Ocean Engineering 235 (2021) 109380

Contents lists available at ScienceDirect

Ocean Engineering
journal homepage: www.elsevier.com/locate/oceaneng

Collision-avoidance navigation systems for Maritime Autonomous Surface

Ships: A state of the art survey
Xinyu Zhang a, *, Chengbo Wang a, ***, Lingling Jiang b, **, Lanxuan An a, Rui Yang c
a
Navigation College, Dalian Maritime University, Dalian, 116026, China
b
College of Environmental Sciences and Engineering, Dalian Maritime University, Dalian, 116026, China
c
Shipping Technology Research Center, China Waterborne Transport Research Institute, Beijing, 100088, China

A R T I C L E I N F O A B S T R A C T

Keywords: The rapid development of artificial intelligence significantly promotes collision-avoidance navigation of mari­
Collision avoidance time autonomous surface ships (MASS), which in turn provides prominent services in maritime environments and
Autonomous navigation systems enlarges the opportunity for coordinated and interconnected operations. Clearly, full autonomy of the collision-
Cognitive navigation
avoidance navigation for the MASS in complex environments still faces huge challenges and highly requires
e-navigation
Maritime autonomous surface ships
persistent innovations. First, we survey relevant guidance of the International Maritime Organization (IMO) and
industry code of each country on MASS. Then, major advances in MASS industry R&D, and collision-avoidance
navigation technologies, are thoroughly overviewed, from academic to industrial sides. Moreover, compositions
of collision-avoidance navigation, brain-inspired cognitive navigation, and e-navigation technologies are
analyzed to clarify the mechanism and principles efficiently systematically in typical maritime environments,
whereby trends in maritime collision-avoidance navigation systems are highlighted. Finally, considering a
general study of existing collision avoidance and action planning technologies, it is pointed out that collision-free
navigation would significantly benefit the integration of MASS autonomy in various maritime scenarios.

1. Introduction accidents caused by human errors or faults, such as the negligent

lookout of the on-duty driver, are common. Autonomous navigation
Research on navigation safety and shipping safety has a long history. effectively replaces human pilots in ship maneuvering and cargo trans­
For decades, one of the most popular ideas in ocean and maritime en­ portation, which greatly reduces the probability of human-induced
gineering research is that how to design more intelligent and safe marine accidents. At the present stage, autonomous technology is
collision-avoidance navigation systems. Maritime safety faces new limited to applications such as unmanned surface vehicles and under­
challenges when the size and the number of ships is increasing. Based on water robots, but unmanned transport cargo ships cannot yet achieve
reports received from the national accident investigation bodies of the fully autonomous navigation. Collision-avoidance navigation systems
EU, over the period 2014–2019, almost half the marine accidents were use a variety of technologies and advanced research. In recent years, the
navigational in nature, including contact, loss of control, collision, and research hotspots of many scholars and experts on autonomous navi­
grounding stranding (EMSA, 2020). Fig. 1 shows causes of accidents to gation decision and planning are roughly divided into path planning,
ships. On the other hand, marine accidents can lead to greater air obstacle avoidance planning, trajectory planning and behavioral
pollution, water pollution by cargo and bunkers. decision-making. Compared with path planning, collision avoidance and
The history of shipping development is the continuous improvement trajectory planning, behavioral decision-making considers time series
of navigation safety and transportation benefits. The problem of ship and space constraints more. Behavioral decision-making systems are
navigation safety has always been a hot issue in the field of marine used to replace crew. Obstacle avoidance and approaching target ports
transportation engineering, and it is also one of the main aspects that are optimized goals. The behavioral decision-making is imitating the
drive the growth of MASS and their technical needs. Every year, marine human crew’s thinking activity or process of ship maneuvering. In each

* Corresponding author.
** Corresponding author.
*** Corresponding author.
E-mail addresses: [email protected] (X. Zhang), [email protected] (C. Wang), [email protected] (L. Jiang).

https://doi.org/10.1016/j.oceaneng.2021.109380
Received 10 January 2021; Received in revised form 24 March 2021; Accepted 19 June 2021
Available online 28 June 2021
0029-8018/© 2021 Elsevier Ltd. All rights reserved.
X. Zhang et al. Ocean Engineering 235 (2021) 109380

1. Ship with autonomy processes and decision support;

2. Remotely controlled ship with seafarers on board;
3. Remotely controlled ship without seafarers on board;
4. Fully autonomous ship.

For the real ship test developed by the autonomous ship, MSC takes
note of the proposal for the development of test criteria submitted by
relevant countries and considers it a necessary work. This criterion
should prevent the safety of autonomous ships and protect the envi­
ronment from pollution.
In June 2019, MSC held the 101th session and approved interim
guidelines for MASS trails, including the risk to safety, security and
protection of the marine environment, listing the compliance mandatory
Fig. 1. Causes of accidents to ships. instruments, manning and qualifications of personnel involved in MASS
trials, human element (including monitoring infrastructure and human-
collision avoidance or transportation process, the optimal navigation system interface), infrastructure for safe conduct of trials, trial aware­
strategy is determined from many schemes in accordance with its own ness, communications and data exchange, reporting requirements and
behavioral constraints. information sharing, scope and objective for each individual trial and
With the development of a new generation of artificial intelligence cyber risk management (MSC.1/Circ.1604, 2019). MSC 102nd session
technology, the autonomy system has been widely adopted in the field of was held in January 2020, just for a status report of the progress of the
driverless vehicle, underwater vehicle, and unmanned aerial vehicle. regulatory scoping exercise (MSC 102/5, 2020).
This work analyses current challenges and opportunities for collision Fig. 2 shows the selected proposals of each session of MSC for MASS
avoidance and navigation planning for maritime autonomous surface on a timeline.
ships. The following contributions are provided:
2.1.2. Industry code of MASS
1. Summary the guidance document of IMO and industry code of each Under the guidance and leadership of IMO, many countries and in­
country on MASS. stitutions in the world have gradually developed their own industry
2. Review of state-of-the-art MASS industry research & development guidelines and codes, which include the scope and function of promot­
and advances in collision-avoidance navigation technology. ing the research and development of MASS according to the specific
3. Characterization of applications for maritime collision-avoidance background and market demand of their own countries. Among them,
navigation systems. the United Kingdom, the United States, China, and Japan have issued
4. Overview of existing and future collision-avoidance navigation several industry codes related to autonomous navigation.
technologies. The UK has established the UK Maritime Autonomous Systems
Regulatory Working Group (MASRWG) to develop a regulatory frame­
2. State-of-the-art autonomous ship and collision-avoidance work for MASS and industry-led behaviors and practices for the safe
navigation technology operation of MASS. The criteria, as shown in Table 1, are the classifi­
cation of MASS in the United Kingdom. The second version of the code of
2.1. Advances in the MASS conduct for the MASS industry was released in November 2018, the
third version of the code of conduct for the MASS industry was released
There is no doubt that the development and application of autono­ in November 2019, which include autonomous ships certification,
mous ship will greatly improve the safety of maritime cargo trans­ registration of MASS, standards for MASS demonstration and testing
portation and reduce the pollution of marine environment caused by areas in British waters, training, skills and qualifications and so on.
marine accidents. Section 2.1 will review and analyze the guidance of Version 4 was released in November 2020, prepared in two parts,
IMO on MASS, the code of different countries on MASS, and the research including Industry Conduct Principles and Code of Practice (Maritime,
and development of MASS industry. 2018, 2019, 2020).
For the United States, ADVISORY ON AUTONOMOUS FUNCTION­
2.1.1. The guidance of IMO on MASS ALITY is issued by American Bureau of Shipping (ABS) in 2020. This
Since Maritime Safety Committee (MSC) agrees to include the issue Advisory mainly analyzes the autonomy development background, in­
of the autonomous ship on its agenda in January 2017 (MSC 98/20/2, dustrial demand and progress. The whole autonomous navigation is
2017), MSC has been committed to guiding various countries and in­ divided into smart, semi-autonomy and autonomy. The definition of
stitutions to discuss the impact and benefits of autonomous navigation each level and the role of the human in the operational decision loop are
technology, research and development, testing, etc. Meanwhile, the elaborated. And the functions and concepts of Remote Control (RC) and
scoping exercise is seen as a starting point. From 2017 to 2020, MSC has Shore Operations Centers (SOC) are described in detail (ABS, 2020). The
made great contributions to the development and application of tech­ autonomy and operational decision loop are shown in Table 2.
nology of maritime autonomous surface ships, and promoted the In 2015, China Classification Society issued the world’s first intelli­
development of technology to practical application. In the 99th session, gent ship code (2015), which proposed the concept, development path
MSC has signed the framework for regulatory scoping exercise as and main structure of intelligent ship (CCS, 2015). In 2018, the
meeting progress, preliminary definitions of MASS, the degrees of au­ GUIDELINES FOR AUTONOMOUS CARGO SHIP was released, including
tonomy included, as well as a methodology for conducting the regula­ specific requirements, objectives, functions, equipment performance,
tory scoping exercise and plan of session (MSC 99/WP.9, 2018). inspection and experiment of key technologies for autonomous naviga­
In December 2018, MSC held its 100th session, which is a milestone tion (CCS, 2018). The scope involves scene perception, navigation
for maritime autonomous surface ships. The framework and methodol­ control, marine engine installation, mooring and anchoring, electrical
ogy for regulatory scoping exercise on MASS is approved. And the de­ equipment, communication and signal equipment, hull structure and
gree of autonomy identified for the purpose of the scoping exercise are safety, fire protection, environmental protection, ship security, remote
(MSC 100/WP.8, 2018): control center, network security, inspection, and certification, etc. In
2019, according to the input of application experience, industry

2
X. Zhang et al. Ocean Engineering 235 (2021) 109380

Fig. 2. Timeline of each session of MSC for MASS, only selected proposals are indicated.

2.1.3. Advances in the MASS industry

Table 1
Autonomy technology of MASS is the integration of numerous
Class of maritime autonomous surface ship in British.
technologies of intelligent ships, including autonomous navigation
Class of Characteristic Notes technology (navigation situation awareness, navigation behavioral
MASS
decision-making, motion control), intelligent engine room operation
Ultra- Length overall <7m and *Derived from MCA and maintenance, ship-shore/ship-ship communication, intelligent hull,
light maximum speed <4 kts High-Speed Craft Code (https://www. integrated trials, and others. With the development of artificial intelli­
Light Length overall ≥ 7m to gov.uk/government/uploads/system
<12m and maximum speed /uploads/at
gence and communication technology, the level of ship automation has
<7 kts tachment_data/file/292155/hsc_ been gradually enhanced. In the past several decades, some countries
Small Length overall ≥12m to 2000_rev06-09_full-compall. pdf) represented by Norway and the United States have played a constructive
<24m where ∇ = role in the field of research and development of unmanned surface ships
Large Length ≥24m (and 100 GT) moulded displacement, in m3,
and autonomous surface cargo ships, all over the words.
High- Operating speed V is not less of the craft corresponding to the
Speed* than V = 7.19 ∇1/6 knots design waterline. In recent years, the Norwegian Fraunhofer CML, as the first organi­
zation to research the demonstration related to the unmanned cargo
ship, completed the project maritime unmanned navigation through
intelligence in network (MUNIN) from 2012 to 2015 to verify the
Table 2
concept of the autonomous ship, which is defined as a ship mainly
Autonomy and operational decision loop.
guided by the autonomous decision support system and controlled by
System Autonomy Integration and Application to Decision Loop
the remote-control operator of the shore control center. The communi­
Levels
Monitoring Analysis Decision Action cation architecture solutions for the autonomous ship bridge, the
1 Smart Machine Machine Human Human autonomous machine room, the shore operation center, and the opera­
2 Semi- Machine Machine Human/ Human/ tors connecting the ship to shore have been developed and verified
Autonomous Machine Machine (Munin, 2016). Sponsorship by the Norwegian Research Council, the
3 Autonomous Machine Machine Machine Machine
University of science and technology of Norway started the autonomous

feedback, research results, international discussion and other aspects,

the code was upgraded and revised. The new version of the Code for
intelligent ships has come into effect on March 1, 2020 (CCS, 2020).
With technologies such as sensing, artificial intelligence, and
internet of things having made rapid progress in various fields, the
ClassNK of Japan issued the “Guidelines for Concept Design of Auto­
mated Operation/Autonomous Operation of Ships (Provisional
Version)” and “Guidelines for Digital Smart Ships” in 2020. The
“Guidelines for Concept Design of Automated Operation/Autonomous
Operation of Ships” provide their design development, installation and
operation of automated operation systems or remote operation systems
(ClassNK, 2020a). The “Guidelines for Digital Smart Ships” specifies the
requirements for the award of class notations to ships equipped with
systems such as energy efficiency analysis, hull or machine monitoring,
sloshing detection and prediction, onboard data processing and data
transmission to shore, route/speed optimization and remote monitor­
ing/operation (ClassNK, 2020b).

