501

1 INTRODUCTION

Digitalization and automation on ship’s bridges are

quite common nowadays in the maritime industry.

Different systems from electronic charts, automation

systems, fuel performance monitoring, or integrated

bridge systems up to decision support systems can be

found on ships worldwide. In recent years, ideas and

applications for autonomous shipping have been

rapidly increasing. In most of today’s ship bridge

systems, decision support systems with different

capabilities are installed and officers of the watch rely

on them. First tests with fully and constrained

autonomous ships are on the way. One of them is the

B0 | BZERO project, with the aim of an autonomous

8-hour watch-free bridge, while the ship is still

manned. The challenge hereby is the definition and

design of the ship’s systems.

1.1 Autonomy Levels

In the transportation industry, different definitions of

autonomy levels exist: Sheridan and ALFUS, SAE

autonomy levels, metro grade of automation, National

Business Aviation Association levels, Lloyd’s Register,

Maritime21 [13, 14] or IMO (International Maritime

Organization) levels of autonomy [5], whereby the

IMO levels (see Figure 1) are the most general and

applicable for maritime applications.

Figure 1. IMO levels of autonomy

Learning to Swim - How Operational Design Parameters

etermine the Grade of Autonomy of Ships

. Ugé

& S. Hochgeschurz

Fraunhofer-Center for Maritime Logistics and Services CML, Hamburg, Germany

Fraunhofer Institute for Communication, Information Processing and Ergonomics FKIE, Wachtberg, Germany

ABSTRACT: In recent years, ideas and applications for autonomous shipping have been rapidly increasing. In

most of today’s ship bridge systems decision support systems with

different capabilities are installed and

officers of the watch rely on them. First tests with fully and constrained autonomous ships are on the way. One

of them is the B0 | BZERO project, with the aim of an autonomous 8-hour watch-free bridge, while the ship is

still manned. The system’s constraints are captured in the operational design domain (ODD) defining all

conditions under which the autonomous system can operate safely. We propose the definition of a preliminary

ODD considering both regulatory and technical restrictions. Furthermore, we present a new way of defining the

level of autonomy of a ship by using the ODD and navigational specifications.

http://www.transnav.eu

the

International Journal

on Marine Navigation

and Safety of Sea Transportatio

Volume 15

Number 3

September 2021

DOI: 10.12716/1001.15.03.02

502

The IMO levels of autonomy look at two aspects:

On the one hand the manning of the ship and on the

other hand the control mode. In level 1 (Decision

Support) the ship’s crew is still on board, decision

support systems and automated processes are

available. Shipboard operations are mainly conducted

manually, but some may be automated. Level 2

(Remote Control with Seafarers) means that the ship’s

crew is still on board, but the ship is controlled from

another location, whereas local crew is available to

intervene if unintended or risky maneuvers or

situations apply. The third level (Remote Control

without Seafarers) also includes remote control, but

here seafarers are not on board. Level 4 (Fully

Autonomous) means that no crew is available on

board and no remote center is controlling the ship.

The shipboard systems are deciding and acting on

their own. Nevertheless, and as we will show, there

are more sublevels or different combinations of those

aspects, and situations where an exact allocation to

one of these levels might not be suitable.

1.2 B ZERO Project

Looking at recent research, the development of MASS

is steadily growing [7]. However, since the era of fully

autonomous and commercially viable ships is not yet

there, a hybrid approach is developed, where

automated and autonomous systems are used, while

still the full or a reduced crew is on board. Whilst

ship’s engine rooms can be already operated without

any crew, the ship’s bridge still needs to be manned at

all times [13]. The project B0 | B ZERO looks into the

possibility of using autonomous systems to achieve an

8-hour watch-free bridge. The designed system is

intended to be built on an existing ship. Therefore,

technical constraints, available sensors and the

planned operational area are part of the project’s work

packages, requiring a thorough planning, specification

and limiting of resources, devices and data or

information.

1.3 Aim

The aim of this paper is to propose a different

approach to categorize levels of autonomy for surface

ships, after describing the outcomes of the

navigational and operational design domain (ODD) of

the B0 | B ZERO project. During the process of

creating the navigational specifications and the ODD

for the B0 | B ZERO project, it became clear that

existing autonomy levels are not sufficient to reflect

the possible variety of vessel autonomy levels. The

need for a further differentiation between more levels

of autonomy arose, because the existing IMO levels

only focus on the two parameters manning and level

of automation of devices. During the B0 | B ZERO

project, the necessity for further parametrization

occurred, in regards to existing regulations, ship sizes

and environmental as well as situational

circumstances. Therefore, this paper takes into

account a greater variety of parameters influencing

the levels of autonomy by considering the findings

from the B0 | B ZERO navigational specifications and

ODD.

