807

1 INTRODUCTION

The Resolution MSC.467 (101) [16] was adopted in the

2019 at the 101st session of the Maritime Safety

Committee (MSC), which is part of the International

Maritime Organization (IMO). The resolution defines

and harmonizes the format and structure of existing

maritime services in the context of e-navigation. The

intense development of technology in the field of

digitization of equipment in maritime applications

begins reaching the point where it will be possible to

further continue the process of replacement of crew

members by machines [13].

There are several reasons for this process. The first

one is the constant development of industrial

technology, which translates into increased durability

and reliability of the manufactured devices. A ship

built of modern materials and components requires

less and less maintenance and the service intervals can

become longer. At the same time, the use of computer

diagnostic techniques allows for early detection of

emerging problems, which allows planning of

specialized service actions that would avert failures

during normal operation.

The second reason for replacing the crew with

machines is the increased precision and safety of the

operations performed. Thanks to this, the most

common cause of accidents, human error, is

eliminated. Years of gathering experience during the

operation of specific systems can be translated into the

language of algorithms and implemented procedures,

which mean that repetitive activities can be carried

Port Tugboat Formation Multi-Agent Control System

W. Koznowski & A. Łebkowski

Gdynia Maritime University, Gdynia, Poland

ABSTRACT: The publication presents the structure of the agent system, tasked with control of the formation of

unmanned port tugboats capable of performing pilot and towing services. The use of autonomous tugboats with

installed software was presented with respect to the existing regulations related to Resolution MSC.467

developed by the Maritime Safety Committee (MSC) belonging to the International Maritime Organization

(IMO), which creates guidelines for the definition and harmonization of the structure and format of maritime

services in the context of e-navigation. The use of a multi-agent system structure enables synergistic cooperation

of tugboats carrying out joint port operations, such as: assistance in maneuvers of ships, precision movement of

ships and other objects in port areas, monitoring and patrolling of port areas, carrying out ice operations,

carrying out inspections of quays, the possibility of assistance in liquidation of petroleum and similar pollutant

spills. The paper presents the structure of the agent system and the description of possible scenarios of port

operations. The control algorithms and the applied methods of artificial intelligence, such as evolutionary

algorithms with elements of fuzzy logic, were discussed. The recorded traffic parameters from the actions

carried out in the simulator of the marine navigation environment were presented.

http://www.transnav.eu

the International Journal

on Marine Navigation

and Safety of Sea Transportation

Volume 15

Number 4

December 2021

DOI: 10.12716/1001.15.04.13

808

out by machine systems that are much faster and more

accurate than when staffed by humans.

The ultimate goal of the ship automation process is

to achieve a level that allows the construction of fully

autonomous marine vessels, initially operating under

continuous human supervision [7], and ultimately not

requiring even this type of control.

A similar trend in the development of automation

applies to typical offshore support units, such as port

tugs, pilot boats or bunker boats. Automation of tug

operations can help to improve the safety [1] of ships

entering and exiting ports. The problems necessary to

solve in the course of the development of these type of

units are mainly related to:

− the necessity of frequent approach / departure

from the quay,

− moving through congested port waters, which

requires tight navigation,

− coexistence (initially) in an environment

dominated by manned units, the behavior of

which, from the point of view of the machine, is

somewhat unpredictable,

− the necessity of performing precise approaches to

the ship being the target of an unmanned unit,

often much larger than it and less maneuverable.

Another direction in the development of transport

systems [9], including maritime transport, is the

gradual departure from fossil fuels, in favor of

alternative propulsion systems [10], using low-carbon

fuels such as LNG [12] or hydrogen, and electricity

stored in energy storage systems or generated from

using fuel cells. The combination of an electric drive

in combination with a battery energy storage is

particularly advantageous, allowing for a high overall

efficiency of such propulsion system [8].

