Possible COLREGs Failures under Digital Helmsman of Autonomous Ships

Abstract

Autonomous navigation will play an important role in the future of the shipping industry. Hence, this study illustrates several concepts that should support the navigation side in relation to the collision avoidance of autonomous ships. The concept of system intelligence, i.e., cloned by human navigators, as the digital helmsman to navigate future vessels is discussed in the first part of this study. That can provide an adequate solution to the ship controllability problem. A collision avoidance framework. i.e., based on fuzzy logic, developed to support the digital helmsman is discussed in the second part of this study. The same collision avoidance framework complements the digital helmsman. Future vessels will navigate in a mixed environment, where manned, remote controlled and autonomous vessels are interacting. Hence, the proposed collision avoidance framework, as a decision support feature based on the respective navigational rules and regulations, should support both humans as well as systems to make appropriate actions in such a navigation environment. It is expected to have an adequate consistency between human and system collision avoidance actions to preserve the integrity of system level intelligence. In the third part of this study, the consistence between human and system decisions/actions in critical collision avoidance situations with the main intention of identifying possible regulatory failure situations in a simulated environment are investigated by using the same collision avoidance framework.

Full text

In Proceedings of the MTS/IEEE OCEANS ’19, Marseille , France, June, 2019.
Possible COLREGs Failures under Digital Helmsman
of Autonomous Ships
Lokukaluge P. Perera
UiT The Arctic University of Norway,
Tromsø, Norway,
Prasad.Perera@uit.no.
Bjørn-Morten Batalden
UiT The Arctic University of Norway,
Tromsø, Norway,
Bjorn.Batalden@uit.no
Abstract—Autonomous navigation will play an important role
in the future of the shipping industry. Hence, this study
illustrates several concepts that should support the navigation
side in relation to the collision avoidance of autonomous ships.
The concept of system intelligence, i.e., cloned by human
navigators, as the digital helmsman to navigate future vessels is
discussed in the first part of this study. That can provide an
adequate solution to the ship controllability problem. A collision
avoidance framework. i.e., based on fuzzy logic, developed to
support the digital helmsman is discussed in the second part of
this study. The same collision avoidance framework complements
the digital helmsman. Future vessels will navigate in a mixed
environment, where manned, remote controlled and autonomous
vessels are interacting. Hence, the proposed collision avoidance
framework, as a decision support feature based on the respective
navigational rules and regulations, should support both humans
as well as systems to make appropriate actions in such a
navigation environment. It is expected to have an adequate
consistency between human and system collision avoidance
actions to preserve the integrity of system level intelligence. In
the third part of this study, the consistence between human and
system decisions/actions in critical collision avoidance situations
with the main intention of identifying possible regulatory failure
situations in a simulated environment are investigated by using
the same collision avoidance framework.
Keywords—Autonomous Ship, Digital Helmsman, COLREGs,
Ship Maneuvering, Collision Avoidance, Regulatory Failures,
Crash Stop Maneuvers, Critical Collision Conditions.
I. INTRODUCTION
Autonomous ships will introduce the next technological
revolution in the shipping industry as a part of Shipping 4.0
[1]. However, the required technologies and infrastructure to
make autonomous navigation in deep-sea shipping are still
underdeveloped. Therefore, it is expected that the first-
generation autonomous vessels will support short-sea shipping
type activities, i.e., ferries and inland shipping. Furthermore,
the existing infrastructure in ports and inland waterways can
also be adopted to make short-sea autonomous shipping a
reality in the near future. However, close vessel encounters in
congested ship routes can be one of the biggest challenges that
the shipping industry can face in such situations. Hence,
collision avoidance in close ship encounters should be an
important research concept to study in the future, and that can
improve the navigation safety of future vessels in short-sea
shipping.
The outcome of a collision risk assessment in a vessel
encounter situation can be used to identify possible collision or
near miss situations. In general, the respective collision risk in
realistic ship encounters is difficult to assess, accurately
because not only vessel positions but also their orientations
should also be considered into the same. Inaccuracy in vessel
navigation details can degrade the collisions risk assessment
process, and that may increase the risk of ship collisions. The
outcome of the collision risk assessment process can eventually
be used as a decision supporting feature for future ships, where
intelligent systems can make navigation decisions. Such a
decision support feature should be based on the respective rules
and regulations of ship navigation. The International
Regulations for Preventing Collisions at Sea 1972 (COLREGs)
facilitates to avoid ship collisions in open sea areas. However,
additional local navigational rules and regulations can be
enforced on vessels especially in confined waters and maritime
traffic lanes. The respective system intelligence under the
decision support feature should be evaluated to verify their
decision consistency and regulatory capabilities in handling
various ship navigation situations, especially under mixed
environmental conditions, where remote-controlled,
autonomous and manned vessels are interacting. The outcome
of this process can improve the navigation safety of future
autonomous vessels.
