployment. From
anecdotal evidence this was often pursued on an informal basis. Such an unstructured approach is rarely, if
ever, found entirely satisfying to all parties. Neither is
it likely to be immune to criticism, from one side or
another. Estimating the likelihood of loss depends on
a number of factors such as the vehicle’s intrinsic reli-
527
Part B | 24
24. Autonomy: Risk Assessment
Mario P. Brito, Gwyn Griffiths
Oceanography and ocean observation in general are ever trending toward both automated
in situ observation and working in extreme environments. These goals can only be met by
de-risking the technology and deployment practices to acceptable levels of risks. A number of
industries have standardized risk management
processes to support the design and development
of their systems. The lack of formal risk assessment of autonomous ocean vehicles has hindered
the potential for true autonomy, which is required for exploring unstructured and unexplored
environments. When discussing risks, different
stakeholders may have different consequences
foremost in mind. For example, the vehicle owner
may be interested in risk of loss, whereas the user
is interested in risk of vehicle unavailability. Other
risks, such as legal risks and risk of collision, affect
all stakeholders. This chapter presents a risk management process using several methods tailored
24.1 Risk Management Process
for Autonomous Ocean Vehicles ............ 528
24.2 Risk of Failure ..................................... 529
24.2.1 Reliability Estimation................. 529
24.2.2 Reliability Modelling.................. 530
24.3 Risk of Collision ................................... 532
24.3.1 Risks of Collision
on or Near the Surface ............... 532
24.3.2 Risks of Collision Underwater ..... 533
24.4 Risk of Unavailability ........................... 534
24.5 Risk of Loss ......................................... 535
24.5.1 Expert Judgment Elicitation
Processes .................................. 536
24.5.2 Survival Prediction .................... 540
24.6 Legal Risks .......................................... 541
References................................................... 542
to autonomous ocean vehicles in which risk assessment is a key component.
Our human curiosity to understand unexplored and
hostile environments has led us to develop state of
the art automated technology capable of meeting our
measurement needs [24.1–3]. Despite having identified
the requirements for greater autonomous ocean vehicles (AOV) intelligence and autonomy, to date, physics
prohibits the deployment of the truly intelligent autonomous underwater vehicle (AUV) [24.4]. Over the
years, for practical implementations, the community
has addressed risk by making the vehicles as simple as
possible, the missions as modest as possible, and the
level of supervision as high as possible [24.4–6].
Nevertheless, over the years there have been a number of vehicle losses. Among the most high-profile
losses are the loss of Autosub2 under the Fimbulisen
ice-shelf, Antarctica on 16 February 2005, during mission 383 [24.7] and more recently the loss of the
Autonomous Benthic Explorer (ABE), during dive 222,
off the coast of Chile, on 5 March 2010. Both losses are
thought to have been caused by technical failure. In the
case of Autosub2, a formal independent inquiry concluded that an abort command or a loss of power were
equally likely to have caused the vehicle loss. There was
no formal independent inquiry into the loss of ABE;
following the accident, the design and operation team
concluded that ABE suffered a catastrophic implosion
of a glass sphere used for providing buoyancy causing
instant destruction of on-board systems [24.8]. Anecdotal evidence exists of losses of smaller vehicles such
as of Remus 100s and undersea gliders but these have
never been formally reported.
Prior to the work presented in this chapter, the risk
of AOV loss may have been estimated by the principal engineer during design and deployment. From
anecdotal evidence this was often pursued on an informal basis. Such an unstructured approach is rarely, if
ever, found entirely satisfying to all parties. Neither is
it likely to be immune to criticism, from one side or
another. Estimating the likelihood of loss depends on
a number of factors such as the vehicle’s intrinsic reli-