85
1 INTRODUCTION
1.1 IntelligentNavigationSystem
TheIntelligentNavigationSystemsisregardedasthe
next generation of main auxiliary driving means.
With the benefits of the improvement of computer
technology,navigationtechnology, and sensors, also
including intelligent arithmetic and methods, these
systems come into practice. The accuracy of the
position, velocity, acceleration and heading
components of the ship is an import
ant part of the
navigation system to predict ship’s location and
heading, as well as an important element of
improvingthesafetyofnavigation.Itshouldbenoted
that the progress of modern technology has been
widely used in a variety of land (Uriasz 2009, Jian,
Wenjunetal.2014),waterway(Zhonglian,Xiuminet
al. 2015) and air (Ruishan, Lei et al. 2016) tra
nsport
systemstomeetthecomplicatednavigationneedsin
recent years. Waterway transport is still in the
progress of these technologies and a considerable
number of navigation functions (ie navigation
prediction, propulsion control and engine control
(Dongzhi, Xinping et al. 2014)) are st
ill operated by
human operators. These lacking navigation features
canalsocompromisetheirrespectiveoperationswith
safeinnarrowandcrowdedwatersandcomplexship
handling at different environment conditions.
Although in the above complex navigation
environment, the crew can st
ill ensure the safety of
navigation,butthedifficultyisstillincreasedwiththe
heighten of the intensity of the ship. The auxiliary
drivingsystems,alsootherAGNSystemsareableto
solve this problem effectively, so as to reduce the
burden of ship operators and reduce the risk of
accidentscausedbyhumanerror.
Ship Domain Model for Multi-ship Collision A
v
oidance
Decision-making with COLREGs Based on Artificial
Potential Field
T.F.Wang,X.PYan&Y.Wang
WuhanUniversityofTechnology,Wuhan,China
NationalEngineeringResearchCenterforWaterTransportSafety,Wuhan,China
Q.Wu
SchoolofLogisticsEngineering,WuhanUniversityofTechnology,Wuhan,China
ABSTRACT:Amulti‐shipcollisionavoidancedecision‐makingandpathplanningformulationisstudiedina
distributed way. This paper proposes a complete set of solutions for multi‐ship collision avoidance in
intelligentnavigation,byusingatop‐to‐bottomorganizationtostructurethesystem.Thesystemisdesigned
with two layers: the collision av
oidance decision‐making and the path planning. Under the general
requirementsoftheInternationalRegulationsforPreventingCollisionsatSea(COLREGs),theperformanceof
distributedpathplanningdecision‐makingforanti‐collisionisanalyzedforbothgive‐wayandstand‐onships
situations,includingtheemergencyact
ionstakenbythestand‐onshipincaseofthegive‐wayship’sfaultof
collision avoidance measures. The Artificial Potential Field method(APF) is used for the path planning in
details.ThedevelopedAPFmethodcombinedwiththemodelofshipdomainta
kesthetargetships’speedand
coursein‐toaccount,sothatitcanjudgethemovingcharacteristicsofobstaclesmoreaccurately.Simulation
resultsindicatethatthesystemproposedcanworkeffectiveness.
http://www.transnav.eu
the International Journal
on Marine Navigation
and Safety of Sea Transportation
Volume 11
Number 1
March 2017
DOI:10.12716/1001.11.01.09
86
1.2 CollisionavoidanceandTheCOLREGs
The mathematical model for detecting the risk of
collisionsbetweenshipsintheparticularoceanareais
divided into two main categories: distance closest
point approach (CPA) and predicted area of danger
(PAD) approach. The CPA approach evaluates the
riskofcollisionbetweenthe
shipandthesurrounding
shipbycalculatingtheminimumcollisiondistanceof
thetwovesselsunderthecurrentroute(Zhang,Zhang
et al. 2015). However, since the difference in vessel
size, heading and speed is not considered, in some
casesthe risk of collision isestimated to differfrom
theactual
situation(Stahlberg,Goerlandtetal.2012).