2 METHODS

2.1 Defining Navigational Specifications

In order to develop the functional and non-functional

specifications of a ship’s navigation system for an 8-

hour watch-free bridge, a navigational specification

was established for the B0 | BZERO project. Those

specifications were broken down into fundamental,

environmental, human-centered, safety and non-

functional requirements.

In a first step the available data sources were

summarized, the corresponding navigational domains

addressed and gaps and missing data identified. In a

second step, the duties of the human officer of the

watch (OOW), the autonomous system (referred to as

Auto-OOW) and the master, closely derived from

human bridge duties, were determined. This means

that generally the Auto-OOW will also pass through

the four process stages of control [11]: acquisition of

information, analysis of information, decision and

action selection and action implementation.

As a third step, restrictions concerning the

autonomous maneuvering of the system were

determined. As a fully autonomous unmanned system

desirably would be able to perform in most or any

circumstances, some restrictions occur for a system

when navigational personnel is included. While

defining the restrictions, it was found that setting the

margin of possible action restraint will lead to a lower

grade of autonomy. Those restrictions were defined in

the ODD.

2.2 Defining the Operational Design Domain

In order to know under which circumstances the

autonomous system can be used, an ODD must be

defined. The ODD describes the conditions in which

the Auto-OOW can safely navigate autonomously [2,

12]. Conditions could be, for example, low traffic

density, good weather, or a sufficient water depth. If

the conditions are no longer met, the ODD is left and

the system must request human assistance [3] or

perform some fallback procedure [15]. A human

officer is then called to the bridge to assess the

situation and take control if necessary. At higher

autonomy levels, the system itself would perform this

so-called fallback with the goal of reaching a system

state of minimal risk [2]. At the highest autonomy

level, i.e., full autonomy, a fallback is not necessary,

since the ODD would then be unrestricted [15].

The process of defining the ODD should start early

in the design process, as defining the ODD also

supports the definition of system and functional

requirements [2]. However, the ODD does not only

help early in the design process, it also supports

evaluating and testing the autonomous system [2].

The ODD can be used to define situations that the

autonomous system must handle on its own as well as

situations in which it requires human assistance. If the

autonomous system does not behave as expected in

the corresponding situations, either the ODD must be

adapted to be in line with the test results or the

autonomous system must be adapted to meet the

requirements of the ODD. Therefore, the process of

defining the ODD is highly iterative [2].

503

Finally, the ODD is also employed during

autonomous operation in order to monitor the current

system state with respect to the ODD. This is referred

to as ODD monitoring [2]. Only, when constantly

monitoring whether the current conditions are within

ODD limits can the system detect possible ODD

violations and request human assistance.

2.3 Methods for gaining the ODD

The process of defining the ODD was started at the

very beginning of the project, as recommended in [2].

As a starting point, a literature search (e.g., [9]) was

used to identify conditions potentially limiting the

ODD of the Auto-OOW. These conditions are referred

to below as ODD factors (see [8]). ODD factors were

subsequently adapted and further refined in close

exchange with the project partners. The close

exchange was particularly important, since it made

sure that potential system boundaries were

considered as ODD factors right away. It also served

to keep all project partners in the loop and to address

their goals for the future capabilities of the

autonomous system.

For these purposes, N = 6 project partners with

nautical (N = 4) and/or technical experience related to

autonomous or sensor systems first completed an

online questionnaire to further specify ODD factors.

The partners with nautical experience possessed on

average 14 years of seafaring experience, with two still

working as navigators at the time of the survey. The

results of the questionnaire were used as a discussion

basis in two online workshops with all project

partners, which eventually produced a preliminary

definition of the ODD and ODD factors. As the ODD

is defined and refined iteratively [2], this preliminary

version has already been adapted to further project

results. The resulting and still preliminary ODD

factors are described in detail below. It is expected

that the definition of the ODD will be further refined

as the project continues.

2.4 Methods for categorization of autonomy levels

For a nuanced definition of autonomy levels the

outputs and findings of the navigational specifications

and the ODD were considered. The fact that a clear

definition of autonomy depends also on other aspects

besides manning and the location of operation (see

IMO levels) arose from the outcomes of the two

working packages in the B0 | B ZERO project. A new

fragmentation into parameters (conceived from

navigational specifications and the ODD) was

derived, which were divided into equivalent

important sub-parameters.

For each parameter a multidimensional

visualization in form of a radar plot (which is also

known as web chart or spider chart) was created. To

avoid confusion within the nautical profession, the

synonym spider chart will be used hereafter. Similar

to a probability distribution function, a profile plotted

on that spider chart represents the relative

distribution measured across more than three

comparative sub-parameters, where each sub-

parameter is represented by an axis on the chart. With

the allocation of a defined value on each sub-

parameter axis, a polygon is created by connecting the

points. The polygon therefore inherits a definite size,

position and shape.