Fig. 3. The architecture of AMOS.

3
X. Zhang et al. Ocean Engineering 235 (2021) 109380

marine operations and systems (AMOS) project research in 2013. The shallow draft, fast maneuvering, and strong endurance. For intelligence,
architecture of AMOS is shown in Fig. 3. It is expected to complete the surveillance, reconnaissance/force protection, anti-mine operations,
research on autonomous ships and robot systems in 2023, and develop precision strike/anti-ship operations and anti-submarine operations.
the structure and operation of safer, smarter, and more Fig. 5 is USA “SPARTAN”. Fig. 6 is Israel “Sea Knight”. Fig. 7 is France
environmentally-friendly ships and offshore intelligent platforms REMORINA and Fig. 8 is Italy SAND, respectively.
(NTNU AMOS, 2017). In October 2016, the Norwegian forum for Fig. 9 is ESM30. Fig. 10 is TIANXING. Fig. 11 is HUSTER-68. Fig. 12
autonomous ships (NFAS) was established to release information about is JHAI No.1. Fig. 13 is LANXINHAO, and Fig. 14 is Zhihai No.1,
international conferences and reports related to MASS, and in October respectively. As shown in Fig. 14, the center for Intelligent Maritime
2017, under the organization of NFAS and SINTEF ocean, Norway, Vehicle of Dalian maritime university of China then developed a plat­
China, the United States and other countries established the interna­ form for “zhihai-1". The research platform consists of USV, which in­
tional network for autonomous ships (INAS), marking the research of cludes an inflatable buoy and UAV landing platform, equipment box and
MASS has been promoted to the national level, even to the international support erected on the buoy. The hull adopts two inflatable buoys to
level (NFAS, 2019; INAS, 2019). The SINTEF ocean laboratory in form a double hull structure, which enhances the navigation stability
Trondheim, Norway, and Kongsberg, a technology company, jointly and the convenience of maintenance: the equipment boxes on both sides
developed autonomous ship named Yara Birkeland, the first electric contain batteries to provide energy for the propulsion system. Electronic
propulsion Unmanned Container Ship in the world. As shown in Fig. 4, devices such as attitude sensors, GPS receivers, microprocessors and
the ship has a length of 70m, a width of 15m, and can carry 100-150teu. other electronic devices are placed in the equipment boxes in the mid­
It has been tested in the water pool of SINTEF since September 29, 2017. dle, which can obtain GPS information, pitch angle, roll angle and
It can use its own installed GPS, radar, and camera to avoid other ships heading angle, and output acceleration information in the attachment
in the channel, and realize auto-docking when arriving at the terminal setting system, which greatly improves the measurement accuracy. The
point. In 2018, the autonomous navigation test from the port of Herøya USV adopts wireless communication components, which can directly
in Norway to the port of Brevik has been realized for the first time. By communicate with external mobile terminals and ground control sta­
the end of November 2020, the ship will be handed over from the tions (Wang et al., 2019a). The wireless communication components
Norwegian shipyard to Yara. After delivery, the vessel will undergo communicate data through ZigBee, and implements effective data
container loading and stability tests before sailing to port and trial sea sending and receiving according to Inter-Integrated Circuit (IIC)
area for further preparation for autonomous navigation (Yara Birkeland, communication protocol.
2020). From the aspect of autonomous surface cargo ship, leading countries
Rolls-Royce of the UK and Stena Line AB of the Swedish ferry com­ and research institutions are the Joint Industry Project (JIP) Autono­
pany will jointly develop the first intelligent ship sensing system. At the mous Shipping of Netherlands, autonomous vessel company Masterly of
“Seminar of Un-manned Ship Technology” held in 2016, the " Devel­ Norway and action planning system project of Nippon Yusen Kaisha Line
opment Plan of Advanced Unmanned Ship Application” was launched. It of Japan. The autonomous shipping project conducts the first autono­
is expected that the use of remote support and specific function opera­ mous ship navigation trials in the North Sea, sets the planned route, and
tions will gradually reduce the appointment of crew members in 2020; avoids obstacles with the experience and lessons. By testing the sce­
remote control of offshore MASS by 2025; remote control of ocean MASS narios on the North Sea, the decision-making process of an autonomous
by 2030; autonomous ocean-going MASS by 2035 (AAWA, 2017). system was able to show in ensuring safe sailing and avoiding collisions
Other countries have also achieved excellent research results in un­ with other vessels. Through the test video, we found that the system can
manned surface vessel (USV). They have their own independently safely avoid obstacles in simple scenarios, such as head-on scenario. It
developed USV, but most of them are used in the military field, such as was concluded that further development of autonomous systems is
the United States, Israel, France, Italy, Japan, Belarus and China, the needed, to cope with complex marine traffic situations in a more effi­
parameters of USV developed by these countries are shown in Table 3 cient way (Autonomous shipping, 2019). NYK has conducted the world’s
(Richter, 2006; Lin and Zhang, 2018; Wan, 2014; Kumar and Kurmi, first MASS trial performed in accordance with the IMO’s Interim
2018). Guidelines for MASS trials as the company begins tests to realize its
According to statistics and comparisons, USVs are used for marine target of manned autonomous ships for safer operations and a reduction
environmental monitoring or military reconnaissance and strike in most in crew workload. During the trial, the SSR’s performance in actual sea
countries. The military use is represented by the “Spartan Scout” in the conditions was monitored as it collected information on environmental
United States. It is the earliest and most versatile. The ship is light, conditions around the ship from existing navigation instruments,
calculated real-time collision risk, automatically determined optimal
routes and speeds that were safe and economical, and then automati­
cally navigated the ship. Through the test video, we found that the ship
avoided the obstacle ship by turning right in the head-on scenario. This
behavior complied with the COLREGS. NYK will analyze the data and
continue to develop SSR into a more advanced navigation support sys­
tem by adjusting the difference between the best course obtained by the
program and the best course determined by professional officer’s judg­
ment. This trial was a big step toward realizing NYK’s goal of manned
autonomous ships (NYK, 2019). Comprising 22 domestic Japanese
companies, such as Japan Marine Science Inc. (project leader), MTI Co.,
Ltd., IKOUS Corporation, NYK and so on, the Designing the Future of
Full Autonomous Ship Project (DFFAS) aims to realize the trial of
autonomous and unmanned navigation. The project plans to carry out a
long-distance demonstration trial within 2021 in congested waters using
a domestic coastal containership, efforts toward practical crewless
maritime autonomous surface ships by 2025 (NYK, 2020; Weathernews,
2020). The schedule for the implementation of autonomous ships in
DFFAS project is shown in Fig. 15. Kongsberg maritime and Masserly
Fig. 4. Yara Birkeland. will equip and operate two zero-emission autonomous vessels for the

4
X. Zhang et al. Ocean Engineering 235 (2021) 109380

Table 3
Various types of USV parameter tables in other countries (Incomplete statistics).
Country Ship name Manufacturing/ Size Endurance Max Research purpose and major Characteristic
application time (m) speed achievements
(kn)

USA Spartan Scout 2001 7/11 8h/28h 50 1) Port surveillance; 2) Force Light hull, shallow draft, fast maneuvering,
protection strong endurance, and can move in very
shallow waters near shore.
Israel Sea knight 2017 11 12h 40.5 1) Surveillance and It is large and capable of high-speed
reconnaissance navigation, and is more stable in heavy
winds and waves. It can sail out of the sea
500 km away from the shore.
France USV REMORINA 2017 9 – – 1) Surveillance and Autonomous decision-making and
reconnaissance; 2) Force automatic obstacle avoidance.
protection
Italy SANDUSV 2018 16 48h 36 1) Search and rescue; 2) Self-righting, and can work in harsh
Environmental monitoring conditions.
Japan Autonomous Ocean 2016 3 >24h – 1) Environmental survey; 2) Consists of two parts: pontoon and
Observation Device (AOV) Data collection planning boat, connected by cables.
Belarus Multifunctional USV 2013 6 120h (313 54 – Stealth with satellite navigation system,
nm) radar and camera.
China ESM30 USV ( 2015 1.15 24h 3.9 1) Environmental sampling Intelligent control terminal and real-time
Yunzhou-tech., 2020) and survey; 2) Data collection remote communication, automatic
obstacle avoidance.
C-38 (Smart Ocean, 2020) 2020 3.8 >7h 2 Monitoring and sampling Multi-point selection and vertical
hierarchical water sample collection.
Skyline One (HEU, 2018) 2017 12.2 1000 km 50 – High speed, long range, independent
monitoring.
M3U (Ma and Sheng, 2017 5 – – 1) Search and rescue Intelligent, unmanned and three-
2018) dimensional efficient search and rescue.
HUSTER-68 (HUST, 2018) 2018 6.8 120 nm 30 Offshore patrol and Sensor equipment includes lidar, binocular
supervision camera, laser rangefinder, optical fiber
combined inertial navigation, etc.
Jinghai-1 (Shanghai 2013 6.28 >130 nm 10 Intelligent measurement Independent and remote-control dual
University, 2018) mode operation.
DMU: Blue signal (Blue 2012/2018 7.02/ – 35 1) Test platform; 2) It has three control modes: full
signal, 2020); Zhihai-1 ( 2 Navigation and control autonomous, semi-autonomous and
Wang and Gao, 2018) systems test; 3) As sea-surface remote-control mode (see Fig. 7).
target system

Fig. 5. USA “SPARTAN”.

Fig. 6. Israel “Sea Knight".

leading Norwegian grocery distributor ASKO. All the technologies

So current state-of-art in autonomous surface cargo ship of MASS
required for unmanned operation are equipped. Meanwhile, Masterly
industry is that it is possible to let a large cargo ship sail autonomously
will ensure ship management and safe operation through its shore-based
over a restricted time and nautical mile, including collision avoidance
remote operation center (Kongsberg, 2020).
decision-making and ship maneuvering, autonomous berthing and
Other research and development projects of autonomous surface
unberthing, to realize remote control and periodicity unmanned
cargo ships have also achieved excellent research results all over the
autonomous navigation. At present, the international research and
world. They have their own autonomous ship and trial sea area, using
development leading projects are all entering the stage of real ship
various algorithms and systems. Some functional tests have been
testing in specific actual sea areas.
implemented, such as smart support navigation, semi-autonomous,
shore based remote control and so on. The detailed comparative infor­
mation of MASS developed by these countries is shown in Table 4.

5
X. Zhang et al. Ocean Engineering 235 (2021) 109380

Fig. 7. France REMORINA.

Fig. 10. TIANXING.

Fig. 8. Italy SAND.

Fig. 11. HUSTER-68.

Fig. 9. ESM30.

2.2. Advances in collision-avoidance navigation technology Fig. 12. JHAI No.1.

Collision-avoidance navigation system plays the role of copilot in the cargo ships depends heavily on intelligent navigation systems with
whole autonomous surface ship system. The problem to be solved is to perception, collision-avoidance decision-making, and control capabil­
determine the obstacle avoidance strategy and collision-avoidance path ities. Thus, the research on the technology of collision-avoidance navi­
through perceiving and learning the maritime safety information of the gation for MASS is mainly divided into perception, collision avoidance,
MASS. Safe and efficient maritime transportation of autonomous surface motion control, and communication.