Therefore, this method is often used in conjunction
with the concept of the ship domain. The PAD
approachevaluatestheriskofcollision by modeling
the predicted track of own ship as an inverted cone
andothertargetships’predictedtrackasaninverted
cylinder.
The overlappingareaof these twopossible
trajectories is the area where there is a risk of
collision, and the limited size, heading and velocity
variations between the vessels can be incorporated
into this method. These two methods are both
considered the ship speed and course conditions of
continuous change,
however the instantaneous
variations of navigation information for example
which caused by the parameter changes are not
considered. In addition, in order to simplified
calculationandsoon,thesemethodsaresimplifiedby
assuming the simple and ideal navigational
conditionssuchasvesselmovementunderastraight
line of deterministic state
and parameter behavior.
Obviously, the complex and volatile actual sailing
situationisquitedifferentfromtheidealstate.
Ontheotherhand,inthecurrentalgorithms,most
ofthe rulesandregulationsontheocean navigation
collisionsituationhavebeenignored.Inpractice,the
phenomenon of neglect of IMO regulations
has also
occurredsometimes,whichhasbecomeanimportant
reason for maritime traffic accidents. Statheros
(Statheros 2008) noted that about 56% of the shipʹs
collisionaccidentswereduetothefactthatthecrew
didnotfollowtherulesandregulations.Amongthese
rulesandregulations,themostimportant one
about
anti‐collisionaretheConventionontheInternational
Regulations for Preventing Collisions at Sea
(COLREGs) (IMO (1972)).The COLREGs rules and
regulations are important documents for IMO to
constrainthebehaviorofshipssailingonthesea.Itis
divided into 5 parts (General, Steering and Sailing,
Lights and Shapes,
Sound and Light signals and
Exemptions) and four Annexes containing technical
requirements. Perera (Perera, Carvalho et al. 2011)
discussed the details of anti‐collision of two ships
meeting in a particular area under the COLREGs.
Furthermore, he presented a method of fuzzy logic
reasoning,whichcanbeusedtoassistcrew
tomake
theshipcollisionavoidancedecision.Eventhoughthe
COLREGs rules and regulations give priority to all
the sailing ships’ obedience to prevent collision
accident, they do not provide certain operating
instructions,especially for the scenarioof multi‐ship
encountering.
COLREGs rules and regulations divide the ships
meetin
awaterareaintotwokinds:thestand‐onship
andthegive‐wayship.Thestand‐onshipistheship
which should get through this area as soon as
possiblebykeepingitsspeedandcourse.Meanwhile,
thegive‐wayshipshouldchangeitsspeedandcourse
in order
to clear this area for the convenience of
stand‐on ship’s pass way. There are 3 kinds of the
ship meeting scenarios, which are the head‐on, take
over and crossing. COLREGs rules and regulations
have been discussed in the recent literatures, for
examplePerera(Perera,Carvalhoetal.2011)
showed
ever kind of the anti‐collision situations, and Jinfen
Zhang(Zhang,Zhangetal.2015)discussedthemulti‐
shipencountercollisionavoidancesituations.
Ship’s automatic path planning, which plays an
importantroleinautonomousvoyage,isthe baseof
mostAGNfunctionslikeauxiliarydriving,intelligent
navigation, unmanned surface vehicle
and so on.
However, most ship’s automatic path planning
researches draw on some research methods in the
field of robots especially self‐driving vehicles
(Lingling and Lei 2014). Due to the working
environment and the relevant traffic laws and
regulationsarecompletelydifferent,ship’sautomatic
path planning must have something
special and
unique.It’sarequirementofCOLREGstobefurther
studied to cope with new problems like auxiliary
drivingorassistmakingdecisionsinautomaticway.