The development of autonomous ship technology,

including tugs, seems inevitable. So far, no operating

systems have been presented, but intensive research

has been carried out in this direction [5], and the first

tests were carried out with the use of full-size tugs by

the co-op between Rolls-Royce and Svitzer [11] and

another one by Samsung [4].

This article presents the concept of a multi-agent

formation control system for unmanned port tugs.

This system has the ability to control one or more

formations of unmanned port tugs, including tugs

using electric propulsion. One of the features of the

presented system is compliance with the COLREGs

convention.

2 MULTI-AGENT SYSTEMS

Multi-agent systems belong to the set of methods used

to search for optimal solutions, classified as artificial

intelligence methods. These systems use standalone

entities called agents that can interact with each other.

Agents are most often computer programs specialized

in the performing specific tasks. A feature of multi-

agent systems is the possibility of cooperation

between various agents. Thanks to this, the results

obtained by the multi-agent system are characterized

by properties that reach beyond the sum of the

specializations of individual agents, and have a high

flexibility of response to the changing environmental

situation. In marine applications, multi-agent systems

can be used to direct ship traffic in congested ports

[14], or inland waters [17]; during oil spill cleanup

operations [18], the evacuation of passenger ships [3],

or for the purpose of ship course keeping [15].

The agent platform is the agents’ area of operation.

It is an environment that enables mutual

communication between agents. In the case of multi-

agent systems in which the agents are physically

dispersed, platforms using communication techniques

based on computer networks are frequently used [6].

Decision priority

Figure 1. An example of a hierarchical structure of an agent

system. The votes of the agents higher in the hierarchy have

a higher weight.

The topology of connections between agents in a

multi-agent system can assume various configurations

[2]. Often employed structures of agent systems are

hierarchical topologies (figure 1), in which decisions

made by agents standing higher in the hierarchy, are

binding on subordinate agents, and team topologies

(figure 2), in which there are sets of closely

cooperating agents (teams) that can communicate with

other teams.

Team A

Team C

Team B

Figure 2. An example of an agent system structure using

teams of closely cooperating agents.

There are also indirect topologies, such as holonic

topology [2] which has a fractal structure, where each

of its components - holons - has a similar internal

structure as the environment in which the holon is

located (figure 3). Each of the holons is an

independent entity, however they may be nested

809

where the master holon contains several child holons

within it. In the holonic topology, one of the agents

inside his holon is appointed as an overriding agent

that can communicate with the environment.

The holonic structure seems to be the correct

topology to be used in the unmanned port tugboat

formation control system, due to the way it

corresponds to the classic manned tug structure

shown in figure 4. At its lowest level there is the tug

itself, which has its own the crew (blue figures) led by

the Captain (white figure). The work of the group of

tugboats providing assistance is coordinated by the

Pilot (yellow figure) on board of the assisted unit,

while the supervision of all operations taking place in

the waters controlled by the port is carried out by the

Port Authority Office.

Figure 3. Holonic structure of an agent system, with

overriding agents marked in boldface.

PORT

AUTHORITY

Figure 4. Command structure in a manned tug assistance

operations. Sailors in blue, captains in white, pilots in

yellow.

Based on the above command structure, figure 5

shows the architecture of the multi-agent formation

control system for unmanned port tugboats. The

equivalent of the crew of a single unmanned tug is a

set of agents responsible for: the operation of the tug's

propulsion mechanisms (P), the navigation situation

around the tug (N), commanded by a designated

commander agent (C), who will be able to

communicate with the environment. All of the

formation tugs report through their agents of

commanders C, under the counterpart of the Pilot –

FP agent (Formation Pilot).

The fractal nature of a single tug's holonic

command structure is reflected at a higher level where

several tugs work together in a formation under the

direction of a formation pilot. Going one step further,

the control structure of several tug formations is again

similar in nature, with several formations under the

command of the Port Authority Office.

PORT

AUTHORITY

Tug formation

Figure 5. Command structure using the multi-agent

formation control system of unmanned tugs. Command (C),

navigation (N) and propulsion (P) agents report to the

formation pilot (FP).