6
X. Zhang et al. Ocean Engineering 235 (2021) 109380

2.2.1. Perception
For the perception, the current shipboard perceived equipment in­
cludes high-definition cameras (HD camera), shipborne radar, milli­
meter wave radar (MMW radar), ECDIS, lidar, Automatic Identification
System (AIS), etc. (Liu et al., 2016). However, for maritime autonomous
surface ships, intelligent perception technology may have reached the
pilot or higher application stage. The problems to be solved include ship
identification, static obstacles perception, visibility impact, speed
perception, distance perception, viewing angle and cost. Cui et al.
(2019) proposed a novel multi-scale ship detection method based on a
dense attention pyramid network (DAPN) in SAR images, to detect
multi-scale ships in different scenarios with extremely high accuracy.
However, radar is typical sensing equipment used to detect ships and
obstacles, but the radar echo cannot scan the shape and appearance of
the target, which affects the ability of collision-avoidance navigation
decision-making. Thus, in the early stage of research on intelligent
perception, many scholars transferred learn perception technology of
unmanned car, to realized maritime obstacle perception using video
Fig. 13. LANXINHAO.
images and HD cameras. Knébel (2020) designed a monocular
camera-based system, to detect obstacles in open sea scenarios and es­
timate surrounding ship’s distance and bearing. Liu et al. (2019) trained
many ship video datasets based on deep learning framework and
cross-layer jump connection policy, to realize automatically recognizing
and tracking dynamic targets. At present, environmental perception is
undoubtedly one of the first tasks facing the research of maritime
autonomous unmanned systems. Especially under poor visibility con­
ditions such as rain, snow, and fog, it will be very difficult for
collision-avoidance navigation system to achieve accurate and rapid
environmental perception. As for this problem, Wright. (2019) explored
the use of machine learning and artificial intelligence techniques as a
tool to combine multiple sensor equipment with collision-avoidance
navigation system. The complementary advantages of multiple sensors
are used to reduce the impact of environmental conditions on the ability
of perception. Han et al. (2020) put radar, lidar and cameras together to
build a new mixed sensor fusion framework. An object ship detection
algorithm had been applied to the mixed sensor platform, to estimate the
encounter information. Finally, they planned the collision-avoidance
navigation behavior and controlled the ship motion by trained this in­
Fig. 14. Zhihai No.1. formation with the international regulations for preventing collisions at
sea (COLREGs). The identification of small moving targets at sea is an
important issue in ship navigation, especially for MASS, it is necessary to
introduce new means to make up for the lack of radar and AIS in

Fig. 15. Schedule for the implementation of autonomous ships in DFFAS project (Weathernews, 2020).

7
 X. Zhang et al.
Table 4
Advances in the autonomous surface cargo ship industry.
Project Institution Collision- avoidance Ship principal Proximity sensors Timeline and status Autonomy
navigation system particulars level*

Yara Birkeland YARA, KONGSBERG. K-NAV, Length:79.5m; Radar; L2/L3

K-BRIDGE AUTOPILOT, Width mld:14.8m; Lidar;

K-NAV AUTOPILOT.
Draught: full 6m, AIS;
ballast 3m;
Speed: Service Camera;
6knots,
Max 13knots; IR camera.
Cargo capacity:
120TEU;
Dead weight:
8

3200 mt.
The Designing the Japan Marine Science Sherpa System for Real IRIS LEADER: Radar; L2/L3, L4
Future of Full Inc. (project leader), NYK ship (SSR) navigation (Periodicity)
Autonomous Ship Line, MTI Co. Ltd. et al. system.
Project (DFFAS) 22 domestic Japanese
companies.

Length: 199.99m; ECDIS;

Breadth: 34.8m; AIS.
DWT: 20853t;

Ocean Engineering 235 (2021) 109380

Draught: 7.9 m;
Speed recorded
(Max/Average):
18.7/17.9 knots.

Project Institution Collision- avoidance Ship principal Proximity Timeline and status Autonomy
navigation system particulars sensors level*

OVERLORD PROGRAM U.S. Navy Command, Control and LUSV: Radar; L3/L4
& Medium/Large Communications (C3). (Intermittent)
(continued on next page)
 X. Zhang et al.
Table 4 (continued )
Project Institution Collision- avoidance Ship principal Proximity sensors Timeline and status Autonomy
navigation system particulars level*

Unmanned Surface
Vessel Plans

Length: 200–300 feet; E-Optical/

Infrared (EO/
IR);
Full load AIS;
displacements: 1000-
2000t;
MUSV: GPS;
Length: 45–190 feet; IMU.
Displacement of
roughly 500t;
Endurance of 4500 nm
or more at 19 knots
9

transit speed or higher.

Mayflower Autonomous ProMare, IBM and a IBM Visual Insights Length: 15m; GNSS; L3/L4
Ship global consortium of computer vision (Intermittent)
partners. technology, IBM edge
systems, IBM ODM.

Width: 6.2m; Radar;

Ocean Engineering 235 (2021) 109380

Max speed: 10 knots; Lidar;
Weight: 5 tons/4535 SATCOM;
KG;
Equipment capacity: AIS;
0.7 tons/700 KG. Camera;
Weather
station.

Project Institution Collision- avoidance Ship principal Proximity Timeline and status Autonomy
navigation system particulars sensors level*

(continued on next page)

 Table 4 (continued )

X. Zhang et al.
Project Institution Collision- avoidance Ship principal Proximity sensors Timeline and status Autonomy
navigation system particulars level*

ZULU MASS ZULU Associates, See AUTOSHIP Project. Length: 90.0m; Radar; L2/L3

Blue Line Logistics, Anglo Beam mid: 15.0m; AIS;

Belgian Shipping
Company
Draft mid: 5.50m; ECDIS;
Air draft limit: 9.1m; RIS (River
Information
System);
Service Speed: 10.5 GPS.
10

knots (85%MCR);
TEU Capacity: 149+
Key technologies of ship China WTRI (project Manned, remote control, ZHIFEI: GPS; L2/L3, L4
intelligent navigation leader), DMU, WHUT, unmanned autonomous (Periodicity)
and control based on CSIC 704, HEU, BRINAV navigation.
ship shore et al. 21 domestic China
cooperation Project companies.

Ocean Engineering 235 (2021) 109380

L: 117.15m; ECDIS;
MB:17.32m; AIS;
MD: 9.9m; G-compass;
Designed draft: 4.8m; Log;
Speed: 12 knots; VDR;
TEU Capacity: 300. MF/VHF;
Radar;
Lidar;
Camera.
X. Zhang et al. Ocean Engineering 235 (2021) 109380

Fig. 16. Classification of MASS collision avoidance with respect to methods.

detecting small moving targets at sea. Chen et al. (2020) combined a hybrid algorithm based on the combination of improved genetic algo­
modified Generative Adversarial Network (GAN) and a Convolutional rithm and improved artificial fish swarm algorithm to realize the safe
Neural Network (CNN)-based detection approach, to design a novel navigation of the unmanned surface vehicle (USV) in the complex and
hybrid deep learning method. Specifically, they generate sufficient multi-static water surface environment. Taking the safety and economy
informative artificial samples of small ships based on the zero-sum game of ship navigation as the objective function, Zeng et al. (2020) proposed
between a generator and a discriminator for training and learning. a hybrid optimization collision avoidance decision algorithm based on
Particle Swam Optimization-Genetic Algorithm (PSO-GA), COLREGS of
2.2.2. Collision avoidance ship navigation considered. Liu et al. (2016) combined the bacterial
From the 1990s, it is roughly divided into four parts: geometric foraging algorithm and particle swarm optimization to optimize the
method, optimization algorithm and bionic algorithm, virtual vector collision avoidance path in the situation of multi-ship encounters. The
and field theory, artificial intelligence method. Fig. 16 shows the clas­ avoidance route was generated through the avoidance angle and timing,
sification of MASS collision avoidance with respect to methods. which improved the global search and local convergence capabilities of
In the first stage, the autonomous collision avoidance navigation the multi-objective optimization algorithm. In recent years, some
designs the model and research method according to the ship pilot’s scholars have also applied model predictive control algorithms to the
avoidance operation process, completely. That is, geometric methods field of ship collision avoidance trajectory optimization, which has
and mathematical models. The early issues of collision avoidance at sea solved the risk avoidance and path optimization in the complicated
have been constantly discussed and studied through radar and radar navigation situation of multiple ships meeting the COLREGS (Johansen
plots. An anti-collision indicator was discussed in two ships encounter et al., 2016; Xie et al., 2019; Eriksen et al., 2019).
situation by Mitrofanov. (1968), reduction of speed and altering heading However, with the rise of driverless technology research and devel­
considered. The concept of a ship domain was first outlines by Goodwin, opment, many scholars have migrated and applied key technologies of
(1975). Davis. (1980) improved the ship domain model and first unmanned vehicles to the fields of intelligent obstacle avoidance and
mentioned an evasion area, then he used these concepts to ship collision autonomous navigation of ships. For road unmanned driving, it has the
avoidance and running aground. Even though 1972 the International most distinctive structural characteristics. This constitutes the third
Regulations for Preventing Collisions at Sea (COLREGS) were drafted, stage, virtual vector, and field theory, of collision avoidance navigation
but the Rule 8 and Rule 16’s keywords, early, large enough, substantial, for autonomous ship. For the problem of obstacle avoidance navigation
safe distance, are not clear the actual maneuvering for collision avoid­ in complex and dynamic obstacle environments, Lyu and Yin. (2019)
ance. Zhao-Lin, W. (1984) applied geometric analysis, through the added security and COLREGS constraints to the corresponding virtual
construction of the calculation model of the target motion element, the field strength and force, improved the artificial potential field, and
Distance to Closet Point of Approach (DCPA) and the Time to Closet constructed a real-time and deterministic obstacle avoidance path
Point of Approach (TCPA), the latest rudder timing and the quantifica­ planning method. Li and Zheng. (2020) used field theory to abstractly
tion of the implementation plan of collision avoidance decision, and simulate the trend of ship navigation, and built a collision avoidance
initially realized the automatic generation of the collision avoidance model based on geometric derivation combined with virtual space
decision of open water ships. Up to now, these geometric and mathe­ electric field and velocity field, which solved the multi-ship avoidance
matical models are still necessary for the most basic applications in problem. The last two years, velocity obstacle (VO) algorithms and the
collision avoidance problems. generalized velocity obstacle (GVO) algorithm are applied to the
In the second stage, with the development of computational intelli­ research of autonomous ship intelligent obstacle avoidance, and con­
gence, optimization algorithms and bionic algorithms are gradually structed an obstacle avoidance technology more in line with actual
applied to the field of ship obstacle avoidance. Zhao (2019) used a research and development for MASS (Huang and van Gelder, 2018;