1.3 ShipDomainandArtificialPotentialField
In the study of ship collision problems, it is an
importantissuetoensurethattheshortest
distanceof
the collision does not occur or the nearest distance
thattwoshipscanpasseachothersafely,whicharea
ismanedshipdomainsinceitwas firstproposedin
the 1970s. As the first question to make sure for
everyone’s study of anti‐collision, how to determine
the
shape and size of ship domain attracts
researchers’ attention. It is obviously that different
definitions and proposals of ship domain have
different shapes and sizes. At beginning of the
research,peopletendtofocusmainlyonthesizeand
shape under standard conditions. People The
standardconditions alwaysmeanalarge
ocean area
andlimitedspeed.Theshapeofshipdomainissetto
a certain shape and the size of ship domain is
considered to be almost constant in a voyage. In
furtherresearch,Peopleattemptatusingaregionof
theshipʹshistoricaldata(liketheAIS data
(Hansen,
Jensenetal.2013))inaspecificseaareatoextractthe
specific information of ship domain. Since ship
domainisakindofimagedescriptionofshipcollision
risk,itisreasonabletobechangedwithdifferentship
speed and other navigation data. Pietrzykowski
(Pietrzykowski and Uriasz 2009) defined a
variable
shipdomainwithdifferentlevelsofsafegrowthand
crewʹsconsciousness.Forexample,theshipdomainis
smaller where the navigator is familiar with the
waters,sinceit’ssaferthanotherscenarios.
The potential field method or the Artificial
PotentialFieldmethod(APF)isaclassicalmethodfor
robot
path planning, and the unmanned surface
vehiclesareregardedasthekindofrobotworkingin
river, lake or sea. A typical application of artificial
potentialfieldtracingalgorithmisshowninFigure1‐
4. The APF method uses virtual gravitational and
repulsive field force to express the relationship
between the
tracing robot and the obstacle and the
87
target.Ingeneral,therelationshipbetweenthetracing
robot and the target is a mutual attraction, and the
valueofthe virtual attractivefielddeclines with the
distance, where the potential energy is
at
U
(Fig.1).
The situation between the tracing robot and the
obstacle is the opposite, the value of the virtual
repulsive field energy increases with the distance,
where the potential energy is
re
U
(Fig.1). Then the
totalpotentialenergyineverypointoftheareaisthe
sumoftheabove:
at re
UU U
(1)
Through analogical basic mechanics knowledge,
the virtual force field can be derived to the virtual
attractiveforceandvirtua l repulsiveforceasfollows:
at re
FF F
(2)
Wherethevirtualattractiveforceis
()
at at
FgradU
,
andthevirtualrepulsiveforceis
()
re re
FgradU
.
By following the total virtual force at any given
position,thepathcanbefound.Theroute plannedby
this method in the area is a set of minimum virtual
force field values, and thatʹs one way to look at it.
(Fig.2 and Fig.3). As its mathematical expression is
elegance and simplicity, the APF method is
particularlyappealing.WhentheAPFmethodisused
inadynamicenvironmentlikemovingobstacles,the
tracing robot have to keep running path planning
algorithmtoavoidalltheobstaclesandthenthepath
planningalgorithmdevelopsintoroutefinding.
Figure1.Thegravityfieldofthetarget(x=10,y=10)andthe
repulsivefieldof7obstacles
Figure2.ThefinalAPFcombinedthegravityfieldwiththe
repulsivefield
Figure3. The final path planned from Starting Point
(x=0,y=0)totheTargetPoint(x=10,y=10)
The intelligent surface vehicle path planning in
calmwaterissimilartothetracingrobot,buttheflow
field in the environment is also a factor to be
considered.RuiSongeta(Liu,Liuetal.2015,Rui,Liu
et al. 2015, Rui, Liu et al. 2016) developed a multi‐
layered
fastmarching(MFM)methodforunmanned
surfacevehiclepathplaninadynamicenvironment.
In this method, they get the flow field in the
environment,thevelocityfieldofthetargetshipsand
thevirtualforcefieldofthe fixedobstaclestogether.
The path that planned use APF method is
a
continuouscurveofadefinitenumericalcomposition.