In a further step the spider charts were used as a

tool to categorize the three primary parameters into

five categories. From the geometrical property of each

shape, the areas can be calculated by decomposing the

closed polygon into triangles, with the vertices being

the (sub)-parameter properties.

Afterwards a closer look onto the distribution of

the areas of the developed shapes was taken. As the

area of polygons in a spider chart increases almost as

a square, rather than linearly, the five categories were

derived from a square function and split into the

categories, respectively. The resulting area of the

shape of each parameter was then used as a

categorical value in order to partition the parameters

for the level of autonomy (see Figure 2).

Figure 2. The profile area of a parameter (left) contributing

to the level of autonomy (right)

This procedure is not only applied to each

parameter, but also to the levels of autonomy. The

profile plotted on that chart represents now the

relative distribution measured across at least three

comparative parameters, where each parameter is

represented by an axis on the chart. The so gained

shapes provide a first visual impression about each

parameter’s influence on the level of autonomy.

3 RESULTS

3.1 A preliminary ODD for the B0 | BZERO project

For the preliminary ODD, a total of 15 ODD factors

were defined and assigned to the following

categories: own ship information, route information,

voyage phases, traffic information, weather and

system failures. The ODD was defined in terms of its

limits (see [2]). When an ODD limit is reached, i.e.,

when the state of an ODD factor becomes critical, the

Auto-OOW requires human assistance. In this case,

the human is called to the bridge to gain awareness of

the current situation and react to it.

Human support will be necessary, for example,

when the Auto-OOW detects a close quarter situation

with a target that has the right of way, so the own ship

is required to give way. In this case, a close quarter

situation was defined by a CPA (closest point of

approach) of less than 1.5 NM and a TCPA (time to

closest point of approach) of less than 12 minutes to

the critical target. If such a situation is detected (the

Auto-OOW should prevent this from happening in

the first place), a human officer will come to the

bridge to have a look at both the situation and the

possible maneuvers the Auto-OOW will offer to solve

504

the situation. Either the human will then decide to

select one of the suggested maneuvers and the control

will remain with the Auto-OOW or the human will

take control himself, i.e., the human performs a watch

takeover. In other situations, the only option of the

human officer will be to take over the watch

immediately. For example, if the wind speed exceeds

8 Beaufort, the human must assume control

immediately without the option to let the Auto-OOW

continue.

These two examples clarify that the ODD factors

were divided into two categories based on what

reaction is necessary when the factor state is critical:

ODD factors requiring immediate watch takeovers

and ODD factors where the human OOW decides

about who controls the own ship. In the latter case, an

evaluation of the human is necessary. Hence, a

distinction was made between watch-takeover and

evaluation factors. The two ODD factor categories

indicate in total three situation states with respect to

the ODD. The current situation is either in scope of

the ODD (no human response required), out of scope

of the ODD (watch takeover required) or unclear and

needs to be evaluated (evaluation required). These

three states are consistent with the categorization of

[3].

Furthermore, it was agreed that the master and

selected personnel should have a comparatively high

degree of decision-making freedom with regard to

ODD limits. Critical values of continuously

measurable ODD factors (such as the roll and pitch

period, the visibility, the wind speed etc.) should be

adjustable by the master and selected personnel, since

the definition of which ODD factor conditions are

critical highly depends on the ship type and the crew.

For example, it depends on the ship type and size,

which roll period is to be seen as critical for

autonomous operation. The goal, however, is to

specify default values for most or all ODD factors.

To account for characteristics of voyage phases that

are difficult to measure and to evaluate by the Auto-

OOW, we determined that voyage phases should be

classified as either autonomy-capable or autonomy-

incapable during voyage planning. Then, when

entering autonomy-capable voyage phases during the

voyage, autonomous operation is feasible, as long as

the current situation is also in scope of the ODD.

However, autonomous operation is in any case

infeasible in voyage phases labelled as autonomy-

incapable. Before entering these voyage phases, the

Auto-OOW has to hand over the watch to the human

officer. Specifically, the following voyage phases

should be autonomy-incapable in the project B0 |

BZERO: shallow waters, VTS areas, dangerous areas

and areas where bad weather is expected. The labels

autonomy-capable and autonomy-incapable will be

tied to waypoints defined during route planning.

Therefore, waypoints are either not ODD relevant,

lead to scheduled ODD exits or scheduled ODD

entries (if all ODD factors allow the entry).

A summary of all ODD factors and the conditions,

in which they are critical to the ODD, as well as the

necessary human responses, is displayed in Table 1.

For some ODD factors, default critical conditions

remain to be defined. For the ODD factor under keel

clearance (UKC) as well as for the factor combination

CPA and TCPA, two distinct default critical

conditions were defined (see Table 1).