II. SHIP INTELLIGENCE
A. Mixed Environment
The first-generation short-sea autonomous ships will
navigate in a mixed-environment with close ship encounters.
The vessel interactions in such environments can compromise
navigation safety since both humans and systems are making
the respective decisions [2]. Humans are considered as rational
decision-makers, because humans systematically select
possible choices based on reasons and facts. On the other hand,
system-based decision-making processes that can mimic
human behavior are yet to be formulated and that will be
purely based on a given set of rules. However, it is expected
that onboard intelligent systems can develop human friendly
decision-making capabilities in the future to support
autonomous vessels. Some decisions can be executed as
collision avoidance actions through rudder and propeller
control systems, therefore ship controllability in relation to
vessel seakeeping and maneuvering behavior should also be
considered.

B. Ship Controllability
Vessels are considered to be under-actuated systems
associated with heavy body inertia. In general, it is impossible
to achieve the full controllability of ocean-going vessels from
rudder and propeller control systems, especially under rough
weather conditions. However, a limited number of over-
powered vessels may be able to preserve their controllability
under calm weather conditions. Such vessels are not the scope
of this study. Various unexpected and undesirable motions
relate to vessel seakeeping and maneuvering behavior can be
expected in under-actuated vessels, i.e. due to the respective
hydrodynamic and wind forces and moments on the hull and
superstructure. Furthermore, vessels are slow responsive
systems with considerable time-delays. Therefore, vessels can
take considerably longer periods to response to rudder and
propeller changes and that can create additional difficulties in
ship controllability.
Even though various advanced controllers have been
proposed by the research community to address such
challenges in ship navigation, that has never been implemented
to realistic ocean-going situations mainly due to system-model
uncertainties. The outcome is that the respective controllers
cannot preserve their robustness and stability in realistic ocean-
going conditions. It can be concluded that ship controllability
is a complex problem and lack of adequate mathematical
models may result in ocean-going vessels with inadequate
controllability. The same issues in ship navigation have often
been ignored by the research community. Therefore, adequate
solutions to overcome the challenges in under actuated ships
have yet to be investigated by the future researchers.
C. Digital Helmsman
Machine learning and artificial intelligence-based
techniques, i.e., deep neural networks, can provide a smart
solution to the ship controllability problem [3]. Deep neural
networks, as a part of deep learning, can provide adequate
system intelligence into ocean-going vessels to navigate as
agent-based systems by having a human behavior based
mathematical framework. However, these deep neural
networks should be trained by human navigators, similar to the
autonomous car [4] to develop such behavior models. In
general, deep neural networks can capture human helmsman
behavior through large-scale real-world ship navigation data
sets. Hence, an agent based on deep learning is classified as the
digital helmsman due to their digital nature and human
behavior in ship navigation. Since future vessels will be
navigated by the digital helmsman, the respective collision
avoidance actions should also be executed by the same system
intelligence. Therefore, it would be an advantage, if the
proposed decision support feature for ship collision avoidance
used by human navigators, can also be used by the digital
helmsman to navigate future vessels.
D. Collision Avoidance
Collision avoidance will be an important part of
autonomous vessels, especially in close ship encounter
situations. It is expected that the digital helmsman executes the
respective collision avoidance actions based on the decision
support feature. Ocean-going vessels are not navigating in
well-defined sea routes, therefore one vessel can observe other
vessels with various course-speed conditions with different
headings in its vicinity. The digital helmsman should be aware
of such vessel encounters and take appropriate actions to avoid
a possible collision or near miss situations, especially in mixed
environments. Since these vessels can have various interactions
within the navigation environment, the collision avoidance
actions taken by both humans as well as systems should have
an acceptable level of consistency. The important concepts
discussed previously in relation to collision avoidance in
autonomous ships are summarized in Figure 1.