However, there is a great difference between the
actualuserequirementswhenitcometothesituation
likeautonomousnavigationandauxiliarydriving,as
the driver cannot steer the ship as accurate as the
robot. To solve this problem, Lee(Lee, Kwon et al.
2004) combined fuzzy logic with APF method and
developedanewautonomousnavigationalgorithm.
2 COLLISIONAVOIDANCEUNDERCOLREGS
2.1 TheCOLREGsrulesandregulations
TheCOLRREGsrulesandregulationsinclude5parts
and38rules.Amongtheserulesandregulations,this
paper mainly focus the Part B Steering and
Sailing
rules. According to the COLREGs, the collision
situationamongtwoshipscanbedividedintohead‐
on,crossingandovertaketotherouteangle.Andthe
own ship should needs to give way to all the ships
thatappearonitsstarboardside,anditisnotastand
on ship until all ships are on the port side. All the
collision situations are briefly described in Fig.4. In
Fig.4, own ship’s position is supposed to be at the
coordinate origin and its course is set to Y‐axis
forward. Then we can define the collision situation
whenthe
targetshipappearsindifferentregionsfrom
Ato F.Whenthetargetshipappearsinregion A,B
andF,itmeansthatthetargetshipisonthestarboard
side, own ship should give its way. Otherwise own
shipisastandonship.
88
Figure4.Briefdescriptionofthecollisionsituations
2.2 TheprepareofPathPlanning
Thefollowing algorithmin2.3‐2.5 mainly references
Jinfen Zhang’s (Zhang, Zhang et al. 2015) research
about the distributed anti‐collision decision support
formulation. As the system in this paper regard the
COLREGs as the top layer and prepare of the path
planning, it simplifies
the output of the Jinfen’s
algorithm.
2.3 ClosestPointofApproach(CPA)
CPAisthemostsimpleandeffectivemeanstopredict
the target ships’ position and to estimate collision
risk. In this paper, CPA is calculated every time to
makesurethatthepathplanningdecisionissafeand
the
shipshavelargeenoughareatomovenomatter
whethershipschangetheirspeedsandcourses.That
means that during the ships take anti‐collision
actions,theCPAshouldkeeplargerthansetminimal
value. For instance, ship1 and ship2 take actions to
avoidcollision,whichisshowninFig.6.Suppose
with
thesametimestarttochangecourse,theship1spend
more time to change course than ship 2 (which
means
1S
T
>
2S
T
).Inthewholeprocessofcompletingthe
collision avoidance of the two ships, there are 3
periodandeachperiodhasitsownCPA.whichislist
below,thesituationisshowninFig.5:
1S
T
1S
T
2S
T
2S
T
s
tar t
T
s
tar t
T
Figure5.CPAcomputationfortwoships
1
1
d
(from
s
tart
T
to
2S
T
):fromstarttotheendofship2
course changing, during this period both of two
shipsareonthechangingcourse;
2
2
d
(from
2S
T
to
1S
T
): from the end of ship2 course
changing to the end of ship1 course changing,
during this period ship1 is still on the changing
course,whileship2isontheoriginalcourse;
3
3
d
(from
1S
T
the end):from the end of ship1 course
changing to a time large enough, during this
periodbothoftwoshipsareontheoriginalcourse.
TheCPAofthistwoships’anti‐collisionactionis
theminimumofthethree.AndthevalueofCPAcan
becalculatedasbelow:
Suppose
thetwomovingships’initialpositionare
111
(, )
S
Pxy
and
222
(, )
S
Pxy
, whose speed
are
1S
v
and
2S
v
,andcourseanglesare
1S
and
2S
.After
ttimethepositionforship
i
is:
SS S S
() (), () sin , cos
Si i i i i i i i i
Pt xtyt x vt y vt
(3)
Duringtheperiod
(0, )T
,thedistanceoftwoships
attime
t
is
22
2
12 1 2 1 2
() () () () () () ()
D
tPtPt xtxt ytyt AtBtC
(4)
where
22
112 2 1 12 2
sin sin cos cosAv v v v
22
121 12 2 1 21 12 2
2 sin sin cos cosBxxv v yyv v
22
12 12
Cxx yy
Theminimumof
()Dt
isatypicalExtremeValue
ProblemofQuadraticFunction.