3.2 Categorization of ODD and navigational limitations

As the restraints of the navigational specifications and

the outcomes of the ODD show, several parameters of

ship’s navigation are affected and influence the grade

of autonomy directly. Hence, it seems useful to

generally classify those parameters into categories for

environment, traffic and own ship (see Table 2).

Table 1. Defined ODD factors with their default critical conditions and required human reactions

__________________________________________________________________________________________________

category ODD factor default critical condition necessary reaction

__________________________________________________________________________________________________

own ship roll period to be defined evaluation

information roll angle at least 10° evaluation

pitch period to be defined evaluation

pitch angle at least 25° evaluation

speed to be defined evaluation

UKC at most 50m watch takeover

at most 0.7 times of the estimated UKC evaluation

voyage phases time to autonomy-incapable area at most 12 minutes watch takeover

route cross track distance to be defined evaluation

information deviation between estimated and at least 15 minutes watch takeover

scheduled arrival time at next waypoint

traffic traffic density heavy traffic (to be defined) watch takeover

information CPA & TCPA Target has to give way, CPA at most 1.5 NM evaluation

and TCPA at most 15 minutes

Own ship has to give way, CPA at most evaluation

1.5 NM and TCPA at most 12 minutes

loss of radar targets loss happens within a radius of at most evaluation

12 NM and target AIS data are not available

weather wind force at least 8 Beaufort watch takeover

visibility at most 3 NM evaluation

failures system failure at least one of the following systems failed: watch takeover

ECDIS, Radar, AIS, gyro compass, echo

sounder, GPS, THD, propulsion system,

steering gear, alarm system, automatic track

control, automatic heading control

__________________________________________________________________________________________________

505

Table 2. ODD and navigational parameters

_______________________________________________

Environment Traffic Own Ship Factors

_______________________________________________

Wave height Number of ships Motion (Roll/Pitch)

Wave direction CPA Speed

Wind speed TCPA UKC

Visibility Voyage Phase

Time of day Cross-track distance

_______________________________________________

Those parameters can be seen as the restraining

settings for any autonomous ship. The narrower the

range of action or limit for each parameter is set, the

lesser the range of possible action is for an

autonomous system. Restraining the parameters

means to set some boundaries, in which the

autonomous system is allowed to freely operate. Each

parameter limit thereby has different impacts on the

navigation. To illustrate this in more detail and to

provide an example, the parameters including their

sub-parameters are described further in the following.

3.2.1 Categorization of Environmental Parameters

Wave height was named as one restricting sub-

parameter for the B0 | BZERO project. As the wave

height varies regionally, seasonally and temporarily, a

ship might encounter all ranges of wave heights on its

voyage. This sub-parameter directly influences the

ship’s motion and can result in heavy rolling or

structural load onto the ship’s hull. Additionally,

increasing ship motions could cause damage to the

cargo, when inappropriately secured. It is anticipated

that the wave height is an essential sub-parameter, as

an autonomous system needs to react to areas of

extreme wave heights, or better to avoid them in the

first place. A system needs to monitor the ship’s

motions, while keeping track using the ship’s engines

and rudder.

A related sub-parameter is the wave direction

relative to the ship. During unfavorable wave

directions and periods, the ship can encounter heavy

motions, which can have severe consequences both for

the crew (when manned) and the ship. Heavy ship

motion can lead to severe motion sickness for the

crew and to parametric rolling [4] followed by

capsizing of the whole ship [16].

Besides wave height and direction, wind speed

was also called as one of the sub-parameters, as wind

speed can cause damage to cargo and loading. Severe

wind speeds in terms of storms and gusts can

influence the course-keeping abilities of the ship.

Wind speed is directly related to wave height, as the

latter is affected and caused by the former and by the

time of exposure to those winds [10].

Another important sub-parameter, originating

from the COLREGs [6], is visibility, which can be

reduced because of fog, dust, sandstorms, heavy rain

or snow. Visibility affects which navigation rules

apply, since they depend on whether another ship is

in sight or not. These human-centered regulations are

currently also in force for automated and autonomous

ships, although those may be more capable than

humans to navigate without optical eye-sight.

However, during the B0 | BZERO project that

parameter was rated as still to be considered while

designing the autonomous system.

Another challenge for automated and autonomous

systems is the time of day. The optical representation

of a ship during daytime is its silhouette, which can be

seen as soon as the ship arrives at the other ship’s

horizon. The silhouette decreases with fading daylight

and at night only the ship’s navigational lights are

visible. At night, it is more challenging to determine

the heading of the other ship and thereby the risk of

collision. Similar to visibility, time of day is a very

human-centered indicator, and during the project

work it was manifested, that the autonomous system

has to identify and react to this parameter.