11
X. Zhang et al. Ocean Engineering 235 (2021) 109380

Huang and Chen, 2019; Shaobo et al., 2020). 2.2.4. Communication

With the development of artificial intelligence technology, the The “Titanic” incident in 1912 made people realize that the primary
research and development of obstacle avoidance navigation technology purpose of maritime radio communication should be to ensure the safety
for MASS has also entered the fourth stage, autonomous obstacle of life and property at sea. Therefore, the first International Convention
avoidance based on artificial intelligence algorithms, such as iterative for the Safety of Life at Sea (SOLAS) was formulated in 1913. One of the
observation and inference, neural networks and fuzzy logic, deep rein­ important achievements is the formulation of minimum requirements
forcement learning, game theory. Fan and Yan, at al. (2020) proposed an for ship radio stations (Nature Publishing Group, 1913). In the past
advanced methodology for maritime accident prevention century, radio communication technology has been widely used in the
decision-making strategy formulation from human factor perspective field of ship navigation, which provides a technical means for effective
based on Bayesian network (BN) and Technique for Order of Preference information exchange and communication for ship-shore and ship-ship.
by Similarity to Ideal Solution (TOPSIS). It has opened a new mode of Since November 1899, the first time in the history of man-made radio
marine accident research. By fusing a hidden obstacle avoidance logic communication has been realized in the United States, which is mainly
layer with an observable and predictable control layer, Wang et al. based on the manual Morse telegraph (Daley, 1977). In the 1970s, telex,
(2020) constructed a two-layer human-like collision avoidance telephone, fax, and other communication methods were gradually
decision-making process, which improved the success rate of obstacle applied to ship communication. In the 1970s and 1980s, narrowband
avoidance of multi-ship encounter scenarios without a coordination direct printing telegraphy (NBDP) and radiotelephone (RT) technology
center. Ahn et al. (2012) combined the fuzzy inference system and the were applied in the ground communication system, and satellite
expert system to the ship collision avoidance system, in which the communication technology was also occasionally used. However, the
membership functions of DCPA and TCPA were determined, and the Morse signal can carry the traffic is also limited, and the operation cost
neural network was used to evaluate the collision risk, madding up for of large wireless telephone station is rising. By the end of the 1980s,
the deficiency of fuzzy logic. In order to improve the autonomous and satellite services had begun to occupy an increasing share in the
intelligent level of adaptive guidance for MASS, deep learning and ship-to-shore communication market. For these reasons, IMO passed the
reinforcement learning are used more to build autonomous navigation International Convention on Maritime Search and Rescue (SAR) in 1979,
systems, to deal with multi-ship collision avoidance (Wang et al., 2019; and proposed to adopt the latest technology to develop the global
Zhao and Roh, 2019; Zhang et al., 2019; Woo and Kim, 2020). For maritime distress and safety communication system (IMO, 1979).
collisions avoidance of MASS, more human-like intelligence will also be Finally, the global maritime distress and safety system (GMDSS) was
developed, and more scholars will transfer game theory to the field of implemented in 1992, when advanced communication technology was
autonomous navigation. widely used in ship communication. Whether it is a ground communi­
cation system or a satellite communication system, at present, the
2.2.3. Motion control communication services recognized by GMDSS are mainly telex
Motion control systems for maritime surface ships, including ship communication, and there is also the single side-band radiotelephone
steering and closed-loop control, have been an active topic of research (SSB) service of the ground system. These services belong to narrowband
since the first mechanical autopilot was constructed by Elmer Sperry in communication services (Wikipedia, 2021).
1911 (Allensworth, 1999). The autopilot was referred to as the “Metal At present, the information exchanged between the ship and shore is
Mike”, which obtain much of the ship maneuvering behavior of a pilot or not only limited to ship reports and some telegrams for instructions and
a navigator. This device did reduce heading error for various complex reports, but also includes massive data information of images, pictures,
sea states using feedback control and automatic gain adjustments voice, and various ship parameters. At the beginning of this century,
(Roberts, 2008). Later in 1922, three-term control was proposed by Voyage data recorder (VDR) and AIS are required to be installed on
Nicholas Minorsky, through the analysis of a position feedback control ships, and various parameters and information of ships are digitized,
system, that is Proportion Integral Differential (PID) control (Minorsky, which lays the foundation for the transition from experience navigation
1922). In the 1960s, thrusters and propellers were applied to control the to digital navigation and makes it possible to transmit various parame­
horizontal motion of ships, such as surge, sway, and yaw, with three ters of ships. For the communication technology of MASS, GMDSS, AIS,
decoupled PID-controller. This was named and known as dynamic and other means of communication will face new challenges in high
positioning (DP) systems. The successful use of Linear Quadratic bandwidth and low delay.
Gaussian (LQG) controllers in ship autopilots and DP systems, and the
availability of more accurate navigation systems like GPS resulted in a 3. Maritime collision-avoidance navigation systems
growing interest in way-point tracking control systems (Holzhüter,
1997). Since 2000, modern control systems were based on a variety of Maritime autonomous navigation systems of collision avoidance can
design techniques such as sliding mode (Zhang et al., 2000), H infinite increase the safety of life at sea. The system assists the Master or officer
control (Sheng. et al., 2006), PID control (Fang et al., 2010; Fang et al., of watch (OOW) in their analysis of encounter situations by simulta­
2012) and neural networks (Sun et al., 2018), to mention only some. neous plotting of all targets in the declared range, to minimize the risk of
In the past two years, for the study on motion control and trajectory collisions. Meanwhile, the safe course or speed is calculated, according
tracking control of ships, to effectively deal with the extremely strong to the COLREGS, aiming at passing from all targets clearly. Therefore, it
unmodeled dynamics, model uncertainty and unknown external inter­ is recommended to develop performance standards that will assist the
ference of the USV, an intelligent self-structured robust adaptive way­ shipping community in proper analysis, design, testing, and approval of
point track tracking control strategy independent of the model is such system.
proposed by Ning Wang and Hamid Reza Karimi (2020), realized a new This section provides a general perspective on navigation systems of
method of precise track tracking control of the surface ship under the collision avoidance for maritime autonomous surface ships and modules
un-known time-varying complex sea conditions, and then proposed a of autonomous navigation systems that are present in the future.
limited time tracking control strategy of the USV, accurately suppress
and cancel external interference and system uncertainty (Ning and Pan, 3.1. Challenges in collision-avoidance navigation systems in an uncertain
2019; Wang et al., 2019a, 2019b). To further support such a cooperative environment
and coordinated manner for USVs, a new intelligent multi-task alloca­
tion and path planning algorithm has been proposed based upon the The complexity of the uncertain environments has a certain impact
self-organizing map (SOM) and the fast-marching method (FMM) by Liu on the rationality and effectiveness of the autonomous navigation
et al. (2019), Zhou et al. (2019), and Tan et al. (2020). behavioral decision for MASS, which is mainly reflected in the closed

12
X. Zhang et al. Ocean Engineering 235 (2021) 109380

Fig. 17. The overall architecture of maritime collision-avoidance navigation system.

Fig. 18. Waypoint-based global route optimization from Oslo to Trondheim.

loop of the whole voyage (Yoo et al., 2018). Three uncertainties include navigation decision reasoning based on self-learning of navigation
the scene elements in the navigation situation, the space-time charac­ situation.
teristics and status of the obstacles, the binary relationship between In the actual voyage, the navigation behavioral decision of MASS still
MASS and the obstacles effective modeling. Therefore, MASS needs faces more uncertainties (Jahnke et al., 2017; Roy et al., 1999; Sormu­
effective description and modeling of behavior decision expert knowl­ nen et al., 2015), such as:
edge base (international maritime traffic rules, good seamanship) based
on scene division, and intelligent collision avoidance decision and

13
X. Zhang et al. Ocean Engineering 235 (2021) 109380

Fig. 19. Based on the massive AIS trajectory big data, inbound and outbound port routes optimization (Taking massive AIS trajectory big data of Tianjin Port of
China as an example).

1. Uncertainty of marine environment (Katsanevakis and Moustakas, sound signal and visual signal (other ship’s light type). Data trans­
2018). The sea is vast and infinite, and human’s understanding of the mitted by the third party includes AIS data, meteorological forecast
sea is very limited. In the voyage, there are not complete kinds of data and tide calendar data. Due to the different characteristics of
environmental prior knowledge. Therefore, there are many un­ these sensors (principle of action, sensing mechanism, data trans­
certainties in the sea areas lacking of environmental prior knowl­ mission), some uncertainty of sensing information will be caused.
edge, including water depth, reef and other disturbing and 3. There is uncertainty in the accuracy of the prediction of obstacle
obstructing factors. motion and collision trajectory (Park et al., 2017; Johansen et al.,
2. The uncertainty of navigation situation information perception (Park 2016; Patterson et al., 2019; Soloperto et al., 2019). The perception
et al., 2019). Due to the rich information, it simply includes the in­ of all the sensors of MASS brings the space state information, and the
formation obtained by the internal sensor, the information obtained whole decision-making process or the navigation process has distinct
by the external sensor and the information transmitted (shared) by space-time characteristics. These sensors cannot detect or report the
the third party. Internal sensors refer to the platform monitoring of behavior intention and motion state of the dynamic obstacles, such
MASS, generally refer to the health status of command data link, the as the motion direction and speed.
operability and health status of sensors identified as critical, the
operability and health status of onboard system (such as propeller, In an uncertain environment, the navigation decision system and
autopilot, collision avoidance system, etc.), watertight information, algorithm should have the ability of situation assessment based multi-
residual fuel, hull integrity, pitch, roll, heave, and ship vibration source heterogeneous information, the ability to infer the motion state
dynamic. External sensors refer to GNSS, bow direction, sea condi­ of dynamic obstacles and the ability to generate the optimal navigation
tion, wind speed and direction, water depth below keel, radar target, strategy, to deal with the above problems and uncertainties.

Table 5 3.2. Design of maritime collision-avoidance navigation systems

Ontology model attribute table.
No. property categories sub-properties object properties
Maritime collision-avoidance navigation system is a complex system
(Liu et al., 2017), which integrates many advanced intelligent
1 location attribute longitude hasLongitude
technologies.
latitude hasLatitude
2 data attribute relative distance distToEntity In this paper, the whole system is divided into five subsystems: global
direction hasDirection
speed hasVelocity
current state currentAttribute
maximum hasMax
minimum hasMin
3 relationship attribute join conditions connectTo
from
to
incorporate relational has
orientation relation hasBehindLeft
hasBehind
hasBehindRight
hasFrontLeft
hasFront
hasFrontRight
hasLeft
hasRight
position relationship isOn
Fig. 20. Block diagram of collision-avoidance navigation decision-
isFrom
making system.

14
X. Zhang et al. Ocean Engineering 235 (2021) 109380

route optimization, navigation situation awareness, navigation behav­

ioral decision, motion control and execution, and high-performance
communication subsystem. As shown in Fig. 17, the overall system ar­
chitecture of the navigation system for maritime autonomous surface
ships is presented, which describes the collaborative relationship among
the five sub-systems.

3.2.1. Global route optimization

The global route optimization subsystem is to set the waypoint with
the help of ECDIS and GPS system in the early stage of cargo trans­
portation for MASS, to realize the calculation and design of relatively
Fig. 22. A block diagram of a PID controller in a feedback loop. r(t) is the
better and safe routes for known obstacles and port-to-port (Krata and
desired process value or setpoint (SP), and y(t) is the measured process value
Szlapczynska, 2018). For the whole voyage of MASS, in terms of data
(PV). u(t) is the overall control function (PID controlle, 2020).
description, global route optimization is equivalent to optimizing the
navigation strategy for global path planning. If there are obstacles or the
relatively safe and practical. Many historical trajectories are fitted to
original route is blocked in the voyage after planning, the global route
obtain recommended routes and recommended waypoints. As shown in
optimization system will conduct quadratic programming to re-plan a
Fig. 19, the global optimization inbound and outbound routes for
reasonable and optimal global route. The commonly used algorithms are
Tianjin port of China are generated based on massive AIS trajectory data
dynamic programming, a *, Dijkstra and trajectory point guidance
technology.
(Zaccone et al., 2017; 2018; Lee et al., 2018; Biyela and Rawatlal, 2019;
Liu et al., 2019).
3.2.2. Navigation situation awareness
Global route optimization is to find a collision free shortest route
Navigation situation awareness system is to use a variety of onboard
from the known starting point to the terminal point according to the
instruments and equipment to actively perceive the internal and
existing information under the influence of obstacles or bad weather in
external information of the ship or marine navigation environment, and
the marine environment. Generally, the constraints of global route
receive the data transmitted by the third party (Sharma and Nazir, 2017;
optimization include time series, spatial constraints, hydrometeorology
Hyvönen et al., 2015). Perception system is the basis of navigation
and COLREGS. In the practical calculation of global route optimization
behavioral decision and motion control of MASS. The accurate percep­
for autonomous ship, the cost setting strategy is often more important
tion information is also an important benchmark of MASS research and
than routing planning and the shortest path selection (Wang and Yan
development. There are many kinds of sensors and shipborne in­
et al., 2020). By setting the port of origin and destination on the elec­
struments in the navigation situation awareness system of MASS, such as
tronic chart, and through the calculation of route optimization and
ECDIS, radar, lidar, HD camera, sonar, AIS, etc., which can obtain
obstacle spatial constraints, multiple waypoints with cost are set to form
high-precision position service information, maritime safety sailing in­
the optimized global route. As shown in Fig. 18, it is waypoint-based
formation, hydrometeorological information, ship dynamic information
global route optimization from Oslo to Trondheim of Norway.
and port information in real time (Raptodimos et al., 2016). This
Whether it is to determine the relationship between waypoints,
multi-source information is fused and processed. Static and dynamic
establish coastal waypoint relationship database, or use various opti­
obstacles are mapped in the ECDIS and sent to the behavioral decision
mization algorithms to solve the shortest path, there are problems such
system.
as difficulty of establishing practical model and poor adaptability. With
Generally, navigation situation awareness can be divided into three
the development of big data technology and its wide application in the
layers: navigation environment information acquisition, scenario un­
maritime field, the optimal route can be obtained by cleaning, fitting,
derstanding, situation assessment and prediction. For the traditional
mining, classifying, and forecasting the massive AIS trajectory data. The
manned ship, the navigation environment information perception
global optimal route generation model based on maritime big data is
generally refers to the acquisition of the position, course, speed, relative
direction, relative distance, hydrology and meteorology of the ship and
the target ship. Scenario understanding refers to using the information

Fig. 21. The ship attitude represented by the ship model of MASS (Motion in 6
degrees of freedom (DOF), Oe xe ye ze is the earth-fixed coordinate system. Fig. 23. VDE system concept and available communication links (Lázaro
Gxb yb zb is the body-fixed coordinate system, and Gis center of ship gravity). et al., 2019).