It should be noted that CPA is the basis of path
planning and the main means of decision‐making
verification.
2.4 CollisionAvoidanceforGive‐WayShips
Thispartofthealgorithmisdesignedformakingsure
thatwhether own ship
isstand‐on ship or give‐way
ship,andthencalculatesthecoursechangerangeand
thespeedchangerangeofthegive‐wayshipandthe
stand‐on ship if necessary. According to the
COLREGsandthemainrequirementsofthesystem,
thetwomainpointsofthealgorithmjust
mentioned
areasfollows.
1 Judgmentofgive‐wayshipandstand‐onship
AsshowninFig.6,theownshipshouldgiveway
to all the target ships apparent on own ship’s
board side, otherwise if there is no ship on the
board side of the own ship, the
own ship is the
standonship.
2 Thespeedchangerangeofthegive‐wayship
Due to the underactivity of the ship’s dynamic
system,althoughthecollisionavoidanceoperation
is rarely performed by changing the speed, it is
still effective in some specific cases. Especially
when the cruise of
the two ships is very small,
which is discussed by perera and Jinfen. Speed
changingisonlyaccordedwhenthecourseangle
is smaller than
, which is a given value. Speed
declines10%oftheoriginalspeedeach time and
updatetheCPAbyusingthemethoddiscussedin
3.2.1. Since the speed change often matches the
track change, the setspeedisreduced by 50% of
theoriginalspeed.Theabovetwostepsare
shown
inFig.6.
89
3 Thecoursechangerangeofthegive‐wayship
This part of the algorithm is the main part, and
focusonthechangerangeofthecourseangleand
thecoursetime.Thecoursechangetimeandangle
changefromtheminimumtothemaximum.Each
stepofthe
changeupdatethevalueofCPAuntil
theCPAislargerthanthegivenvalue.Thispartof
thealgorithmisshowninFig.7.
Figure6.Theprocedureofownshipgive‐waydecision
2.5 CollisionAvoidanceforStand‐OnShips
AccordingtotheCOLREGs,thestandonshipshould
keep its speed and course to pass the meeting
situation,atthebasesofsafety.However,theproblem
is much complex in the real world, sometimes the
stand on ship cannot tackle the problem without
changecourseaswellasthespeed.Forinstance,the
targetshipmaynotbeabletogiveitswayintime.By
studythesituationthatthestandonshiphastotake
measures,itisobviouslythatthemeasures thatstand
onshiptakeisontheopposite
ofthegivewayship.It
means that the stand on ship should turn left and
acceleratecomparedwiththegivewayship’sturning
right and slow down. Since the COLREGs
requirementthatin the processofavoidingcollision
thegivewaysshiphastotakethemeasuresasmuch
as
possible from the stand boatʹs stern through the
formation of the port side of the port side of the
situation.Inordertoavoidaconflictwiththeaction
of give way ship, the action of the stand on ship
should be facilitated by the bow from the give
way
ship.Soitis the reasonable for the stand on ship to
apply the above collision avoidance strategy. Based
on the above analysis, the anti‐collision decision for
stand on ship can be made in the same way of the
give way ship. The only different is turn left and
accelerate.
Figure7.Theownshipchangecoursepartisshown
3 PATHPLANINGUSINGAPFMETHOD
3.1 theModelofthePotentialField
APF method will be used in this part, in order to
contributethepathplanningindetails.Thepurpose
of building the target shipsʹ potential field is to
describe the impactof the target ships. Through the
90
establishment of the target ship potential field, the
ownshipandthetargetshipcankeepasafedistance
by altering the course and the speed. For the target
ship,the riskofthevicinityofthetargetshipinthe
vertical and horizontal distribution along the target
ship is
not uniform, taking the characteristics of the
ship and the actual situation of navigation. For
example,themainfactoraffectingthesafetyofaship
inthehorizontaldirectionisthedistancebetweenthe
ships and the speed is also an important factor in
additiontothedistance.Therefore,the
modellingof
the target ship potential field will also be combined
with the vertical and horizontal characteristics of
different models of design. In this paper, the main
innovation of the target ship potential field is the
model used the ship’s longitudinal influence as the
skeleton,to extendthewaytoachieve
the impactof
theshipdistributionofthedescription.