The previously mentioned environmental sub-

parameters are split into six sections each and can be

seen in Table 3. It can be stated that low range sections

restrict the ship’s autonomy to a higher degree than

upper range sections, whereas upper range sections

allow more decision and reaction freedom for the

autonomous system. For the sub-parameter time of

day, the entries are used twice to cover all six sections

due to a shortage of possible options. The

environmental restrictions are independent from each

other, which means that a heterogenous distribution is

possible and expectable. The sections are used as the

scale on the vertices of the spider chart.

Table 3. Environmental restriction parameter sections

_______________________________________________

Scale Wave Wave Wind Visibility Time of

height direction force day

[m] [Bft] [nm]

_______________________________________________

1 < 2 none < 2 15+ Day

2 2-4 head 2-4 12-15 Day

3 4-6 bow 4-6 8-12 Twilight

4 6-8 beam 6-8 6-8 Twilight

5 8-10 quartering 8-10 2-6 Night

6 10+ following 10+ 0-2 Night

_______________________________________________

In the following three different random application

cases of different restrictions due to environmental

sub-parameters are shown in Figure 3 and explained

in the following. The vertices (scale of sections) of

each sub-parameter in the use cases create a polygon.

The smaller the area of the polygon is, the lower the

degrees of freedom for the autonomous system are.

Application Case 1. No environmental restrictions

apply. The ship system is free to maneuver within all

environmental conditions. This means that the ship’s

system has to be able to navigate in every wave height

and direction, at every wind speed, as well as during

limited to no visibility and at every time of the day.

The area of the polygon is the biggest area in this

application case compared to the other two.

Application Case 2. Medium restrictions apply.

This application case is taken from the B0 | BZERO

project, where a safe maneuvering frame is developed

for a real application on a ship. Safe operating limits

are determined to be wind speeds up to 8 Bft and

wave heights up to 6 m. Wave direction, visibility and

daytime are unrestricted. Since in comparison to case

1, wind speed and wave height are restricted, the area

of the polygon is smaller indicating a lower overall

autonomy level.

Application Case 3. Large restrictions apply. The

autonomous system is only entitled to maneuver

inside very narrow limits of wind speeds up to 4 Bft,

wave heights up to 4 m, and wave directions only

506

from a heading direction. It is also limited to very

good visibility (up to 15 nm) and daytime use only.

Due to the large restrictions, the area of the polygon is

small, indicating very restricted autonomy.

Figure 3. Comparison of three different application cases

within environmental restrictions

Away from the three presented application cases,

various other cases with varying combinations of

environmental restrictions are possible. What one can

see in Figure 3 is that with narrowed limits the

polygon’s area decreases. Expressed in geometry

terms this means that the enclosed area A of a

polygon P is dependent on the location of the vertices

p on each axis. The largest possible area is given in

case 1, whereas case 3 only covers a very small area.

The environmental restrictions can be summarized

into one parameter named “environment” by

assigning each polygon size to one of five categories

ranging from worst to good (see Table 4). The

polygon area size ranges from 2.5 up to 25.

Table 4. Assigning different polygon area sizes to categories

representing the overall environmental condition

_______________________________________________

Area Category

_______________________________________________

< 2.5 Worst

2.5 - 4 Bad

4 - 9 Medium

9 - 16 Fair

16 - 25 Good

_______________________________________________

3.2.2 Categorization of Traffic Parameters

A similar categorization as for the environmental

parameters was carried out for the traffic and own

ship parameters. The three traffic sub-parameters are

the number of ships, the CPA- and the TCPA-values

(see Table 2). The focus of the selection of traffic

parameters lays mainly on the functionalities and

capabilities of the autonomous system. The number of

ships in the vicinity of the own ship is an indicator of

traffic density and was also a limiting factor when it

comes to processing power of navigational devices.

Within the B0 | BZERO project “heavy traffic”

remains to be defined, since its definition depends on

the capabilities of the autonomous system. The

outcomes of the pending system tests will provide a

better insight into these. As a starting point for

categorizing different traffic densities, the number of

ships within a range of 12 nm (normal optical sight) as

displayed in Table 55 will be used. CPA- and TCPA-

values were categorized based on standing orders and

collision regulations and can also be retrieved from

Table 5. Autonomous systems that are able to deal

with the smallest CPA and TCPA values have the

highest decision making and action freedom. Again,

similar to the environmental restrictions the

navigational parameters are arbitrary and do not

necessarily depend on each other.

Table 5. Navigational parameter sections

_______________________________________________

Scale Number of ships within 12 nm CPA TCPA

_______________________________________________

1 < 5 5+ NM 30+ min

2 20 3-5 NM 18-30 min

3 55 2-3 NM 12-18 min

4 115 1-2 NM 6-12 min

5 200+ <1 NM < 6 min

_______________________________________________

After traffic parameters were categorized similar to

the environmental parameters, the list displayed in

Table 6 emerged.