15
X. Zhang et al. Ocean Engineering 235 (2021) 109380

Fig. 24. MCP framework concept (EfficienSea2 solution, 2020).

obtained by ship borne navigation aids to calculate the parameters such planning and trajectory planning, corresponding to the MASS, that is,
as DCPA and TCPA. Situation assessment is the prediction and deduction the control of the marine telegraph and rudder (Hanson and Hanson,
of ship navigation situation in the future dynamic obstacle environment 2017; Huang, 2018). There is also a feedback control layer based on the
based on the situation understanding, including the prediction of other integrated error of ship attitude variables in this subsystem. On the
ship’s trajectory and motion. However, the existing research on navi­ voyage of autonomous navigation for MASS, there are often some errors
gation situation awareness is simply to analyze and judge the encounter between the actual navigation and the plan due to the uncertainty of the
scenario based on COLREGS (Sharma et al., 2019). ocean current, swell, and other environments. Therefore, the control
For MASS, navigation situation awareness system realizes the navi­ system will be feedback controlled again based on these errors. On the
gation situation estimation and scene division with the help of the bi­ one hand, the decision instructions can be adjusted in real-time for
nary relationship analysis. Using ontology model and the idea of divide re-planning to better conform to the current navigation behavior. On the
and conquer, the multi-source heterogeneous information obtained by other hand, the navigation behavioral and motion can be corrected. The
the navigation situation awareness layer of MASS is clustered into ship’s navigation behavior can avoid uncertain risks.
different scene entities, and binary or multivariate attributes are From the perspective of ship attitude control, the feedback control
established. Table 5 shows the attribute table of ontology model, part of MASS is not essentially different from that of ordinary ship. Both
including location attribute, data attribute and relationship attribute. are based on a certain preset trajectory, considering the error of current
ship attitude and planned route, and continuously tracking feedback
3.2.3. Decision-making control. In maneuvering, a marine craft experiences motion in 6 degrees
The navigation behavioral decision system is the core part of the of freedom (DOFs) (Fossen, 2011). The MASS attitude represented by
whole MASS - navigation brain (Xinping et al., 2019). The system takes the ship model is in a three-dimensional coordinate system, and the
the results of the perception system as input and collects all the infor­ MASS attitude can be fully described by surge, roll, sway, pitch, heave,
mation of the navigation situation, including not only the current po­ yaw. The ship attitude represented by the ship model of MASS is shown
sition, speed, and course of the autonomous ship, but also the in Fig. 21.
information of obstacles. The decision system of a maritime autonomous As for the feedback control system of MASS, the problem we need to
navigation system is to determine the route and navigation strategy of solve is to control the ship to follow the space-time trajectory of global
the MASS on the basis of knowing navigation safety information (Xue route optimization and collision-avoidance navigation decision-making
et al., 2019; Shaobo et al., 2020). system as far as possible. The distinguishing feature of the PID controller
The decision-making system takes the output of the global route is the ability to use the three control terms of proportional, integral and
optimization system as the guidance information. And gives the driving derivative influence on the controller output to apply accurate and
behavior instructions of the unmanned ship through the system self- optimal control. Early PID controller was developed by observing the
learning, including longitudinal steering avoidance, such as left navigation behavior of officers in maneuvering a vessel on course in the
rudder, right rudder, etc., and lateral variable speed avoidance, such as face of varying disturbance such as wind and sea state (Ma et al., 2019;
acceleration, deceleration, parking, etc. In the future development of Zhang et al., 2020; Chen et al., 2020; Wikipedia, 2020). The structure of
collision-avoidance decision making, it is found that the traditional soft a typical PID feedback control system is shown in Fig. 22. Where e(t)
computing and artificial intelligence algorithm must be combined to represents the current tracking error, and the tracking variable error can
select the most suitable algorithm according to the needs of the scene, be the longitudinal/transverse error, angle/curvature error or the
and adaptive collision-avoidance navigation decision-making in an un­ comprehensive error of some attitude state variables of ship. The P
certain environment. As shown in Fig. 20, it is the block diagram of the controller represents the feedback to the current error, and its gain is
collision-avoidance navigation decision-making system. controlled by KP . The I and D controllers represent integral and differ­
ential terms respectively, and their gains are controlled by KI and KD
3.2.4. Control and execution respectively.
After the decision instruction is given by the decision system of
maritime autonomous navigation system, the control and execution 3.2.5. High-performance communication
system of MASS will execute the instruction, mainly including speed The high-performance communication subsystem guarantees the

16
X. Zhang et al. Ocean Engineering 235 (2021) 109380

data or safety information distribution and sharing between satellite, and improves the intelligent communication technology of MASS (MCP
ship, and shore. The same as UAVs/UGVs communication, there are four consortium, 2019).
types of MASS communication services: (1) MASS-to-MASS for data and
control links; (2) MASS-to-Shore Control Center (SCC) for control and 4. Trends in maritime collision-avoidance navigation systems
commands link; (3) MASS-to-Ground Wireless Nodes for MASS-aided
data dissemination and collection; and (4) MASS-to-Satellite system. 4.1. Analysis on trend of maritime collision-avoidance navigation systems
Due to the need to transmit many sensor information and equipment
status information, as well as radar images, sea video and so on between There are two main types of collision-avoidance navigation systems
the ship and shore, the communication volume is large. Therefore, the for MASS: rule-based navigation system and learning algorithm-based
unmanned ships put forward high bandwidth, low delay, low cost, and navigation system. The rule-based collision-avoidance navigation sys­
other requirements for the maritime communication system. tem is to divide the behavior of the MASS, establish the behavior rule
As for ship collision avoidance, the high-performance communica­ base according to the COLREGS, knowledge, experience, and traffic
tion system is particularly important. AIS, GMDSS, and other systems are regulations, divide the ship state according to different environmental
mainly used for collision avoidance of ships. However, the growth in AIS information, and determine the navigation strategy according to the rule
has been such, that in some of the most crowded waters the system is as logic. The representative methods are finite state machine and expert
of today already overloaded. The International Association of Marine system. At present, the research of learning algorithm-based collision-
Aids to Navigation and Lighthouse Authorities (IALA) started the work avoidance navigation system has achieved remarkable results. Accord­
on the Very-high-frequency Data Exchange System (VDES) (Report ing to different principles, it can be divided into deep learning-related
ITU-R M.2371-0, 2015). Rather than an evolution of AIS, VDES is a navigation methods and machine learning-based navigation methods,
communications system encompassing different communications sub­ such as fuzzy models.
systems, which including AIS and Application Specific Messages (ASM) Rules-based and learning algorithm-based collision-avoidance navi­
channels. Furthermore, VDES has a third subsystem, called VDE, which gation technologies have their advantages and disadvantages. For a rule-
allows higher rate communications, and is highly flexible to be able to based collision-avoidance navigation system, its advantages are the al­
support a variety of services in the future. A key characteristic of VDES is gorithm logic is clear, strong interpretability, strong stability, and easy
that it does not only support direct ship-to-ship and ship-to-shore to model. The system operation does not require high processor per­
communication, but it also foresees a satellite component specifically formance. The model has strong adjustability and expansibility, and it
for VDE. Fig. 23 shows the VDE system concept and available commu­ can realize more complex combination functions through the layering of
nication links (Lázaro et al., 2019; Golaya and Yogeswaran, 2020). state machine. It has advantages in the breadth traversal of function
With the development and support of e-navigation technology, the scenarios. However, the navigation behavior is incoherent due to the
Maritime Connectivity Platform (MCP) has been developed and tested, state cutting conditions. The trigger conditions of the behavior rule base
which is a communication framework enabling efficient, secure, reli­ are easy to overlap, resulting in system failure; the insufficient depth
able, and seamless electronic information exchange between all autho­ traversal of the scene makes it difficult to improve the accuracy of the
rized maritime stakeholders across available communication systems. system decision-making, and there is a bottleneck for the improvement
AS is shown in Fig. 24, MCP has three core components, an identity of the performance of the complex condition processing and algorithm.
registry, a service registry, and a messaging service. In general, MCP For learning algorithm-based collision-avoidance navigation system, its
integrates the information resources of ship end and shore end, improves advantages are it has the advantage of scene traversal depth. For a
the level of in-depth development of information resources and certain subdivision scene, it is easier to cover all navigation situations
comprehensive utilization of unmanned technology, promotes the deep through the big data system. The scale of the navigation algorithm can
integration of cloud computing and maritime management and services, be simplified by using the network structure. Meanwhile, some

Fig. 25. Brain-inspired goal-directed navigation. win is input connection weight.wsys is internal connection weight.wout is output connection weight.wback is feedback
connection weight.

17
X. Zhang et al. Ocean Engineering 235 (2021) 109380

Fig. 26. E-navigation overall technical architecture.

navigation system. The top layer uses finite state machine to traverse
hierarchically according to the scene. The bottom layer uses a
learning algorithm to apply in modules based on the specific scene.
The research focus of this method is how to connect the finite state
machine and the learning algorithm model reasonably and the
overfitting/underfitting problems.
3. In the wake of developments in technology that combined rule-based
with learning-based, technology brain-inspired and cognitive navi­
gation has become more capable of conquering uncertainty in com­
plex navigation situations. It integrates perception, decision-making,
planning, and control to realize the intellectualization and human-
analogy of collision avoidance navigation systems.
4. E-navigation development lays a technical foundation for the con­
struction of the autonomous collision-avoidance navigation system
in the aspects of intelligent perception, intelligent navigation deci­
sion, intelligent communication, and intelligent control.