The longitudinal potential of the target ship is
calculatedas follows. The local coordinate system is
established based on the direction of the target ship
body direction and the intersection of the shipʹs
horizontal axis and the vertical axis. The
entire
longitudinalpotentialfielddistributionisapiecewise
function,thecalculationformulaasfollows:
ship
r
ship r
r
Up
v
Apv
KS
v
()(0)
0(0)
Ι
(5)
where,
s
hip
U
is the ship potential constant, which represents
themaximumvalueoftheshipʹspotentialfield
r
v
is the relative speed of the own ship and the
target ship, when the speed of the two in the same
direction, the speed of the own ship is greater than
thetargetship
0
r
v
,onthecontrary
0
r
v
.
K
isthelongitudinaldistancebetweentheownship
andthetargetvessel.
S
isasetsafedistance,where
r
Sv Ts
min
,
T
is
thesystemdelay, whichisrelatedtosensordelayand
calculation delay，
min
s
is a set of extended safety
distances.
When the point p is in the area
(which is
shown in Fig.9.), The longitudinal potential field is
distributed as a constant
s
hip
U
. Considering the
relative velocity
r
v
which is the target ship and the
ownship,when
0
r
v
,therelativedistancebetween
the target ship and the ship is gradually increased,
and the longitudinal potential value is 0. When
0
r
v
,itindicatesthatthedistancebetweentheship
and the target ship is gradually reduced. The more
dangerousthepotentialvalueis,themoretherelative
velocity is, the more dangerous and the potential
value is positively correlated with the relative
velocity.Theabove‐mentionedlongitudinalpotential
distributionreflectsthe
riskdistributionofthetarget
vesselinitslongitudinaldirection,theinfluenceofthe
shiponitssurroundingenvironmentinitstransverse
direction,anditsoverallpotentialfieldisformedon
the basis of longitudinal, which is shown in Fig.10,
Fig.11,Fig.12.
The calculation of the target shipʹs potential
is
obtained by multiplying the longitudinal potential
field
s
hip
A
by the transverse potential field. The
horizontal calculation rule is a kind of Gaussian
function. The whole calculation method is shown in
theformula(6)
imn sh p
v
D
eA
2
2
exp
2
(6)
where,
Disthepotentialdistanceofmovingshipindifferent
areawhichisshowninFig9
v
is the convergence coefficient of the shipʹs
potential, and determines the horizontal
influencerangeofthepotentialfield.
v
2
l
2
l
2
d
2
d
2
d
y
2
d
y
2
l
x
1
2
l
x
v
22
22
ld
xy
22
22
ld
xy
2
2
1
22
ld
xy
v
2
2
1
22
ld
xy
v
Figure8.Thepotentialdistanceofmovingship
3.2 casestudyofthePotentialField
1 As discussed above, the potential field of a ship
whichpositionis(0,0),courseangleis0,speedis
10,andtheOutlineDimensionis20×5isshownin
fig from origin, x‐y, x‐z and y‐z four angles, the
shape
ofthepotentialfieldofoneshipisshownin
Fig10.
Figure9.Fourvisualanglesofthepotentialfieldofoneship
2 The potential field of two crossing ships from
origin, x‐y, x‐z and y‐z four angles is shown in
fig11whichposition,courseangle,speedandthe
shapeisshownintable1.Thesuperpositionofthe
virtual force field when the ship is approaching
canbeclearlyseen
inthefigure.