Each ship encounter situation should be analyzed to
estimate the respective collision risk. Then, the same collision
risk should be used to formulate appropriate collision
avoidance decisions. It is expected that the respective collision
avoidance decisions should also be similar to human decisions.
Since the digital helmsman is trained by human navigators, the
respective decisions should also be understood by such
systems. The combination of those two components, i.e. the
collision risk estimation and collision avoidance decisions, can
be categorized the decision support feature. That can be used
by both humans as well as systems to make appropriate
collision avoidance actions. The collision avoidance actions
that can be taken by a ship navigator divided into two
categories: speed change (to increase or decrease) and course
change (to the starboard and port). That should be done
through ship propulsion and rudder control systems. In general,
the alteration of course is the most preferred collision
avoidance action, since low speed conditions can also reduce
ship maneuverability.
Since both humans and systems convert these collision
avoidance decisions into actions, their regulatory compliance
should also be evaluated. Therefore, any inconsistency between
human and system collision avoidance actions can be
eliminated to preserve the integrity of system level intelligence.
This study investigates the respective regulatory compliance in
relation to possible COLREGs failure situations under several
simulated system decision making situations in ship
navigation. That outcome can also help to improve the
behavior of the digital helmsman and identify possible
modification on the respective navigation rules and regulations
to support future vessels.
III. THE COLREGS
A. Decision Consistency
It is expected that the first-generation autonomous vessels
should follow the existing rules and regulations of the
COLREGs due to the mixed environment. However, the
COLREGs will be interpreted by both humans as well as
systems during these ship encounters. To avoid possible
collisions or near-miss situations, the interpretations should
have a high level of consistency. Developing a collision
avoidance framework with adequate consistency between
Fig. 1. The important components of a ship collision avoidance framework.

human and system interpretations can be an appropriate
solution. A mathematical framework based on fuzzy logic, i.e.,
a computational intelligence approach, that has been developed
previously to support the same consistency between human and
system interpretations of the COLREGs is considered in this
study [5].
B. Fuzzy Logic in COLREGs
Since fuzzy logic can imitate a human type decision-
making process, many human friendly industrial applications
are based on the same [6]. Fuzzy logic is considered as the
method to formulate the COLREGs into a human friendly
decision support feature. Hence, the decision support feature
can support the digital helmsman to navigate future vessels
and avoid a possible collision or near miss situations. The
respective fuzzy membership functions are formulated in a
way so that the ship navigators’ understanding of the
COLREGs rules and regulations in relation to various ship
encounter situation can be captured into a mathematical
framework. Furthermore, this mathematical framework should
be exhaustive, consisting of all possible ship encounter
situations, the input fuzzy membership functions, the collision
avoidance decisions, and the output fuzzy membership
functions - that should be taken by ship navigators.
One should note that the collision avoidance decisions
should be based on the COLREGs rules and regulations.
Those rules and regulations are incorporated by mapping the
input and output fuzzy membership functions of a Mamdani
type Fuzzy inference system [7]. The Mamdani type Fuzzy
membership functions can represent an IF-THEN-ELSE type
rule-based logical structure. Hence, that can provide a good
mathematical framework to capture the respective regulatory
structure of the COLREGs. The COLREGs should be aligned
with the fuzzy membership functions to achieve these
objectives, as mentioned before. Further details on this fuzzy
logic based mathematical framework for ship collision
avoidance are presented in the work of Perera et al. [5, 8]. The
same framework can support the implementation of the
COLREGs rules and regulations under the digital helmsman.
While systems are making decisions, there is a possibility that
some regulatory failures can occur through the same
framework and that should be investigated further. Hence, the
decision support feature is evaluated in a simulated
environment for the regulatory failure situations, assuming
that the COLREGs are interpreted by the digital helmsman.
C. Collision Risk Estimation
Any multiple vessel encounter situations can be separated
into multiple two-vessel encounter situations. This collision
avoidance framework is developed with respect to a two-
vessel encounter situation (see Figure 2). It is also assumed
that these are power-driven vessels. The vessel that has the
collision avoidance framework is named the “own vessel” and
the other vessel is named the “target vessel”. It is expected
that both own- and target vessels are approaching a close ship
encounter situation, where the respective collision risk should
be estimated (COLREGS, 1972 - Rule 7 Risk of collision).
The collision risk in a ship encounter situation should take
account of the respective collision avoidance decisions (see
Figure 1) (COLREGS, 1972 - Rule 8 Action to avoid
collision) [9].