The task of the propulsion (P) agent is to control

and supervise the tug's propulsion mechanisms. In the

case of a conventional drive, the scope of tasks

includes: starting and stopping the drive motors,

monitoring the operating parameters of the drive

system (temperatures, pressures, levels, etc.), handling

emergency situations such as: exceeding the

permissible ranges of operating parameters, failures

disabling some of the mechanisms, unexpected leaks.

In the case of alternative drives, e.g. hybrid or fully

electric drives, there are also function of supervision

over energy storage and its periodic charging.

The duties of the navigation (N) agent include:

supervision of the navigational situation during the

tug's movement and during the performance of

precision maneuvers, determining the route of

passage if it is necessary to travel longer distances,

maintaining the position of the tug during the

movement of the tug formation in a preset formation.

The navigation agent on the leader tug has an

additional task of determining formation anti-

collision maneuvers, in the event of a collision threat

with other ships. The navigation agent has at his

disposal navigation devices available on board the

tug, including RADAR + ARPA, AIS and GPS with

satellite compass function. The precision nature of the

maneuvers performed requires a high-resolution GPS

receiver using RTK technology.

The agent C commanding the tugboat

communicates with the environment, i.e. with the

other tugboats, through their C agents and the FP

formation pilot agent. The FP formation pilot

provides him the tasks to be performed, e.g. the task

of sailing the route from the current position to the

area of the given ship as a leader, or as one of the

other tugboats of the formation.

810

2.1 Possible tug formation application scenarios

The presented multi-agent system for controlling the

formation of port tugboats allows performing of

operations typical for tugs, such as assisting ships

during maneuvers (entering and leaving the port,

mooring and unmooring) or the precision movement

of ships and other objects in port waters.

The possibility of equipping the tugs with

additional specialized instruments, the formation

control capabilities allow, among others:

− monitoring and patrolling of port areas - when

equipping tugs with vision systems,

− carrying out waterway deicing actions - assuming

that the hulls of the tugs have the appropriate ice

class,

− carrying out inspections of quays - when tug boats

are equipped with appropriate vision systems and

sensors,

− automated monitoring of the depth of port basins

and fairways - with the implementation of a

connection of the agent system with the tug's

onboard echo sounder,

− assisting in the cleanup of spills of petroleum

substances and other pollutants - with the use of

specialized equipment, e.g. scoops or by using the

tugs to erect floating barriers.

c) d)

Figure 6. Formation shapes used for tug formation

movements. a) line formation, b) echelon formation, c) V

formation, d) column formation. Formation leader shown in

green, other members in red. Formation follows the

direction of leader (shown with arrow).

By using appropriate formation shapes, it is

possible to perform various additional tasks during

the movement of the tug formation. Possible

formations are shown in figure 6. The possibility of

monitoring the depth of waters in the port area,

carried out during the movement of the tugboats,

seems to be extremely interesting. For this task, the

column formation (figure 6d) is particularly useful, in

which the echo sounder of each tugboat can collect

information about a slightly different fragment of the

bottom than, for example, in the case of a line

formation (figure 6a).

3 PRELIMINARY TESTING OF PORT TUGBOAT

FORMATION MULTI-AGENT CONTROL

SYSTEM

To verify the work of the algorithms of the multi-

agent port tugboat formation control system, the

proprietary navigation environment simulator was

used. This simulator makes it possible to study the

behavior of ships in a common simulated

environment, and to visualize the test results in a 3D

view and in the form of result files.

Using network technologies, each holon

representing one tugboat, and containing three agents:

command (C), navigation (N) and propulsion (P) was

launched on one of the simulator's computers. The

master computer controlling the simulation acted as

the formation pilot (FP), giving orders to individual

tugs.