91
Table1.Thedetailsoftwoships
_______________________________________________
Ship Position Course Speed Outline
AngleDimension
_______________________________________________
Ship1 (0,0)01020×5
Ship2 (40,‐40) 9010 20×5
_______________________________________________
3 Thepotentialfieldoffourcrossingdifferentships
fromoriginandx‐yvisua langlesisshowninfig12
whichposition,courseangle,speedandtheshape
isshownintable2.Thesuperpositionofthevirtual
force field when the ship is approaching can be
clearlyseeninthefigure.
Table2.Thedetailsoffourdifferent ships
_______________________________________________
Ship Position Course Speed Outline
AngleDimension
_______________________________________________
Ship1 (100,100)210 530×10
Ship2 (100,‐100) 135 1020×6
Ship3 (‐50,70)‐601010×3
Ship4 (‐150,‐100) 601030×5
_______________________________________________
Figure10. Four visual angles of the potential field of two
crossingships
Figure11. The originand x‐y visual angle of the potential
fieldoffourcrossingships
4 THEDISTRIBUTEDMULTI‐SHIPCOLLISION
AVOIDANCE
This paper proposes a complete set of solutions for
multi‐ship collision avoidance in intelligent
navigation,by usingatop‐to‐bottom organizationto
structurethesystem.Thesystemisdesignedwithtwo
layers: the decision‐making layer for path planning
andthecontrol
layerforpathfollowing.Fig.13show
thesystem’sorganizationandelements.
Theuplayeristhedecision‐makinglayer,whichis
designed to prepare for path planning. First step of
the decision‐making layer is the collision avoidance
functionunderCOLREGs.Thispartofthealgorithm
isdesignedformaking
surethatwhetherownshipis
stand‐on ship or give‐way ship, should keep the
speedandcourseorchangethem.Thesecondpartof
theuplayer’sfunctionispathplanningindetailsby
usingtheAPFmethod.
Course Change
Range
Path Planning
Bound
Obstacle
Tracking
COLREGs
Speed Change
Range
Course keeping
Speed keeping
Give-way
Ship
Stand-on
Ship
Accurate
Noaccurate
CPA
Navigation
&Sensor
Stand-on or
Give-way
Norisk
Establish
Database
System
Setting
Virtual Force
Field Build
Target Ship
Information
Target Ship Prediction
Information
Obstacle
Identification
Start
SystemPreparation&DataPreprocessing
Decision-MakingunderCOLREGs PathPlanning
Risk
if Necessary
if Necessary
Planned Speed
Path Planning
Decision
Figure12.Thesystem’sorganizationandelements
5 CONCLUSIONS
Amulti‐shipcollisionavoidancedecision‐makingand
pathplanningformulationisstudiedinadistributed
way.Thispaperproposesacompletesetofsolutions
for multi‐ship collision avoidance in intelligent
navigation,by usingatop‐to‐bottom organizationto
structurethesystem.Thesystemisdesignedwith
two
layers: the collision avoidance decision‐making and
thepathplanning.Underthegeneralrequirements of
the International Regulations for Preventing
Collisions at Sea (COLREGs), the performance of
distributed path planning decision‐making for anti‐
collisionisanalyzedforbothgive‐wayandstand‐on
ships situations, including the emergency
actions
taken by the stand‐on ship in case of the give‐way
ship’s fault of collision avoidance measures. The
ArtificialPotentialFieldmethod(APF)isusedforthe
pathplanningindetails.ThedevelopedAPFmethod
combined with the model of ship domain takes the
targetships’speedandcourseinto
account,sothatit
canjudgethemovingcharacteristicsofobstaclesmore
accurately.Simulationresultsindicatethatthesystem
proposedcanworkeffectiveness.