The navigation area of the own vessel is divided into three
regions: general collision risk region, critical collision risk
region and vessel domain. These regions are introduced to
formulate the COLREGs into fuzzy logic adequately through
the respective membership functions. Three simplification
steps are introduced in this situation (Figure 2). One
assumption is that the vessels are moving in straight-line
trajectories. Secondly, the vessel course-speed vectors and
headings are in the same direction [10]. Even though vessels
navigation with complex maneuvering trajectories have been
presented in the literature [11], such situations are considered
as beyond the scope of this study. Thirdly, the study does not
include situations for which Rule 19 Conduct of vessels in
restricted visibility, as this rule cross out the rules 11 to 18
(Conduct of vessels in sight of one another).
One should note that having a trajectory intersection point
between the own and target vessels does not mean that it is a
possible collision or near-miss situation. These vessels can
pass the intersection point at different time intervals with
relatively safe distances. Therefore, the collision risk between
these vessels can be negligible. If the collision risk is
negligible, none of the vessels should make any course or
speed changes. The vessel should keep their course-speed
vectors in such situations concerning the COLREGs.
However, this concept is often ignored by the research
community, especially under path planning type algorithms,
where unnecessary course speed change actions can be taken
by the respective vessels.
A possible collision situation between two vessels can be
determined by the relative bearing (see Figure 2), i.e., if any
relative bearing changes cannot be observed within a
considerable time period and the distance between the vessels
diminish, the vessels are heading towards a possible collision
or near miss situation. This bearing and distance concept is
transformed into the relative motion between two vessels,
where the relative navigation trajectory of the target vessel
with respect to the own vessel is estimated. The concept is
adopted to illustrate the mathematical definition of a possible
ship collision situation under this framework: if the relative
navigation trajectory of the target vessel intercepts the own
vessel domain, then that situation is categorized as a collision
or near-miss situation. In general, the relative course-speed
vector of the target vessel can be used to estimate the relative
navigation trajectory of the target vessel. The distance and
time for the closest point of approach both in relative and
actual scales in a close ship encounter situation can be
calculated to quantify the respective collision risk.
D. Head-on and Overtaking Situations
Vessels can have various navigation limitations and that
should be considered during collision avoidance maneuvers.
Ocean-going vessels can have speed limitations due to their
engine-power configurations. A vessel with a large
displacement also has a large moment of inertia, making it

time consuming to alter its speed. This applies for both
acceleration and deceleration situations. Furthermore, vessels
under slow speed conditions can lose their maneuverability.
This is one of the reasons the COLREGs has highlighted safe
speed to avoid collision or near-miss situations (COLREGs,
1972 - Rule 6 Safe speed). These speed conditions can also
relate to vessel maneuverability with special reference to
stopping distance and turning ability. Since the turning ability
of a vessel relates to its speed, adequate ship speed should be
maintained to preserve vessel maneuverability. One should
note that the vessel with inadequate maneuverability can result
in possible collision or near miss situations with COLREGs
violations.
Fig. 2. Mathematical framework for collision risk estimatio
This section discusses the respective collision avoidance
actions in accordance with the COLREGs for the general
collision risk region (see Figure 2). In general, three ship
encounter situations are illustrated in the COLREGs:
overtaking, head-on and crossing. One should note that these
situations can represent all possible ship encounters.
Even though the COLREGs does not specify the respective
side one vessel should overtake another (COLREGs, 1972
Rule 13 Overtaking), the outcome of this study suggests that
the overtaking vessel should pass on the port side of the
overtaken vessel (with enough time and distance). That
outcome is influenced by the head-on situation, where the
vessels should pass each other port-to-port (COLREGs, 1972
Rule 14 Head-on situation). Rule 13 differs from Rule 14 and
15 as the rule applies even when there is not a risk of collision
– it is enough to be overtaking the other vessel, coming up
with the vessel ahead. The nature of the fairway and
navigational waters in combination with the traffic situation
may make it necessary to pass on the starboard side, but in this
study, the passing side for an overtaking vessel should be on
the port side. Passing on the port side limits the vessel’s
opportunity to make alteration of course to starboard for other
vessels. On the other hand, passing on starboard side limits the
overtaken vessel to make course alterations to starboard.