4.2. Cognitive navigation and its thought of brain-inspired realization

Fig. 27. Framework of autonomous navigation system based on e-Navigation. With the continuous development of brain, neuroscience, and arti­
(C&E stands for control and execution center. CAN stands for collision- ficial intelligence technology, the integration of perception, cognition,
avoidance navigation system. MSP stands for maritime service portfolios.). path planning and behavior decision-making inspired by the brain
navigation mechanism of insects and mammals has been greatly devel­
machines have self-learning performance, and the machines can refine oped. From the input of the original perception information to the direct
the environmental characteristics and decision attributes by themselves, output of collision-avoidance navigation decision, it presents the intel­
which is convenient for system optimization iteration. But it is difficult ligent behavior of human end-to-end navigation. It has the potential to
to modify the model due to the poor interpretability of the navigation improve robustness, accuracy, real-time response, autonomous intelli­
decision results of the algorithm. The learning algorithm does not have gence, and computational efficiency. Generally, it can be divided into
the advantage of the breadth of scene traversal, and the learning models brain-inspired spatial cognition layer and brain-inspired goal-directed
used in different scenes may be completely different. As is known, the navigation layer (Vijesh et al., 2013). For intelligent ships, the concept
navigation effect depends on the quality of data. Insufficient samples, of “navigation brain” was first proposed by academician Yan Xinping’s
poor data quality, and unreasonable network structure will lead to team. “Navigation brain” system is an artificial intelligence system
overlearning, under learning, and other problems. based on reinforcement learning, which gradually replaces the human
According to the advantages and disadvantages of the two methods, brain with a machine brain to realize the development of ship intelli­
the development trend of collision-avoidance navigation system for gence and even unmanned. The system is an artificial intelligence sys­
MASS can be summarized as follows: tem for ship intelligent navigation, which is composed of three
functional spaces: perception, cognition and decision execution (Xinp­
1. The rule-based algorithm will still be widely used in a collision- ing, 2017; 2019; Xue, Yan et al., 2019).
avoidance navigation system. It will be used as the top-level archi­ Brain-inspired spatial cognition can solve the self-motion informa­
tecture of the navigation system and the subdivision solution of some tion extracted from the brain-inspired perception process of environ­
specific problems, and more hybrid structures will be used. The ment and the analyzed environment beacon information through the
research focus of this method will be to solve the reasonable self-organizing group discharge activity of brain navigation cells. The
decision-making problem of the “gray area” of state division, and the navigation information is transformed into specific navigation cell
overlapping of trigger conditions of behavior rule base of navigation. discharge activities in the brain. At the same time, the location infor­
2. The combination of rule-based methods and learning algorithm- mation and beacon information are related and stored by the location
based methods will be more used in the collision-avoidance cells, completed construction of cognitive map.
Through unsupervised learning, the intelligent neural connections

18
X. Zhang et al. Ocean Engineering 235 (2021) 109380

between multiple navigation cells and multiple action cells are estab­ strategy (Im et al., 2018; Porathe and Rødseth., 2019; Ahn., et al., 2019;
lished by brain-inspired goal-directed navigation, in the cognitive map. Jeong et al., 2018). It integrates e-navigation technology and autono­
Navigation cells mainly include head facing cells, position cells and mous ship technology. In the data center, it establishes a database of the
boundary cells, which represent various navigation information such as information sensed by the MASS based on the standard of S-100, and
orientation, position and obstacles. Action cells mainly represent a va­ transmits information to the ship and the shore-based platform through
riety of navigation behaviors, such as steering, acceleration and decel­ the maritime cloud, to realize the autonomous navigation of MASS
eration (Yang et al., 2020). The intelligent behavior inspired by human without collision.
navigation is realized. The brain-inspired goal-directed navigation in The four major issues that e-navigation technology system mainly
Fig. 25. solves are perception, data, standard and transmission. From the com­
mon research of e-navigation and autonomous navigation of MASS, e-
4.3. Collision-avoidance navigation based on e-Navigation navigation development lays a technical foundation for the construction
of autonomous navigation system in the aspects of intelligent percep­
In May 2006, at the 81st meeting of the Maritime Safety Committee tion, intelligent navigation decision, intelligent communication, and
(MSC) of the IMO, the Seven Countries Proposal-"Development of e- intelligent control.
Navigation Strategy” was adopted and adopted by The International
Association of Marine Aids to Navigation and Lighthouse Authorities 5. Conclusion
(IALA). E-navigation refers to the coordinated collection, integration,
exchange, display, and analysis of maritime information onboard and The importance of maritime autonomous navigation systems is un­
onshore by electronic means to enhance the navigational capabilities of deniable and the opportunity for coordinated and interconnected op­
berths and other related services to improve the level of safety and se­ erations is clear. MASS may finish intelligent navigation through shore
curity at sea and protect the marine environment (IMO MSC 81/23/10). remote control center with long distance to the operations, so that
The e-navigation concept was put forward to meet the rapid devel­ dependence on more autonomy infrastructures, such as maritime
opment of autonomous navigation technology and navigation assistance autonomous navigation system, collision avoidance decision support
methods. It aims to achieve the optimization of maritime transportation systems, or motion control systems must be expected. The cost, reli­
by integrating the existing navigation assistance technology and tools. ability, performance, and availability of such systems are important
The e-navigation technology framework mainly includes three elements, issues.
the ship environment, the shore-based support environment, and the Moreover, there is a wide variety of scenarios with different collision
communication system. Ship environment refers to supporting the avoidance decision requirements with respect to data-rates, latency, and
collection, integration, exchange, display, and analysis of all informa­ importance. These are, for instance, command and control data
tion provided by ship-based sensors. Shore-based support environment (telemetry), sensor data for situation awareness, payload sensor data,
refers to shore-based technical services that support shore-based appli­ collision avoidance transponder broad-casts, and status information.
cations, such as search and rescue, VTS, ports, and MSI (Maritime Safety Therefore, autonomous navigation strongly depends on the system au­
Information) services, etc. Communication systems refer to the tonomy level and situation needs.
communication equipment and communication links between ships- This work reviews the major advancements in maritime collision-
ships, shore-ships. To this end, the overall technical architecture of e- avoidance navigation technologies applied in several different sce­
navigation can be simply described as the three sides of the coin, as narios, from transportation to scientific research. Moreover, it highlights
shown in Fig. 26. The front and back sides of the coin represent the ship how available technologies and systems can be composed to efficiently
environment and the shore-based support environment, and the side of and effectively handling in maritime obstacle environments.
the coin represents the link ship communication system with the shore Existing and prototype maritime autonomous surface ships,
(Wang et al., 2017). collision-avoidance navigation technologies are characterized,
At present, with the development of shipping industry, ships have describing their requirements and capabilities. Additionally, the design
shown the characteristics of large-scale, specialized, high-speed, and of maritime collision-avoidance navigation systems is highlighted,
intelligent. E-navigation tries to integrate the existing navigation tech­ considering the availability and performance of different autonomy
nology to maximize the safety of ship navigation and improve the effi­ levels. The discussed are aligned with current trends in the collision-
ciency of maritime cargo transportation. MASS combined with artificial avoidance navigation system, e-navigation technologies, and brain-
intelligence technology greatly reduces the impact of human factors on inspired cognitive navigation.
maritime transportation safety and improves the level of ship navigation
safety. The combination of e-navigation technology and maritime Author contributions
autonomous navigation technology can effectively promote the devel­
opment of intelligent and information technology of maritime trans­ Xinyu Zhang: Conceptualization, Methodology, Validation, Formal
portation and enhance the safety of navigation. analysis, Investigation, Resources, Writing - original draft, Writing -
To effectively improve the safety level of maritime transportation review & editing, Supervision, Project administration, Funding acqui­
and combine the autonomous navigation with e-navigation, the overall sition. Chengbo Wang: Conceptualization, Methodology, Investigation,
technical framework of the autonomous navigation system of MASS Writing – original draft, Writing - review & editing. Lingling Jiang:
based on e-navigation is shown in Fig. 27, and autonomous navigation is Conceptualization, Methodology, Investigation, Writing - review &
developed based on the intelligent environment state information editing. Lanxuan An: Methodology, Formal analysis, Writing - original
perception, intelligent navigation decision and intelligent communica­ draft, Writing - review & editing. Rui Yang: Methodology, Writing –
tion of e-navigation (Gao Zongjiang et al., 2017). The application of original draft, Writing - review & editing.
e-navigation technology lays a foundation for the development of
autonomous navigation of MASS, and promotes the implementation of Funding
e-navigation strategy.
E-navigation relies on four major issues of perception, data, stan­ This work was supported by the National Natural Science Foundation
dards, and transmission, and moves from theory to practical application. of China (grant no.51779028), and the National Key Research and
The application of e-navigation technology lays a foundation for the Development Program of China (grant no. 2018YFB1601502).
development of autonomous navigation technology of MASS, and the
development of MASS also promotes the implementation of e-navigation

19
X. Zhang et al. Ocean Engineering 235 (2021) 109380

Declaration of competing interest Foundation of India 16, 119–131. https://doi.org/10.1080/