The simulation was carried out, covering 3

successive stages of the tug formation job: creation of

the formation from the group of tugs waiting at

berths, the transition of the formation of tugs in the set

formation shape from the rally point to the vicinity of

the vessel waiting for assistance, and finally

surrounding of the assisted vessel by the formation

tugs.

The first stage of the job was to create a formation

using the required number of tugs. In the case of the

simulated maneuver, it was assumed that the

formation will consist of 4 identical tugs. It was

chosen to employ the linear formation, using one of

the tugs as the leader of the formation. The leader was

selected by checking which of the four available tugs

was closest to the vessel waiting for assistance.

Leader

Tug 2

Tug 4

Tug 3

> R

Figure 7. The principle of determining the order of tugs in a

linear formation.

Tug 3

Leader

Tug 2

Tug 4

Rally point

waypoint

Figure 8. The point at which the assembly of a linear

formation begins, when the leader reaches the rally point

811

and the rest of the formation members are within a set

distance from it.

After selecting the leader, he takes over the task of

laying out the route to the vicinity of the ship waiting

for assistance. This task is the responsibility of his

navigational agent (N), which defines the route as a

list of waypoints, the first of which is in the vicinity of

the available tugs and is also the rally point of the

formation. The last of the waypoints lies in the

vicinity of the ship waiting for assistance.

The formation leader sends information about the

location of the rallying point to the other tugboats, at

the same time engages his tug’s propulsion and

proceeds towards it at a speed of 80% of the speed

defined as the formation movement speed Vf. The

remaining tugs determine their distance Rt from the

rally point as shown in figure 7, based on the

inequality:

1Tn Tn



(1)

where:

RTn – distance of n-th tugboat from the rally point,

RTn+1 – distance of (n+1)-th tugboat from the rally

point.

Then, the process of forming a line formation is

initiated, where each tug boat moves towards the rally

point at such a speed, that when the leader reaches

this point, each of the tugs is at a predetermined

distance Ln defined by the equation:

( )

L n D= − 

(2)

where:

Ln – Distance of n-th tugboat from the leader,

n – number of the tug, n leader = 1,

Df – distance between neighboring tugboats in the line

formation.

When the formation leader reaches the rally point,

he starts to turn towards the next waypoint on the

computed path. The remaining tugs, already at

appropriate distances Ln from the assembly point,

move towards it, simultaneously adjusting their speed

to the same as leader's speed, as shown in figure 8.

Then each of the tugs, reaching the assembly point,

starts a turn, from then on following the tug

preceding them, keeping a defined distance Df. The

assumed distance is selected in such a way that in the

event of a failure on the preceding vessel, it would be

possible to perform an avoidance maneuver. Another

aspect taken into account when determining the value

of Df is the amount of hydrodynamic resistance that

could affect the hull of the tug if it was moving in the

water stream of the preceding vessel. When the last

tug, after reaching the rally point, positions itself at a

distance Df from the penultimate tug, the line

formation creation process is completed and the

leader along with the other tugs accelerates to the set

speed Vf.

After assembling the formation, it moves to the

place of assistance operation. The journey is divided

into phases of: leaving the port, and then the phase of

straight passage towards the waiting ship, during

both respecting the right of way. During the

movement phase, the tugboats maintain the speed Vf

equal to 8.5 knots and the mutual distance Df,

therefore, for the purposes of anti-collision

calculations, the formation length LF can be assumed

according to the equation:

( )

F H f

L L k D= + − 

(3)

where:

LF – Total length of line formation [m],

LH – Length of one tugboat [m],

k – Number of tugboats in the formation.

Table 1 shows the parameter values used for

simulation in the navigation environment simulator.

Both the number of tugs k and the distance between

the tugs Df have a direct impact on the operation of

the formation, especially in congested waters, because

for the purposes of performing anti-collision

maneuvers, the whole formation is treated as single

object.

Table 1 Parameters used during simulation.