Table 6. Traffic parameter categorization

_______________________________________________

Area Category

_______________________________________________

< 1.5 Low

1.5 – 4 Few

4 - 9 Average

9 - 16 Increased

16 - 25 High

_______________________________________________

3.2.3 Own Ship Factors

Own ship factors were categorized similarly to

environmental and traffic parameters (see Table 7).

The sub-parameters were motion (roll and pitch),

speed, under keel clearance, voyage phase and cross

track distance. As the holistic ship’s motion is very

ship and ship-type specific, roll and pitch angles will

be determined. The ship’s speed is another sub-

parameter which is very ship specific. Therefore, the

categorization will not take the absolute speed into

account, but the percentage of design speed.

Furthermore, the under-keel clearance will not be

categorized in absolute meters, but in relation to the

draft of each ship. Special attention is paid to the

voyage phases, which were determined during the B0

|BZERO project as time slots and areas of a ship’s

voyage correlating to the passage planning. Here, the

categories were chosen as follows: The least degree of

freedom is assumed, when the ship is only allowed to

navigate in declared areas. Further decision-making

scope is achieved by allowing the ship to navigate

during sea-passage up to the highest level of freedom,

the berthing. The higher level of a voyage phase does

always include also the lower levels, i.e., a ship that

can manage fairways autonomously (section 4) will

also be able to manage sea passages autonomously

(section 3). The cross-track error will also not be

categorized in absolute numbers, but in relation to the

ship’s length.

Table 7. Own ship sub-parameter sections

_______________________________________________

Scale Roll/ Speed UKC Voyage XTD

Pitch [% of [m phase [m

Angle design perm perm

speed, kn] draft] LOA]

_______________________________________________

1 2°/5° 70-80 32+ Declared areas 20+

2 10°/20° 50-85 16 Sea passage 10

3 25°/30° 40-90 8 Fairways 5

4 40°/50° 20-95 4 Pilotage 2

5 40°+/50°+ 0-100 < 2 Berthing < 1

_______________________________________________

507

The categorization of own ship factors based on

the polygonal areas leads to the categories displayed

in Table 8.

Table 8. Categorization of own ship factors

_______________________________________________

Area Category

_______________________________________________

< 1 Most restricted

1 - 4 Heavily restricted

4 - 9 Moderately restricted

9 - 16 Less restricted

16 - 25 Least restricted

_______________________________________________

3.3 Level of autonomy

Looking into the restrictions by the explained

categories and the aspects of autonomy stated in

section 1.1 it is clear that a combination of them lead

to different levels of autonomy. Therefore, a

categorization of inputs is conducted and for different

constraints the outcomes can be seen in Table 9.

A last step for a new declaration of levels of

autonomy is the categorization of the in Table 9

assigned parameters according to the polygon area,

using the aforementioned approach. That means that

possible polygon areas are divided into categories

according to a square function. In detail the results

can be seen in Table 9.

Table 9. Proposed autonomy levels

_______________________________________________

Area Level of autonomy

_______________________________________________

< 2.5 Assistance

2.5 - 4 Partial Automation

4 - 9 Conditional Autonomy

9 - 16 High Autonomy

16 - 25 Full Autonomy

_______________________________________________

The lowest level of autonomy within this

framework is called “Assistance”, as it has similar

characteristics as a decision support system. It can be

seen that the lowest possible parameter categories are

used, which means the least possible freedom for the

navigational systems and still manning on the ship,

which makes it a support system. Within partial

automation, conditional and high autonomy, some

parameters are set to an advanced level giving the

autonomous system more degrees of freedom. As this

model proposes a free distribution of the parameters,

it is not precisely defined which parameter lead to one

of those levels. Different combinations are possible, as

it can be seen from the autonomy application cases

(see Figure 4). Although some application cases seem

to have a large degree of decision freedom, singular

restriction narrows the end result. That leads to the

fact, that the application cases green and yellow fall

into the category conditional autonomy and the

application cases purple and red fall into the high

autonomy level. In the full autonomy level, the

parameters are nearly set to full degree of freedom

and the autonomous system is free to maneuver at

almost any circumstance.

3.4 Use Cases

For the demonstration of use cases, the parameters are

arbitrary. In Figure 4, five different use cases are

shown. It can be seen that the restriction of parameters

reduces the polygon area, leading to a lesser level of

autonomy. The applications were taken from realistic

nautical conceivable situations, shortly described in

Table 11.