Therefore, a consistence between these two similar ship
navigation situations should be established by adopting the
port passing concept. There is not a definite angle defining the
difference between a head-on situation and a crossing
situation, and this has resulted in several collisions. Rule 14 do
state that if there is doubt about the situation, it shall be
reacted as if the situation is a head-on situation. The relative
motion of one vessel with respect to the other vessel has
considerable similarities in both situations. Therefore, an
inconsistence, i.e. a possible regulatory failure situation,
between these two similar ship navigation situations can
avoided by the port-to-port approach, specially under the
digital helmsman.
E. Crossing Situations
The most complex collision avoidance actions in ship
navigation can be executed during crossing situations
(COLREGs, 1972 Rule 15 Crossing situation). Such a
situation with additional navigational details is presented in
Figure 3 and the COLREGs provide adequate navigation
guidelines to navigate in such vessel encounter situations. It is
regulated that the vessels coming from the starboard have the
priority to navigating in ship encounter situations. The ships
have a higher and lower priority in an encounter situation are
called the stand-on and give-way vessels, respectively (see
Figure 2). Therefore, the give-way vessel (COLREGs, 1972
Rule 16) should take early collision avoidance actions to avoid
a possible collision or near-miss situation and the stand-on
vessel should keep its course and speed (COLREGs, 1972
Rule 17). A succession of small course and speed changes by
both vessels should be avoided, since that may result in a
possible collision or near-miss situation.
One should note that any change in course or speed
conditions by the stand-on vessel at the general collision risk
region can result in a possible violation of the COLREGs,
(COLREGS, 1972 Rule 8 Action to avoid collision and Rule
17 Action by stand-on vessel). Furthermore, that can confuse
the give-way vessel. However, if the give-way vessel is not
taking any collision avoidance actions, then the stand-on
vessel should take appropriate actions to avoid a possible
collision or near-miss situation. It is expected that such
collision avoidance actions by the stand-on vessel should be
taken in the critical collision risk region. Therefore, vessel
encounter situations in the critical collision risk region should
take emergency measures to avoid possible collision or near
miss situations. The COLREGs do not provide adequate
details on possible collision avoidance actions specially for
these situations, therefore human navigator knowledge and
experiences should be taken to avoid possible collision or near
miss situations. Since the COLREGs require vessels to take
any measures to avoid possible collision or near miss
situations even when the give-way vessel is violating the

COLREGs, the stand-on vessel should prepare for crash stop
type of maneuvers to avoid possible collision or near miss
situations in this region.
IV. CRITICAL COLLISION RISK SITUATIONS
A. Ship Close Encounters
Four critical risk situations, where the own vessel is having
close encounter situations with the target vessel, are
considered (see Figure 3, 4, 5, and 6) to identify possible
regulatory failure situations. It is also assumed that the target
vessel has not honored the COLREGs in these situations and
that can increase the complexity in the respective ship
encounter situation. Furthermore, the respective collision risk
in a ship encounter situation can only be preserved by such a
assumption. i.e., if both vessels take early collision avoidance
actions, then the collision risk can be eliminated, quickly,
therefore, none of the vessels should take any further actions.
Hence, the collision avoidance framework cannot be
evaluated, properly. The worst-case scenarios with higher
collision risk in ship encounters are considered in this study to
identify possible regulatory failure situations due to the same
reasons.
Figure 3 represents a situation in the critical collision risk
region, where the target vessel as the give-way vessel is
heading towards a trajectory intersection point, i.e., a clear
COLREGs violation with crossing from the port. Hence, the
own vessel as the stand-on vessel should take necessary
actions to avoid a possible collision or near miss situation.
One should note that the green arrows in this figure represent
the collision avoidance actions that should have been taken by
the vessels in accordance with the COLREGs and the red
arrows represent the collision avoidance actions that have
been taken by the vessels due to the criticality of the situation
i.e. can be a COLREGs violation. Therefore, the own vessel
as the stand-on vessel has taken a crash stop type maneuver to
make a trajectory that is parallel to the target vessel trajectory.
Since no intersection between these trajectories, those actions
have avoided a possible collision or near miss situation.
Figure 4 represents a situation, where the target vessel as
the give-way vessel is heading towards a trajectory
intersection point, a clear COLREGs violation with crossing
from the port/overtaking. Hence, the own vessel as the stand-
on vessel has taken a crash stop type maneuver to make a
trajectory that is parallel to the target vessel trajectory. Since
no intersection between these trajectories, those actions have
avoided a possible collision or near miss situation.