_______________________________________________

Parameter Symbol Value

_______________________________________________

Number of tugboats in the formation k 4

Tugboat length LH 25m

Mutual distance between tugboats Df 150m

Line formation length LF 475m

_______________________________________________

Figure 9 shows the data view from the navigation

environment simulator, showing the trajectories of the

simulated vessels: four tugs, one vessel waiting for

assistance and one other vessel, placed against the

map of the Port of Gdynia. The route of the other ship

crossing from the southern part of the port to the

north was deliberately synchronized with the route of

the formation passage in order to create a collision

situation with the tugs of the formation.

Tug 1

Tug 2

Tug 3

Tug 4

Other ship

Target

vessel

A-C

Maneuver

Figure 9. The paths of simulated vessels against the map of

the Port of Gdynia. Visible assembly of the line formation,

sailing in formation in the port area, performing an anti-

collision maneuver after leaving the port heads, and

reaching the destination next to the ship waiting at port

road.

812

Rally

point

Port

exit

Target

vessel

Figure 10. Stages of line formation assembly in the Port of

Gdynia. Visible features: rally point, path in the port

(circles), anti-collision maneuver waypoints (triangles), final

waypoint (square), and unperturbed transition path

(dashed line).

Figure 9 shows the reaction of the formation leader

who, after identifying the threat of a collision with the

other ship coming from starboard, made a starboard

turn in order to pass behind the stern of the other

ship. The determination of the anti-collision maneuver

in accordance with COLREGs rules resulted in the

modification of the formation transition route by

adding additional waypoints, directing the formation

to the route leading behind the other ship. Additional

waypoints are shown in figure 10 as triangular

markers, the actual route as a solid line, and the

originally planned route as a dashed line.

The first additional waypoint lies on the original

route from the port exit area to the point in the

vicinity of the waiting vessel. This waypoint is created

at the place where the leader of the tugboat formation

starts the turn that marks the beginning of the anti-

collision maneuver. Subsequent tugs reaching this

point follow the leader's footsteps, also making a turn,

heading to the second waypoint located in such a way

that the entire formation could safely bypass the other

ship, maintaining a linear formation at all times. After

the tug boats have reached the second additional

waypoint, the route continues to the original

destination, located in the vicinity of the vessel

awaiting assistance, marked with a square in Figure

10.

After individual tugs reach the destination point

next to the vessel awaiting assistance, the line

formation is gradually disbanded and the tugs are

directed to their designated work stations around the

vessel. The trajectories of the tugs at this stage of the

formation's operation are shown in figure 11.

Tug 1

(leader)

Tug 2

Tug 3

Tug 4

Figure 11. The phase of surrounding the target ship by the

formation member tugs. Visible final formation waypoint

marked with a square symbol, the silhouette of the ship

awaiting assistance, and the imaginary bow-stern line of the

awaiting ship.

The arrangement and routes of the tugs around the

target ship, shown in figure 11, required the tugs 1

and 2 to perform additional maneuvers in order to

safely reach the opposite side of the ship than the

direction from which the formation came. The

maneuver consisted of avoiding the stern and the bow

of the target vessel at a safe distance, and then going

to assigned positions. Due to the favorable location of

the designated work stations in relation to the bow-

stern line of the target vessel (not needing to cross this

line), the tugboats 3 and 4 could proceed to their

positions immediately after reaching the final

formation waypoint.

An additional condition considered while

determining the placement of tugs around the assisted

vessel was to assign the places furthest from the final

formation waypoint to the first two tugs. This made it

possible to remove the need to avoid the tugs already

in the required positions if the assisted vessel was

positioned exactly with its stern or bow towards the

arrival direction of the tugs. Figure 12 shows a 3D

view of the final mutual position of the tugs and the

assisted vessel.

Tug 1

Tug 2

Tug 3

Tug 4

Figure 12. 3D view from the navigational simulator,

showing the situation around the target vessel at the end of

the target surrounding phase.