Figure 4. Comparison of different autonomy application

cases

Table 10. Parameters for level of autonomy

__________________________________________________________________________________________________

Scale Manning Control mode Sea area Environment Traffic density Own ship factors

__________________________________________________________________________________________________

1 Fully manned on duty Local control Open sea Good Low Most Restricted

2 Fully manned off duty Local supervision Coastal Fair Few Heavily Restricted

3 Reduced manning Remote control Traffic separation Medium Average Moderately restricted

off duty schemes

4 Reduced/no manning, Autonomous Confined waters Bad Increased Less Restricted

but passengers on board supervision

5 No manning Autonomous Confined & Worst High Least Restricted

control shallow waters

__________________________________________________________________________________________________

508

Table 11. Application situation for autonomy level presentation

__________________________________________________________________________________________________

Manning Control mode Sea area Environment Traffic density Own ship factors

__________________________________________________________________________________________________

Blue Fully Decision support Confined and Worst High Least restricted

system with able shallow waters

human OOW

Red Fully Remote Confined waters Bad Increased Less restricted

Green Reduced Remote Open sea Bad Average Moderately

restricted

Purple Reduced manning Autonomous Coastal Medium Few Less restricted

+ passengers

Yellow No manning Autonomous Open sea Fair Low Most restricted

__________________________________________________________________________________________________

4 DISCUSSION

Commercially viable fully autonomous ships are

rather unrealistic in the near future [1]. Therefore,

some form of constrained autonomy, as in the B0 |

BZERO project, is pursued. The project aims at

developing a constrained autonomous system that can

handle most situations autonomously, while it still

requires human assistance on the bridge when certain

situations are encountered. To know exactly when

human assistance is required, an ODD and

navigational specifications must be defined and

specified. The aim of this paper was twofold. First, we

wanted to provide a preliminary definition of the

ODD and navigational specifications of a constrained

autonomous vessel using the B0 | BZERO project as

an example. Second, we aimed at providing a

categorization of important parameters that need to be

considered when evaluating a vessel’s autonomy

based on our defined ODD and navigational

specifications.

As our defined ODD and navigational

specifications show, several parameters must be

considered when determining the exact manifestation

of a vessel's autonomy. The parameters can be divided

into several categories and can take on various values

and forms. Our categorization makes it possible to

flexibly select which values and forms lie within the

ODD and which do not. The more possible parameter

values and forms are included in the ODD, the higher

the degree of the ship’s autonomy. Due to the high

flexibility of the categorizations, the degree of

autonomy can be tailored to specific ship conditions,

to the needs of a shipping company and to any system

limitations. The parameters’ values and forms selected

for the ODD can also subsequently be used to develop

scenarios to test the performance of the selected

degree of autonomy [2]. Such tests also provide

insights into which parameters need to be adjusted

and to what extent.

It remains to be tested, in general, how the ODD

proves itself in practice. Currently, the defined ODD

is still very strongly oriented towards human

capabilities and less towards system boundaries. By

conducting more tests with the help of the defined

scenarios, the focus could shift more towards system

boundaries. Furthermore, individual ODD factors

such as heavy traffic or critical values for the roll and

pitch period remain to be defined. For certain ODD

factors such as visibility, it is not yet clear, how they

can be measured. Another limitation of the ODD is

that the human still possesses considerable decision-

making freedom. The human, for example, is

responsible for classifying voyage phases either as

autonomy-capable or autonomy-incapable during

voyage planning. The high degree of decision making

freedom is accompanied by a high degree of human

responsibility and allows room for human error [1].

Furthermore, it is proposed to define the levels of

autonomy of a ship based on the limitations of the

parameters manning, control mode, sea area,

environment, traffic density and own ship restrictions.

The visualizations have shown that is possible to

derive a categorization of autonomy levels with a

further segmentation of parameters and sub-

parameters. Nevertheless, further application cases as

well as further segmentation and parameter

description might lead to different categories and

results. Limits also occur in the description of wave

height e.g., as monster waves might be experienced.

That leaves also the discussion open for using

limitations for some parameters, as they might be

exceeded or undercut. Further criticism can be

directed towards the arbitrary combination of

parameters, as some are not totally free combinable, as

“no manning” with “decision support systems”.

Another set-back of that system is the composition

of parameters. As the model is designed to use only

the area of the created polygons, it is no longer

possible to infer the exact degrees of freedom for the

individual parameters from the area. The same occurs

during the specification of parameter’s degrees of

freedom via specifying the sub-parameters. That is

why a continuously moderate level of sub-parameters

would obtain the same level of autonomy as a

combination of high- and low-level sub-parameters.