09733159.2020.1845456.
Goodwin, E.M., 1975. A statistical study of ship domains. J. Navig. 28, 328–344. https://
The authors declare that they have no known competing financial doi.org/10.1017/S0373463300041230.
interests or personal relationships that could have appeared to influence Han, J., Cho, Y., et al., 2020. Autonomous collision detection and avoidance for ARAGON
the work reported in this paper. USV: development and field tests. J. Field Robot. 37, 987–1002. https://doi.org/
10.1002/rob.21935.
Hanson, B.B., Hanson, T.E., 2017. U.S. Patent No. 9,669,904. U.S. Patent and Trademark
Acknowledgements Office, Washington, DC.
Harbin Engineering University, 2018. Unmanned surface vehicle of school of naval
architecture. http://sec.hrbeu.edu.cn/2018/0925/c443a201600/page.htm.
The authors thank the Editor-in-Chief, Technical Editor, and anon­ (Accessed 20 September 2019).
ymous referees for their invaluable comments. Holzhüter, T., 1997. LQG approach for the high-precision track control of ships. In: IEE
Proceedings-Control Theory and Applications, vol. 144, pp. 121–127. https://doi.
org/10.1049/ip-cta:19971032.
References Huang, R., 2018. Maritime intelligent real-time control system based on UAV. In:
Proceeding of International Conference on Robots & Intelligent System (ICRIS),
Aawa, 2017. The advanced autonomous waterborne applications. https://www.rolls- Changsha, China, pp. 10–12. https://doi.org/10.1109/ICRIS.2018.00011, 26-27
royce.com/media/press-releases.aspx#27-11-2017-rr-opens-first-ship-intelligence-e May 2018.
xperience-space. (Accessed 27 November 2020). Huang, Y., Chen, L., et al., 2019. Generalized velocity obstacle algorithm for preventing
Abs, 2020. ADVISORY ON AUTONOMOUS FUNCTIONALITY. ship collisions at sea. Ocean. Eng. 173, 142–156. https://doi.org/10.1016/j.
Ahn, J.H., Rhee, K.P., You, Y.J., 2012. A study on the collision avoidance of a ship using oceaneng.2018.12.053.
neural networks and fuzzy logic. Appl. Ocean Res. 37, 162–173. https://doi.org/ Huang, Y., van Gelder, et al., 2018. Velocity obstacle algorithms for collision prevention
10.1016/j.apor.2012.05.008. at sea. Ocean. Eng. 151, 308–321. https://doi.org/10.1016/j.
Ahn, J., Joung, T.H., Kang, S.G., Lee, J., 2019. Changes in container shipping industry: oceaneng.2018.01.001.
autonomous ship, environmental regulation, and reshoring. Journal of International Huazhong University of Science and Technology, 2018. Industrial technology research
Maritime Safety, Environmental Affairs, and Shipping 3, 21–27. https://doi.org/ institute of guangdong huazhong university of science and technology. http://www.
10.1080/25725084.2019.1678564. hustmei.com/document/201805/article1761.htm. (Accessed 20 September 2019).
Allensworth, T., 1999. A short history of Sperry Marine. Litton Marine Systems. Hyvönen, M., Rajala, M., Virtanen, A., et al., 2015. Assistive situation awareness system
https://www.sperrymarine.com/corporate-history/sperry-marine. (Accessed 10 for mobile multimachine work environments. IEEE Trans. Intell. Transport. Syst. 16,
March 2021). 3403–3413. https://doi.org/10.1109/TITS.2015.2453208.
Autonomous shipping, 2019. First autonomous manoeuvring vessel trials held on North Im, I., Shin, D., Jeong, J., 2018. Components for smart autonomous ship architecture
Sea. https://autonomousshipping.nl/2019/03/27/first-autonomous-manoeuvrin based on intelligent information technology. Procedia computer science 134, 91–98.
g-vessel-trials-held-on-north-sea/. (Accessed 24 October 2020). https://doi.org/10.1016/j.procs.2018.07.148.
Biyela, P., Rawatlal, R., 2019. Development of an optimal state transition graph for Imo, 1979. International convention on maritime search and rescue (SAR). https://www.
trajectory optimisation of dynamic systems by application of Dijkstra’s algorithm. imo.
Comput. Chem. Eng. 125, 569–586. https://doi.org/10.1016/j. org/en/About/Conventions/Pages/International-Convention-on-Maritime-Search
compchemeng.2019.03.025. -and-Rescue-(SAR).aspx. (Accessed 21 March 2021).
Blue signal, 2020. Intelligent unmanned surface vehicle to realize autonomous Imo’s 81st Maritime Safety Committee Document (MSC 81/23/10): Japan, Marshall
navigation in real sea conditions. http://szb.dlxww.com/dlrb/html/2020-07/02/ Islands, Netherlands, Norway, Singapore, United Kingdom and United States
content_1565007.htm?div=-1. (Accessed 8 December 2020). Propose "Development of Electronic Navigation Strategy".
CCS, 2015. The Code for Intelligent Ship (2015). Inas, 2019. International network for autonomous ships. http://www.autonomous-ship.
CCS, 2018. Guidelines for Autonomous Cargo Ship. org/events.html. (Accessed 20 September 2019).
CCS, 2020. The Code for Intelligent Ship (2020). Jahnke, A., Arp, H.P.H., Escher, B.I., et al., 2017. Reducing uncertainty and confronting
Chen, Z., Chen, D., et al., 2020. Deep learning for autonomous ship-oriented small ship ignorance about the possible impacts of weathering plastic in the marine
detection. Saf. Sci. 130, 104812 https://doi.org/10.1016/j.ssci.2020.104812. environment. Environ. Sci. Technol. Lett. 4, 85–90. https://doi.org/10.1021/acs.
Chen, Y.Y., Lee, C.Y., Tseng, S.H., Hu, W.M., 2020. Nonlinear optimal control law of estlett.7b00008.
autonomous unmanned surface vessels. Appl. Sci. 10, 1686. https://doi.org/ Jeong, S.H., Shim, J.H., Choi, K.S., Son, Y.C., 2018. Analysis and design of common
10.3390/app10051686. platform core technology for maritime autonomous surface ships. The Journal of
ClassNK, 2020a. Guidelines for Concept Design of Automated Operation/Autonomous Advanced Navigation Technology 22, 507–513. https://doi.org/10.12673/
Operation of Ships (Provisional Version). January 2020. jant.2018.22.6.507.
ClassNK, 2020b. In: Guidelines for Digital Smart Ships, first ed. August 2020. Johansen, T.A., Perez, T., Cristofaro, 2016a. Ship collision avoidance and COLREGS
Cui, Z., Li, Q., Cao, Z., Liu, N., 2019. Dense attention pyramid networks for multi-scale compliance using simulation-based control behavior selection with predictive hazard
ship detection in SAR images. IEEE Trans. Geosci. Rem. Sens. 57, 8983–8997. assessment. IEEE Trans. Intell. Transport. Syst. 17, 3407–3422. https://doi.org/
https://doi.org/10.1109/TGRS.2019.2923988. 10.1109/TITS.2016.2551780.
Daley, A.J., 1977. HF radio marine telex. IEEE Trans. Veh. Technol. 26, 199–202. Johansen, T.A., Cristofaro, A., Perez, T., 2016b. Ship collision avoidance using scenario-
https://doi.org/10.1109/T-VT.1977.23680. based model predictive control. IFAC-PapersOnLine. 49, 14–21. https://doi.org/
Davis, P.V., Dove, M.J., Stockel, C.T., 1980. A computer simulation of marine traffic 10.1016/j.ifacol.2016.10.315.
using domains and arenas. J. Navig. 33, 215–222. https://doi.org/10.1017/ Katsanevakis, S., Moustakas, A., 2018. Uncertainty in marine invasion science. Frontiers
S0373463300035220. in Marine Science 5, 38. https://doi.org/10.3389/fmars.2018.00038.
EfficienSea2 solution, 2020. Maritime connectivity platform. https://efficiensea2.org/s Knébel, 2020. Detecting Obstacles from Camera Image at Open Sea. M.A. Thesis. Aalto
olution/maritime-connectivity-platform/. (Accessed 20 March 2021). University, Espoo.
Emsa, 2020. Preliminary annual overview OF marine casualties and incidents 2014- Kongsberg, 2020. Kongsberg maritime and MASSTERLY to equip and operate two zero-
2019. http://www.emsa.europa.eu/accident-investigation-publications/annual-over emission autonomous vessels for asko. https://www.kongsberg.com/maritime/abo
view.html. (Accessed 18 September 2020). ut-us/news-and-media/news-archive/2020/zero-emission-autonomous-vessels/.
Eriksen, B.O.H., Breivik, M., et al., 2019. The branching-course model predictive control (Accessed 18 September 2020).
algorithm for maritime collision avoidance. J. Field Robot. 36, 1222–1249. https:// Krata, P., Szlapczynska, J., 2018. Ship weather routing optimization with dynamic
doi.org/10.1002/rob.21900. constraints based on reliable synchronous roll prediction. Ocean. Eng. 150, 124–137.
Fan, S., Zhang, J., Blanco-Davis, E., Yang, Z., Yan, X., 2020. Maritime accident https://doi.org/10.1016/j.oceaneng.2017.12.049.
prevention strategy formulation from a human factor perspective using Bayesian Kumar, A., Kurmi, J., 2018. A review ON unmanned water surface vehicle. Int. J. Adv.
Networks and TOPSIS. Ocean. Eng. 210, 107544 https://doi.org/10.1016/j. Res. Comput. Sci. 9, 95–99. https://doi.org/10.26483/ijarcs.v9i0.6145.
oceaneng.2020.107544. Lázaro, F., Raulefs, R., Wang, W., Clazzer, F., Plass, S., 2019. VHF Data Exchange System
Fang, M.C., Zhuo, Y.Z., Lee, Z.Y., 2010. The application of the self-tuning neural network (VDES): an enabling technology for maritime communications. CEAS space Journal
PID controller on the ship roll reduction in random waves. Ocean. Eng. 37, 529–538. 11, 55–63. https://doi.org/10.1007/s12567-018-0214-8.
https://doi.org/10.1016/j.oceaneng.2010.02.013. Lee, S.M., Roh, M.I., Kim, K.S., Jung, H., Park, J.J., 2018. Method for a simultaneous
Fang, M.C., Lin, Y.H., Wang, B.J., 2012. Applying the PD controller on the roll reduction determination of the path and the speed for ship route planning problems. Ocean.
and track keeping for the ship advancing in waves. Ocean. Eng. 54, 13–25. https:// Eng. 157, 301–312. https://doi.org/10.1016/j.oceaneng.2018.03.068.
doi.org/10.1016/j.oceaneng.2012.07.006. Li, Y., Zheng, J., 2020. Real-time collision avoidance planning for unmanned surface
Fossen, Thor I., 2011. In: Handbook of Marine Craft Hydrodynamics and Motion Control, vessels based on field theory. ISA Trans. 106, 233–242. https://doi.org/10.1016/j.
first ed. John Wiley & Sons, West Sussex, United Kingdom, pp. 1–575. isatra.2020.07.018.
Gao, Zongjiang, Zhang, Yingjun, et al., 2017. Research summary of unmanned ship, 43. Lin, Xin-long, Zhang, Bi-sheng, 2018. Technical development and operational
Journal of Dalian Maritime University, pp. 1–7. https://doi.org/10.16411/j.cnki. application of unmanned surface combat system. Journal of Unmanned Undersea
issn1006-7736.2017.02.001. Systems 26, 107–114.
Golaya, A.P., Yogeswaran, N., 2020. Maritime communication: from flags to the VHF Liu, B., Wang, S.Z., et al., 2019. Ship recognition and tracking system for intelligent ship
data exchange system (VDES). Maritime affairs. Journal of the National Maritime based on deep learning framework. TransNav: International Journal on Marine