Figure 13 shows the plot of changes in wind speed

during the simulation. Figure 14 presents the course

and rudder angle of the leader tugboat, tug 1, during

the course of the simulation. The leader tug’s speed

plot has been shown in the figure 15. The visible dips

in the speed are associated with the turns while

underway. The initial 80% setpoint of Vf speed is seen

in the beginning of the plot. It is intended to allow the

rest of the tugs to settle behind the leader as the

formation was being assembled.

813

Figure 13. Wind speed plot in a simulated navigation

environment simulator environment.

Figure 14. Plot of the course and the angle of the rudder of

tug 1 - the leader of the formation.

Figure 15. Speed plot of tug 1 - formation leader.

4 CONCLUSIONS

The article presents the concept of a multi-agent traffic

control system for unmanned port tugboats. The

trends in the development of modern ship traffic

control systems are presented. A holonic agent

connection structure is described, corresponding to

the traditional manned tug command system. The

cooperation of agents placed on individual tugs

allows performing complex operations, such as

creating specific formations, moving in formation

along a specific route or a precise approach to the ship

waiting for assistance.

The results of simulation studies confirm the

possibility of controlling the formation of autonomous

tugs with the use of an agent system, including the

controlled assembly and disbanding of formations

with a specific shape, and the possibility of

commanding the passage of formation in a specific

shape along a given route. It is possible that the

example shown using a linear formation shape would

have limited applicability in congested waters and

with larger number of tugs. Further research could

consist in testing other formations, e.g. multi-column

formations or employing dynamically changing

distance between tugs. Reducing the mutual distance

would shorten the formation length, but would

require a reduction in speed. A lower speed would

extend the transition time, but could also facilitate the

performance of anti-collision maneuvers.

Additionally, the possibility of reacting to

unforeseen disturbances on the route in the form of

other ships on a collision course with the formation

was successfully tested. The sailing formation under

the command of the leader made a controlled turn

maneuver, in accordance with the rules of the right of

way of the sea, in order to avoid the other vessel

having the right of way. After performing the anti-

collision maneuver, the formation correctly returned

to the interrupted task. After reaching the vicinity of

the target vessel, the formation orderly disbanded and

the tugs took up positions around the vessel, ready for

the assist. The use of a multi-agent system may

contribute to an increase in the level of safety in a

given body of water and a reduction of fuel, which

translates into a reduction in exhaust emissions to the

atmosphere.

ACKNOWLEDGEMENT

This research was funded by a research project of the

Electrical Engineering Faculty, Gdynia Maritime University,

Poland, WE/2021/PZ/08 titled “Optimization of low-

emission surface vessel control processes.”.

REFERENCES

1. Abramowicz-Gerigk, T., Burciu, Z.: Application of Ship

Motion Simulation in Reliability Assessment of Ship

Entrance into the Port. TransNav, the International

Journal on Marine Navigation and Safety of Sea

Transportation. 10, 4, 613–617 (2016).

https://doi.org/10.12716/1001.10.04.10.

2. Balaji, P.G., Srinivasan, D.: An Introduction to Multi-

Agent Systems. In: Srinivasan, D. and Jain, L.C. (eds.)

Innovations in Multi-Agent Systems and Applications -

1. pp. 1–27 Springer Berlin Heidelberg, Berlin,

Heidelberg (2010). https://doi.org/10.1007/978-3-642-

14435-6_1.

3. Couasnon, P., de Magnienville, Q., Wang, T., Claramunt,

C.: A Multi-agent System for the Simulation of Ship

Evacuation. In: Kawai, Y., Storandt, S., and Sumiya, K.

(eds.) Web and Wireless Geographical Information

Systems. pp. 63–74 Springer International Publishing,

Cham (2019).

4. Demonstration of Autonomous and Remote-Controlled

Ship Operations: https://www.maritime-

executive.com/article/demonstration-of-autonomous-

and-remote-controlled-ship-operations, last accessed

2021/05/17.