Figure 5 represents a situation, where the target and own
vessels are in a head-on situation with a trajectory intersection
point, a clear COLREGs violation. The owe vessel has taken
necessary actions, i.e. turn to the starboard, to avoid a possible
collision or near-miss situation. Since no intersection between
these trajectories, those actions have avoided a possible
collision or near miss situation. Therefore, executing those
collision avoidance actions under the proposed framework
with the digital helmsman can also improve the navigation
safety.
Figure 6 represents a critical collision risk situation in ship
navigation. This situation has identified by executing various
collision avoidance simulations under the respective collision
avoidance framework. The outcome shows that this is a
possible regulatory failure situation. This figure represents a
situation, where the target and own vessels are heading
towards the trajectory intersection point, a clear COLREGs
violation with crossing from the port/head-on. The main issue
in this situation is that the decision support feature had
difficulties to identify this either as a crossing or head-on
situation. Even the bearing between the own and target vessels
may not be able to clarify this classification. Therefore, the
validity and applicability of the respective stand-on and give-
way vessel concepts can also be challenged. Any collision
avoidance actions taken by the stand-on vessel can be either
be COLREGs violations or regulatory failures.
The own vessel can take two possible crash stop type
maneuvers in this situation, i.e. turn to the starboard or port.
Even though the own vessel has enough navigation space, the
port turn can be a violation of the COLREGs. If the target
vessel decides to turn starboard as the give-way vessel, then
that may result in a possible collision or near-miss situation
with a trajectory intersection point. On the other hand, the
starboard turn can also intercept the navigation trajectory of
the target vessel. That may also result in a possible collision or
near-miss situation. One should note that it would be difficult
to make a crash stop maneuver from the starboard due to the
angle between the respective trajectories. The outcome shows
that the own vessel can stop in front of the target vessel with
lost maneuverability and that can also be a possible collision
or near miss situation. Therefore, executing these collision
avoidance actions under the digital helmsman can challenge
the navigation safety in the respective vessels in this situation.
On the other hand, this situation can also happen in the general
collision risk region. Therefore, the required measures to
avoid such a situation should be taken.
V. CONCLUSIONS
Several possible regulatory failure situations are evaluated
in a simulated environment under the proposed collision
avoidance framework in this study. One should note that these
simulations may not be able to capture all possible ship
collision situations, therefore additional approaches yet to be
investigated. However, this evaluation is focused in the critical
collision risk region, where close ship encounter situations can
be observed. One critical situation with several possible
collision or near miss conditions under the digital helmsman is
identified in this study. This situation is categorized as the
head-on slightly to port situation and that has been further
investigated to identify the respective reasons. It is noted that
the critical collision avoidance actions that can be taken by the
own vessel even as the stand-on vessel are limited in these
situations and the outcome either be a COLREGs violation or
a regulatory failure. Hence, the definitions of head-on and
crossing situations and stand-on and give-way vessels are
challenged in this situation even under human navigators.

Fig. 3. A crossing from the port situation.
Fig. 4. A crossing from the port/overtaking situation
Fig. 5. A head-on situation
Fig. 6. A crossing from the port/head-on situation

This situation can occur as an intermediate state of ship
encounter situations, i.e. in both the general collision risk
region and critical collision risk region. Such situations cannot
be ignored as an improbable or isolated situation due to these
reasons. Furthermore, the complexity in this critical collision
situation can be increased in several directions and that are
discussed in the following section.
These close ship encounters are further simplified by
assuming that the vessels are moving in straight line
trajectories and their course-speed vectors and headings are in
the same direction. These assumptions may not hold in
realistic ocean-going situations, therefore the following
outcomes are expected:
• If the course-speed and heading vectors in these
vessels are having significant differences; therefore
these vessels can be moving in parabolic type
maneuvers; then that can introduce additional
complexities into the collision risk estimation.
• If this situation is associated with more than two
vessels, various collision avoidance decisions can
overlap and even cancel each other in some
situations.
• If any constrains in the vessel maneuverability can
introduce additional difficulties in executing collision
avoidance actions.
These outcomes can further degrade the collision
avoidance actions that are taken by humans in critical collision
situations. Furthermore, the respective collision avoidance
actions, by the digital helmsman can also be challenged in
such a situation, since the respective decisions are based on a
set of navigation rules and regulations, i.e. the COLREGs.