20
X. Zhang et al. Ocean Engineering 235 (2021) 109380

Navigation and Safety of Sea Transportation 13, 699–705. https://doi.org/ Richter, M., 2006. Operational Manning Considerations for Spartan Scout and Sea Fox
10.12716/1001.13.04.01. Unmanned Surface Vehicles (USV). Naval Postgraduate School Monterey Ca
Liu, H., Deng, R., Zhang, L., 2016a. In: The Application Research for Ship Collision Graduate School of Business and Public Policy.
Avoidance with Hybrid Optimization Algorithm. 2016. IEEE International Roberts, G.N., 2008. Trends in marine control systems. Annu. Rev. Contr. 32, 263–269.
Conference on Information and Automation (ICIA), Ningbo, pp. 760–767. https:// https://doi.org/10.1016/j.arcontrol.2008.08.002.
doi.org/10.1109/ICInfA.2016.7831921. Roy, N., Burgard, W., Fox, D., Thrun, S., 1999. Coastal navigation-mobile robot
Liu, Zhixiang, et al., 2016b. Unmanned surface vehicles: an overview of developments navigation with uncertainty in dynamic environments. In: Proceeding of IEEE
and challenges. Annu. Rev. Contr. 41, 71–93. https://doi.org/10.1016/j. International Conference on Robotics and Automation (Cat. No. 99CH36288C),
arcontrol.2016.04.018. Detroit, MI, USA, 10-15 May, vol. 1, pp. 35–40. https://doi.org/10.1109/
Liu, S., Li, L., Tang, J., et al., 2017. Creating autonomous vehicle systems. Synthesis ROBOT.1999.769927.
Lectures on Computer Science 6, i–186. https://doi.org/10.2200/ Shanghai University, 2018. Unmanned surface vehicle engineering research institute.
S00787ED1V01Y201707CSL009. http://www.jhai.shu.edu.cn/Default.aspx?Tabid=35148. (Accessed 20 September
Liu, C., Mao, Q., Chu, X., Xie, S., 2019a. An improved A-star algorithm considering water 2019).
current, traffic separation and berthing for vessel path planning. Appl. Sci. 9, 1057. Shaobo, W., Yingjun, Z., Lianbo, L., 2020. A collision avoidance decision-making system
https://doi.org/10.3390/app9061057. for autonomous ship based on modified velocity obstacle method. Ocean. Eng. 215,
Liu, Y., Song, R., Bucknall, R., et al., 2019b. Intelligent multi-task allocation and 107910 https://doi.org/10.1016/j.oceaneng.2020.107910.
planning for multiple unmanned surface vehicles (USVs) using self-organising maps Sharma, A., Nazir, S., 2017. Distributed Situation Awareness in pilotage operations:
and fast marching method. Inf. Sci. 496, 180–197. https://doi.org/10.1016/j. implications and challenges. International Journal on Marine Navigation and Safety
ins.2019.05.029. of Sea Transportation 11, 289–293. https://doi.org/10.12716/1001.11.02.11.
Lyu, H., Yin, Y., 2019. COLREGS-constrained real-time path planning for autonomous Sharma, A., Nazir, S., Ernstsen, J., 2019. Situation awareness information requirements
ships using modified artificial potential fields. J. Navig. 72, 588–608. https://doi. for maritime navigation: a goal directed task analysis. Saf. Sci. 120, 745–752.
org/10.1017/S0373463318000796. https://doi.org/10.1016/j.ssci.2019.08.016.
Ma, Yong, Sheng, Guangming, 2018. A Sea and Air Autonomous Cooperative Search and Sheng, L., Ping, Y., Yan-yan, L., Yan-chun, D., 2006. Application of H infinite control to
Rescue M3U Platform. Hubei: CN108563242A, 2018-09-21. ship steering system. J. Mar. Sci. Appl. 5, 6–11. https://doi.org/10.1007/s11804-
Ma, X., Wang, L., Xu, J., 2019. Active vibration control of rib stiffened plate by using 006-0041-8.
decentralized velocity feedback controllers with inertial actuators. Appl. Sci. 9, Smart, Ocean, 2020. Zhihui C series multifunctional unmanned ship of Smart Ocean
3188. https://doi.org/10.3390/app9153188. Technology. http://www.tjzhocean.com/pro/wurenchuan/zhineng/178.html.
Maritime, U.K., 2018. Maritime Autonomous Surface Ships UK Code of Practice (V2). (Accessed 18 October 2020).
Society of Maritime Industries. Soloperto, R., et al., 2019. Collision avoidance for uncertain nonlinear systems with
Maritime, U.K., 2019. Maritime Autonomous Surface Ships UK Industry Conduct moving obstacles using robust Model Predictive Control. In: Proceeding of 18th
Principles and Code of Practice (V3). Society of Maritime Industries. European Control Conference (ECC), Naples, Italy, 25-28 June, pp. 811–817.
Maritime, U.K., 2020. Maritime Autonomous Ship Systems (MASS) UK Industry Conduct https://doi.org/10.23919/ECC.2019.8796049.
Principles and Code of Practice 2020 (V4). Society of Maritime Industries. Sormunen, O.V.E., et al., 2015. Uncertainty in maritime risk analysis: extended case
Mcp consortium, 2019. Maritime connectivity platform. https://maritimeconnectivity. study on chemical tanker collisions. Proc. IME M J. Eng. Marit. Environ. 229,
net/. (Accessed 20 March 2021). 303–320. https://doi.org/10.1177/1475090213515640.
Minorsky, N., 1922. Directional stability of automatically steered bodies. Journal of the Sun, M., Luan, T., Liang, L., 2018. RBF neural network compensation-based adaptive
American Society for Naval Engineers 34, 280–309. control for lift-feedback system of ship fin stabilizers to improve anti-rolling effect.
Mitrofanov, O., 1968. An anti-collision indicator. J. Navig. 21, 163–170. https://doi.org/ Ocean. Eng. 163, 307–321. https://doi.org/10.1016/j.oceaneng.2018.06.011.
10.1017/S0373463300030319. Tan, G., Zou, J., Zhuang, J., Wan, L., Sun, H., Sun, Z., 2020. Fast marching square method
MSC International Maritime Organization, 2017. Maritime Autonomous Surface Ships based intelligent navigation of the unmanned surface vehicle swarm in restricted
Proposal for a Regulatory Scoping Exercise. MSC 98/20/2, February. waters. Appl. Ocean Res. 95, 102018 https://doi.org/10.1016/j.apor.2019.102018.
MSC International Maritime Organization, 2018a. Regulatory Scoping Exercise for the Vijesh, M., Iyengar, S., et al., 2013. A navigation algorithm inspired by human
Use of Maritime Autonomous Surface Ships (MASS). MSC 99/WP.9, May. navigation. In: Proceedings of 2012 IEEE/ACM International Conference on
MSC International Maritime Organization, 2018b. Regulatory Scoping Exercise for the Advances in Social Networks Analysis and Mining, Istanbul, Turkey, 26-29 Aug.
Use of Maritime Autonomous Surface Ships (MASS). MSC 100/WP.8, December. 2012, vol. 225. IEEE, New Jersey, USA, pp. 1309–1314. https://doi.org/10.1109/
MSC International Maritime Organization, 2019. INTERIM GUIDELINES FOR MASS ASONAM.2012.225, 2013.
TRIALS. MSC.1/Circ.1604, June. Wan, Jie-xi, 2014. Status and development trends of foreign military unmanned surface
Msc International Maritime Organization, 2020. Status Report – Progress of the boats. National Defense Science & Technology 35, 91–96.
Regulatory Scoping Exercise. MSC 102/5, January. Wang, T., Wu, Q., et al., 2020. Autonomous decision-making scheme for multi-ship
Munin, 2016. Maritime unmanned navigation through intelligence in network. http:// collision avoidance with iterative observation and inference. Ocean. Eng. 197,
www.unmanned-ship.org/munin/. (Accessed 20 September 2019). 106873 https://doi.org/10.1016/j.oceaneng.2019.106873.
Nature Publishing Group, 1913. International Conference on the Safety of Life at Sea. Wang, C.B., et al., 2017. e-Navigation and its data standard analysis. Journal of World
Nature 92, 355. https://doi.org/10.1038/092355a0. Shipping 40, 22–24. https://doi.org/10.16176/j.cnki.21-1284.2017.01.005.
Nfas, 2019. Norwegian forum for autonomous ships. http://nfas.autonomous-ship.org/i Wang, C., Zhang, X., Cong, L., et al., 2019. Research on intelligent collision avoidance
ndex-en.html. (Accessed 20 September 2019). decision-making of unmanned ship in unknown environments. Evolving Systems 10,
Ning, Wang, Pan, Xinxiang, 2019. Path-following of autonomous underactuated ships: a 649–658. https://doi.org/10.1007/s12530-018-9253-9.
translation-rotation cascade control approach. IEEE/ASME Transactions on Wang, N., Sun, Z., Jiao, Y., Han, G., 2019a. Surge-heading guidance based finite-time
Mechatronics. 24, 2583–2593. https://doi.org/10.1109/TMECH.2019.2932205. path-following of underactuated marine vehicles. IEEE Trans. Veh. Technol. 68,
Ning, Wang, Reza Karimi, Hamid, 2020. Successive waypoints tracking of an 8523–8532. https://doi.org/10.1109/TVT.2019.2927893.
underactuated surface vehicle. IEEE Transactions on Industrial Informatics 16, Wang, Ning, Gao, Yuncheng, et al., 2018. Implementation of an integrated navigation,
898–908. https://doi.org/10.1109/TII.2019.2922823. guidance and control system for an unmanned surface vehicle. 2018 Tenth
Ntnu, A.M.O.S., 2017. Research at centre for autonomous marine operations and International Conference on Advanced Computational Intelligence (ICACI) 717–722.
systems. https://www.ntnu.edu/amos/research. (Accessed 20 September 2019). https://doi.org/10.1109/ICACI.2018.8377549.
Nyk, 2019. NYK conducts world’s first maritime autonomous surface ships trial. https Wang, N., Jin, X., Er, M.J., 2019b. A multilayer path planner for a USV under complex
://www.nyk.com/english/news/2019/20190930_01.html. (Accessed 18 September marine environments. Ocean. Eng. 184, 1–10. https://doi.org/10.1016/j.
2020). oceaneng.2019.05.017.
Nyk, 2020. NYK to participate in crewless maritime autonomous surface ship trial Wang, K., Li, J., Huang, L., Ma, R., Jiang, X., Yuan, Y., et al., 2020. A novel method for
project. https://www.nyk.com/english/news/2020/20200615_01.html. (Accessed joint optimization of the sailing route and speed considering multiple environmental
18 September 2020). factors for more energy efficient shipping. Ocean. Eng. 216, 107591 https://doi.org/
Park, J.S., Park, C., Manocha, D., 2017. Intention-Aware Motion Planning Using Learning 10.1016/j.oceaneng.2020.107591.
Based Human Motion Prediction arXiv preprint. arXiv:1608.04837. Weathernews, 2020. The world’s first crewless ship trial run. https://global.weatherne
Park, J., Choi, J., Choi, H.T., 2019. COLREGS-compliant path planning considering time- ws.com/news/13521/. (Accessed 18 October 2020).
varying trajectory uncertainty of autonomous surface vehicle. Electron. Lett. 55, Wikipedia, 2020. PID controller. https://en.wikipedia.org/wiki/PID_controller.
222–224. https://doi.org/10.1049/el.2018.6680. (Accessed 18 October 2020).
Patterson, A., Lakshmanan, A., Hovakimyan, N., 2019. Intent-aware Probabilistic Wikipedia, 2021. Global maritime distress and safety system. https://en.wikipedia.or
Trajectory Estimation for Collision Prediction with Uncertainty Quantification arXiv g/wiki/Global_Maritime_Distress_and_Safety_System. (Accessed 21 March 2021).
preprint. arXiv:1904.02765. Woo, J., Kim, N., 2020. Collision avoidance for an unmanned surface vehicle using deep
Porathe, T., Rødseth, Ø.J., 2019. Simplifying interactions between autonomous and reinforcement learning. Ocean. Eng. 199, 107001 https://doi.org/10.1016/j.
conventional ships with e-Navigation. J. Phys. Conf. 1357, 012041 https://doi.org/ oceaneng.2020.107001.
10.1088/1742-6596/1357/1/012041. Wright, R.G., 2019. Intelligent autonomous ship navigation using multi-sensor
Raptodimos, Y., Lazakis, I., et al., 2016. Ship sensors data collection and analysis for modalities. TransNav: International Journal on Marine Navigation and Safety of Sea
condition monitoring of ship structures and machinery systems. In: Proceeding of Transportation 13, 503–510. https://doi.org/10.12716/1001.13.03.03.
Smart Ship Technology, London, UK, vols. 26–27, pp. 1–11. January 2016. Xie, S., Garofano, V., et al., 2019. Model predictive ship collision avoidance based on Q-
Report Itu-R M.2371-0, 2015. Selection of the Channel Plan for a VHF Data Exchange learning beetle swarm antenna search and neural networks. Ocean. Eng. 193,
System. ITU. 106609 https://doi.org/10.1016/j.oceaneng.2019.106609.

21
X. Zhang et al. Ocean Engineering 235 (2021) 109380

Xinping, Y., 2017. “Navigation brain” intelligent navigation system. http://www.intel Zaccone, R., Martelli, M., 2018. A random sampling-based algorithm for ship path
navi.com/. (Accessed 10 January 2021). planning with obstacles. In: Proceedings of the International Ship Control Systems
Xinping, Y., Feng, M., Jialun, L., et al., 2019. Applying the Navigation Brain System to Symposium-ISCSS, Glasgow, UK, pp. 170–178. https://doi.org/10.24868/issn.2631-
Inland Ferries. Proceeding of 18th Conference on Computer and IT Applications in 8741.2018.018, 2 – 4 October 2018.
the Maritime Industries, Tullamore, pp. 156–162. Zeng, Y., Zhang, J., et al., 2020. Collision avoidance decision-making based on particle
Xue, J., Chen, Z., Papadimitriou, E., et al., 2019a. Influence of environmental factors on swarm optimization and genetic algorithm. Navigation of China 43, 1–6+28.
human-like decision-making for intelligent ship. Ocean Engineering 2019 186, Zhang, R., Chen, Y., Sun, Z., Sun, F., Xu, H., 2000. Path control of a surface ship in
106060. https://doi.org/10.1016/j.oceaneng.2019.05.042. restricted waters using sliding mode. IEEE Trans. Contr. Syst. Technol. 8, 722–732.
Xue, J., Wu, C., Chen, Z., Van Gelder, P.H.A.J.M., Yan, X., 2019b. Modeling human-like https://doi.org/10.1109/87.852916.
decision-making for inbound smart ships based on fuzzy decision trees. Expert Syst. Zhang, X., Wang, C., Liu, Y., Chen, X., 2019. Decision-making for the autonomous
Appl. 115, 172–188. https://doi.org/10.1016/j.eswa.2018.07.044. navigation of maritime autonomous surface ships based on scene division and deep
Yang, C., Liu, J.Y., Xiong, Z., et al., 2020. Brain-inspired navigation technology reinforcement learning. Sensors 19, 4055. https://doi.org/10.3390/s19184055.
integrating perception and action decision: a review and outlook. Acta Aeronautica Zhang, Q., Zhang, M., Hu, Y., Zhu, G., 2020. Error-driven-based adaptive nonlinear
Astronautica Sinica 41, 023280. https://doi.org/10.7527/S1000-6893.2019.23280 feedback control of course-keeping for ships. J. Mar. Sci. Technol. https://doi.org/
(in Chinese). 10.1007/s00773-020-00741-6.
Yara, Birkeland, 2020. Yara Birkeland status. https://www.yara.com/news-and-media/p Zhao, J., 2019. Research on Obstacle Avoidance Problem of Surface Unmanned Vehicle
ress-kits/yara-birkeland-press-kit/. (Accessed 18 October 2020). Based on Improved Genetic Algorithm. Master’s Thesis. Tianjin University of
Yoo, B., et al., 2018. Ship route optimization considering on-time arrival probability Technology, Tianjin, China.
under environmental uncertainty. In: Proceeding of OCEANS-MTS/IEEE Kobe Zhao, L., Roh, M.I., 2019. COLREGs-compliant multiship collision avoidance based on
Techno-Oceans (OTO), Kobe, Japan, 28-31 May, pp. 1–5. https://doi.org/10.1109/ deep reinforcement learning. Ocean. Eng. 191, 106436 https://doi.org/10.1016/j.
OCEANSKOBE.2018.8559308. oceaneng.2019.106436.
Yunzhou Intelligent Technology Co Ltd, 2020. Products. http://www.yunzhou-tech. Zhao-Lin, W., 1984. Quantification of action to avoid collision. J. Navig. 37, 420–430.
com/Products/detail/id/50.html. (Accessed 18 October 2020). https://doi.org/10.1017/S0373463300019949.
Zaccone, R., Figari, M., 2017. Energy efficient SHIP voyage planning BY 3D dynamic Zhou, X., Wu, P., Zhang, H., Guo, W., Liu, Y., 2019. Learn to navigate: cooperative path
programming. J. Ocean Technol. 12, 49–71. planning for unmanned surface vehicles using deep reinforcement learning. IEEE
Access 7, 165262–165278. https://doi.org/10.1109/ACCESS.2019.2953326.

22