814

5. Du, Z., Reppa, V., Negenborn, R.R.: Cooperative Control

of Autonomous Tugs for Ship Towing. IFAC-

PapersOnLine. 53, 2, 14470–14475 (2020).

https://doi.org/10.1016/j.ifacol.2020.12.1448.

6. Kravari, K., Bassiliades, N.: A Survey of Agent

Platforms. Journal of Artificial Societies and Social

Simulation. 18, 1, 11 (2015).

https://doi.org/10.18564/jasss.2661.

7. Neumann, T.: Telematic Support in Improving Safety of

Maritime Transport. TransNav, the International Journal

on Marine Navigation and Safety of Sea Transportation.

12, 2, 231–235 (2018).

https://doi.org/10.12716/1001.12.02.02.

8. Nuchturee, C., Li, T., Xia, H.: Energy efficiency of

integrated electric propulsion for ships – A review.

Renewable and Sustainable Energy Reviews. 134, 110145

(2020). https://doi.org/10.1016/j.rser.2020.110145.

9. Przybylowski, A.: Global Trends Shaping Life Quality in

Agglomerations with Particular Emphasis on Mobility in

Seaport Agglomerations. TransNav, the International

Journal on Marine Navigation and Safety of Sea

Transportation. 13, 3, 615–620 (2019).

https://doi.org/10.12716/1001.13.03.18.

10. Przybylowski, A.: Sustainable Transport Planning &

Development in the EU at the Example of the Polish

Coastal Region Pomorskie. TransNav, the International

Journal on Marine Navigation and Safety of Sea

Transportation. 6, 2, 277–286 (2012).

11. Rolls-Royce demonstrates world’s first remotely

operated commercial vessel: https://www.rolls-

royce.com/media/press-releases/2017/20-06-2017-rr-

demonstrates-worlds-first-remotely-operated-

commercial-vessel.aspx, last accessed 2021/05/17.

12. Rutkowski, G.: Study of New Generation LNG Duel Fuel

Marine Propulsion Green Technologies. TransNav, the

International Journal on Marine Navigation and Safety

of Sea Transportation. 10, 4, 641–645 (2016).

https://doi.org/10.12716/1001.10.04.14.

13. Rybczak, M., Rak, A.: Prototyping and Simulation

Environment of Ship Motion Control System. TransNav,

the International Journal on Marine Navigation and

Safety of Sea Transportation. 14, 2, 367–374 (2020).

https://doi.org/10.12716/1001.14.02.13.

14. Singh, A.J., Nguyen, D.T., Kumar, A., Lau, H.C.:

Multiagent Decision Making For Maritime Traffic

Management. AAAI. 33, 01, 6171–6178 (2019).

https://doi.org/10.1609/aaai.v33i01.33016171.

15. Wang, C., Yan, C., Liu, Z.: Leader-following Consensus

for Second-order Nonlinear Multi-agent Systems Under

Markovian Switching Topologies with Application to

Ship Course-keeping. International Journal of Control,

Automation and Systems. 19, 1, 54–62 (2021).

https://doi.org/10.1007/s12555-019-0395-8.

16. Weintrit, A.: Initial Description of Pilotage and Tug

Services in the Context of e-Navigation. Journal of

Marine Science and Engineering. 8, 2, (2020).

https://doi.org/10.3390/jmse8020116.

17. Xu, W., Wu, S.: Ship Agent model for traffic flow

simulation in inland waterway. IOP Conf. Ser.: Mater.

Sci. Eng. 768, 7, 072104 (2020).

https://doi.org/10.1088/1757-899X/768/7/072104.

18. Ye, X., Chen, B., Li, P., Jing, L., Zeng, G.: A simulation-

based multi-agent particle swarm optimization

approach for supporting dynamic decision making in

marine oil spill responses. Ocean & Coastal

Management. 172, 128–136 (2019).

https://doi.org/10.1016/j.ocecoaman.2019.02.003.