Therefore, adequate solutions to overcome such critical
collision situations within the COLREGs should be
investigated to make autonomous shipping a reality.
ACKNOWLEDGMENT
This work has been conducted under an internal funded
project of UiT The Arctic University of Norway - that supports
towards autonomous ship navigation in the arctic region. This
work is also supported by the MARKOM2020 project, a
development project for maritime competence established by
the Norwegian Ministry of Education and Research in
cooperation with the Norwegian Ministry of Trade and
Industry
REFERENCES
[1] O. J. Rodseth, L. P. Perera, and B. Mo,"Big data in shipping -
Challenges and opportunities," In Proceedings of the 15th International
Conference on Computer Applications and Information Technology in
the Maritime Industries (COMPIT 2016), Lecce, Italy, May, 2016, pp.
361-373.
[2] L.P. Perera and B. Murray, "Situation Awareness of Autonomous Ship
Navigation in a Mixed Environment under Advanced Ship Predictor," In
Proceedings of the 38th International Conference on Ocean, Offshore
and Arctic Engineering (OMAE 2019), Glasgow, Scotland, UK, June,
2019 (OMAE2019-95571).
[3] L. P. Perera,"Autonomous Ship Navigation under Deep Learning and
the challenges in COLREGs," In Proceedings of the 37th International
Conference on Ocean, Offshore and Arctic Engineering (OMAE 2018),
Madrid, Spain, June, 2018 (OMAE2018-77672).
[4] A. Fridman, et al., MIT Autonomous Vehicle Technology Study: Large-
Scale Deep Learning Based Analysis of Driver Behavior and Interaction
with Automation lessArXiv2017.
[5] L.P. Perera, J.P. Carvalho and C. Guedes Soares, "Intelligent ocean
navigation & Fuzzy-Bayesian decision-action formulation," IEEE
Journal of Oceanic Engineering, vol 37, no 2, 2012, pp 204-219.
[6] M. L. Smith, “Sensors, appliance control and fuzzy logic,” IEEE
Transactions on Industrial Applications, vol. 30, no. 2, pp. 305–310,
1994.
[7] T. Ozen, J. M. Garibaldi, and S. Musikasuwan, “Modeling the variation
in human decision making,” in IEEE Annual Meeting of the Fuzzy
Information Processing - NAFIPS 04, vol. 2, June 2004, pp. 617–622.
[8] L.P. Perera, V. Ferrari, F.P. Santos, M.A. Hinostroza, and C. Guedes
Soares, "Experimental Evaluations on Ship Autonomous Navigation &
Collision Avoidance by Intelligent Guidance," IEEE Journal of Oceanic
Engineering, vol. 40, no. 2, 2015, pp 374-387.
[9] IMO, COLREGS - International Regulations for Preventing Collisions
at Sea, 1972.
[10] L.P. Perera, J. P. Carvalho and C. Guedes Soares, "Fuzzy-logic based
decision making system for collision avoidance of ocean navigation
under critical collision conditions," Journal of Marine Science and
Technology, vol 16, 2011, pp 84-99.
[11] L.P. Perera, and C. Guedes Soares, "Collision Risk Detection and
Quantification in Ship Navigation with Integrated Bridge Systems,"
Journal of Ocean Engineering, vol. 109, 2015, pp. 344-354.
[12] B. Murray and L.P. Perera, "AIS-based Probabilistic Vessel Trajectory
Prediction for Collision Avoidance in Future Vessels," In Proceedings of
the 38th International Conference on Ocean, Offshore and Arctic
Engineering (OMAE 2019), Glasgow, Scotland, UK, June, 2019
(OMAE2019-95963).
[13] L.P. Perera, J.P. Carvalho and C. Guedes Soares, "Solutions to the
Failures and Limitations of Mamdani Fuzzy Inference in Ship
Navigation," IEEE Transactions on Vehicular Technology, vol. 63, no.
4, 2014, pp 1539-1554.
[14] L.P. Perera and B. Murray, "Situation Awareness of Autonomous Ship
Navigation in a Mixed Environment under Advanced Ship Predictor," In
Proceedings of the 38th International Conference on Ocean, Offshore
and Arctic Engineering (OMAE 2019), Glasgow, Scotland, UK, June,
2019 (OMAE2019-95571).