5 CONCLUSIONS

In this paper the approach of defining the operational

design domain for an autonomous navigation system

is presented. Together with the outcomes of the

navigational specifications of the B0 | BZERO project

a new system for the determination of the level of

autonomy is proposed. The theoretical framework

behind our approach relies on the elaboration of ODD

parameters for the use case of an 8-hour unmanned

bridge as well as on the fact that the data on spider

charts create polygon shapes allowing to measure

multidimensional performance as well as categorizing

of parameters. Applying this approach shows a wider

spread of distributing parameters to the levels of

autonomy than in the IMO model, as well as the

opportunity to determine already the level of

autonomy of a ship in the early stage of specification

of the systems. Until now, this paper serves as

theoretical foundation, the practical use has to be

shown in the future and it has to prove its potential.

509

REFERENCES

1. Abilio Ramos, M., Utne, I.B., Mosleh, A.: Collision

avoidance on maritime autonomous surface ships:

Operators’ tasks and human failure events. Safety

Science. 116, 33–44 (2019).

https://doi.org/10.1016/j.ssci.2019.02.038.

2. Colwell, I., Phan, B., Saleem, S., Salay, R., Czarnecki, K.:

An Automated Vehicle Safety Concept Based on

Runtime Restriction of the Operational Design Domain.

In: 2018 IEEE Intelligent Vehicles Symposium (IV). pp.

1910–1917 (2018).

https://doi.org/10.1109/IVS.2018.8500530.

3. Farah, H., Bhusari, S., Gent, P. van, Babu, F.A.M.,

Morsink, P., Happee, R., Arem, B. van: An Empirical

Analysis to Assess the Operational Design Domain of

Lane Keeping System Equipped Vehicles Combining

Objective and Subjective Risk Measures. IEEE

Transactions on Intelligent Transportation Systems. 22,

5, 2589–2598 (2021).

https://doi.org/10.1109/TITS.2020.2969928.

4. France, W.N., Levadou, M., Treakle, T.W., Paulling, J.R.,

Michel, R.K., Moore, C.: An Investigation of Head-Sea

Parametric Rolling and Its Influence on Container

Lashing Systems. Marine Technology and SNAME

News. 40, 01, 1–19 (2003).

https://doi.org/10.5957/mt1.2003.40.1.1.

5. International Maritime Organization: Autonomous

Shipping,

https://www.imo.org/en/MediaCentre/HotTopics/Pages/

Autonomous-shipping.aspx, last accessed 2021/03/26.

6. International Maritime Organization: Colregs-

international regulations for preventing collisions at sea:

Articles of the Convention on the International

Regulations for Preventing Collisions at Sea. (1972).

7. Kim, M., Joung, T.-H., Jeong, B., Park, H.-S.:

Autonomous shipping and its impact on regulations,

technologies, and industries. null. 4, 2, 17–25 (2020).

https://doi.org/10.1080/25725084.2020.1779427.

8. Koopman, P., Fratrik, F.: How many operational design

domains, objects, and events? Presented at the AAAI

Workshop on Artificial Intelligence Safety (2019).

9. Mathes, S., Herberg, J., Berking, B., Behnke, J., Jonas, M.:

Functional scope and model of integrated navigation

systems - a toolbox for identification and testing. (2001).

10. Meucci, A., Young, I.R., Aarnes, O.J., Breivik, Ø.:

Comparison of Wind Speed and Wave Height Trends

from Twentieth-Century Models and Satellite

Altimeters. Journal of Climate. 33, 2, 611–624 (2020).

https://doi.org/10.1175/JCLI-D-19-0540.1.

11. Parasuraman, R., Sheridan, T.B., Wickens, C.D.: A model

for types and levels of human interaction with

automation. IEEE Transactions on Systems, Man, and

Cybernetics - Part A: Systems and Humans. 30, 3, 286–

297 (2000). https://doi.org/10.1109/3468.844354.

12. Porathe, T., Hoem, Å., Rødseth, Ø., Fjørtoft, K., Johnsen,

S.O.: At least as safe as manned shipping? Autonomous

shipping, safety and “human error.” In: Haugen, S.,

Barros, A., Gulijk, C. van, Kongsvik, T., and Vinnem, J.E.

(eds.) Safety and Reliability – Safe Societies in a

Changing World. pp. 417–425 CRC Press, London

(2018).

13. Rødseth, Ø., Nordahl, H.: Definitions for Autonomous

Merchant Ships. (2017).

https://doi.org/10.13140/RG.2.2.22209.17760.

14. Rødseth, Ø.J.: Defining Ship Autonomy by

Characteristic Factors. In: 19-26. pp. 19–26 SINTEF

Academic Press, Busan, Korea (2018).

15. SAE International: J3016B: Taxonomy and Definitions

for Terms Related to Driving Automation Systems for

On-

Road Motor Vehicles,

https://www.sae.org/standards/content/j3016_201806/,

last accessed 2021/05/10.

16. Zakaria, N.M.G.: Effect of ship size, forward speed and

wave direction on relative wave height of container

ships in rough seas. The Institution of Engineers. 72, 3,

21–34 (2009).