ACKNOWLEDGMENTS
Theresearchwas sponsoredbygrantsfromthe Key
Project in the National Science & Technology Pillar
Program(GrantNo.2015BAG20B05)
REFERENCES
Dongzhi,M. A., Y. A. N. Xinping, Z. Jiangbin and Z. Tao
(2014).ʺRemoteAcquisitionandTransmissionSystemof
Vessel Energy Efficiency Information.ʺ Journal of
Transport Information and Safety 32(4): 92‐96,107.(In
Chinese)
92
Hansen,M.G.,T.K.Jensen,T.Lehn‐Schiøler, K.Melchild,
F. M. Rasmussen and F. Ennemark (2013).ʺEmpirical
ship domain based on AIS data.ʺ Journal of Navigation
66(06):931‐940.
Jian,L.U.,Z.Wenjun,Y.HaifeiandJ.Jun(2014). ʺAnalysis
of Rear‐end Risk Based on the Indicator
of Time to
Collision.ʺ Journal of Transport Information and Safety
32(5):58‐64,76.(InChinese)
Lee,S.‐M.,K.‐Y.KwonandJ.Joh(2004).ʺAfuzzylogicfor
autonomous navigation of marine vehicles satisfying
COLREG guidelines.ʺ International Journal of Control
AutomationandSystems2:171‐181.
Lingling, L. I.
and Q. I. U. Lei (2014).ʺVessel Traffic
AccidentForecastingUsingaCombinationGrayNeural
Network Model.ʺ Journal of Transport Information and
Safety(3):110‐113,118.(InChinese)
Liu, Y., W. Liu, S. Rui and B. Richard (2015).ʺPredictive
navigationof unmannedsurface vehiclesin adynamic
maritime environment when using
the fast marching
method.ʺInternationalJournalofAdaptiveControl&Signal
Processing.
Perera,L.P.,J.P.CarvalhoandC.G.Soares(2011).ʺFuzzy
logic based decision making system for collision
avoidance of ocean navigation under critical collision
conditions.ʺ Journal of Marine Science and Technology
16(1):84‐99.
Pietrzykowski,Z.
andJ.Uriasz(2009).ʺTheshipdomain–a
criterion of navigational safety assessment in an open
seaarea.ʺJournalofNavigation62(01):93‐108.
Rui, S., Y. Liu and R. Bucknall (2015).ʺPath planning
algorithmforunmannedsurfacevehicleformationsina
practical maritime environment.ʺ Ocean Engineering 97:
126‐144.
Rui,
S.,Y.LiuandR.Bucknall(2016).ʺTheangleguidance
pathplanningalgorithmsforunmannedsurfacevehicle
formationsbyusingthefastmarchingmethod.ʺApplied
OceanResearch59:327‐344.
Ruishan, S. U. N., W. Lei and L. I. U. Junjie (2016).ʺAn
Overview and Prospect of Civil Aviation Safety in
China.ʺJournalofTransportInformationandSafety34(4):1‐
5,69.(InChinese)
Stahlberg, K., F. Goerlandt, J. Montewka and P. Kujala
(2012).ʺUncertainty in Analytical Collision Dynamics
Model Due to Assumptions in Dynamic Parameters. ʺ
TransNav, the International Journal on Marine Navigation
andSafetyofSeaTransportation,Vol.6,No.1:
47‐54.
Statheros,T.(2008).ʺAutonomousShipCollisionAvoidance
Navigation Concepts, Technologies and Techniques.ʺ
JournalofNavigation61(1):129‐142.
Uriasz, Z. P. J. (2009).ʺThe Ship Domain – A Criterion of
NavigationalSafety Assessment inanOpen SeaArea.ʺ
JournalofNavigation62(1):93‐108.
Zhang, J., D. Zhang, X.
Yan, S. Haugen and C. G. Soares
(2015).ʺA distributed anti‐collision decision support
formulation in multi‐ship encounter situations under
COLREGs.ʺOceanEngineering105:336‐348.
Zhonglian,J.,C.H.U.XiuminandY.A.N.Xinping(2015).
ʺDevelopments and Prospects of Intelligent Water
Transport:A Review of the Symposium
on Intelligent
Water Transport, in Conjunction with the 10th Annual
Meeting of China ITS.ʺ Journal of Transport Information
andSafety(6):1‐8.(InChinese)
