320
2 PIDVCAINSTRUCTIONS
2.1 GoalandPrincipleofPIDVCA
The aim of PIDVCA algorithm is to provide crews
safe, scientific and economic decision‐making for
vessel collision avoidance by a machine (computer),
whichcanfollowstheCOLREGSandhas theability
tosimulatetheordinarypracticesofcrewsandreflect
theirseamanship.Itssecurityisreflectedintheoutput
of the avoidance decision that is premised on the
maximized DCPA, and its scientific is embodied in
the organic integration of quantitative analysis of
vessels ’ relative motion geometric graphic and
qualitativeanalysisofCOLREGSandexpertempirical
knowledge, and its economy
is reflected in its
minimizedtrackoffset,whilesatisfytheprecondition
ofthesafety.
In the co‐existence stage of manned and
unmannedships,forthePIDVCAcanbeacceptedby
the sailors to achieve concerted avoidance between
manned ship and unmanned ship, algorithm’s
principles
areasfollows.
First,followthespiritofCOLREGS.
Second, simulate the ordinary practices of crews
andreflecttheirexcellentseamanship.
Thirdly, reflect the concretization, clearness and
reasonable extension about important concepts and
provisionsoftheCOLREGSinpractice.
COLREGSisveryclearabouttherulesofconduct
oftwovessels
inanyconditionofvisibility,andthere
is no specific advice or specific method for collision
avoidance of multiple vessels, only the wordsʺ
ordinary practice of seamen, excellent seamanshipʺ
are used. According to crew training and expert
consultation,ʺordinarypracticeofseamen” means
the making‐decisions for collision avoidance of
multiplevesselsthatcanbewidelyacceptedbycrews,
for example, adopt relevant requirements of
COLREGS about two vessels to avoid key vessel of
the multiple vessels ’ encounter situation.
“seamanship of crews” is short for excellent
seamanship is expressed in excellent crews making
reasonableandeffectiveavoidancedecisions,concrete
embodiment
in the correctly evaluation of risk of
collision, safety distance approaching (SDA),
avoidancetime,avoidancemeasureandresumetime,
as well as inthe flexible reactioncapacity to handle
complicatedsituation.
2.2 PIDVCAalgorithmcomposition
Tothe problemthat the collisionsituationisendless
andunrepeatableinthenavigationdue
tothevessels’
differenttypes,scales,speeds,sailingwatersandthe
angles of the two vessels’ rendezvous, PIDVCA
adoptsthepersonalizedmachinelearningmethodto
realize the purpose of automatic target perception,
cognitive target, generation, verification and
optimizationdecision.
Accordingtotheprinciple,PIDVCAconsistsofthe
initial PIDVCA generation algorithm
and the
PIDVCAverificationandoptimizationalgorithm.The
initialPIDVCAgenerationalgorithmiscomposed of
the target intercrossing characteristics recognition
algorithm, target potential danger judgment
algorithm,encounter attributerecognitionalgorithm,
the shipʹs avoidance attribute recognition algorithm
for dangerous targets, encounter situation analysis
and classification algorithm, dynamic hazard
assessment algorithm
and PIDVCA plan generation
algorithm. PIDVCA verification and optimization
algorithm is made up of predicting target potential
risk analysis algorithm, ordinary algorithm of
simulation of ordinary practice and seamanship
algorithm of excellent seamanship, space searching
optimization algorithm, algorithm of space‐time
searching optimization, algorithm of coordination
collisionavoidanceoptimization,PIDVCA plan
local
dynamicoptimizationalgorithm,PIDVCAgeneration
algorithm and emergency decision algorithm of
dangeroussituation.
PIDVCAalgorithmisdividedintotwo‐vesseland
multi‐vessel PIDVCA generation and optimization
algorithm considering two‐vessel and multi‐vessel
encounter situation. For the two‐vessel PIDVCA
generation and optimization algorithm, the initial
PIDVCA generation
algorithm can basically achieve
satisfactory avoidance effect, When necessary, it is
supplemented by coordinated collision avoidance
decision optimization algorithm, spatial searching
optimization algorithm and PIDVCA plan dynamic
optimization algorithm under the guidance of
dynamic optimization objective function. For
example, in Condition of restricted Visibility, when
the own‐ship avoids the target
ship near the left
beam, it needs to be supplemented by the local
PIDVCA plan dynamic optimization algorithm to
avoidproducingalargetrackoffset.ForthePIDVCA
generation and optimization algorithm of multi‐
vessel,itisnecessarytoperformthepredictiontarget
ship potential risk analysis algorithm of PIDVCA
verification and optimization algorithm, and
supplementedbyotheralgorithmswhennecessary.
3 ANALYZEOFTHERELATIVEMOTION
GEOMETRICCHANGERULESOFCROSSING
INCONDITIONOFNVISIBILITY
3.1 Encountersituationdescription
TakingthecrossingsituationandintheOpenwaters
asanexample,itisassumedthatthetwovesselshave
the
samescaleanddifferentspeeds,andtheaffecting
of vessels’ maneuvering are ignored, and the SDA
betweenthetwovesselsisthesame,setas1seamile,
and the DCPA is zero. In order to facilitate the
comparativeanalysis,thetwovesselsareeachotherʹs
target‐vesselsand
altercoursetorightforavessel.
FIG. 1 and 2 are geometric diagrams of relative
motionoftwo vesselsin thesamecrossing situation
astheirleftcrossingandrightcrossing.Thatis,when
the portside vessel as the own‐ship,the other vessel
on her right, and when the
starboard vessel is the
own‐ship, the other vessel on her left. For ease of
comparison, two independent relative motion
geometriesarepresentedintheonegeometry.Inthe
figures,C0andV0respectivelyrepresenttheheading
andvelocityoftheown‐ship,VtandVrrespectively
representthetrue
andrelativevelocityofthetarget‐
ship,andACistheturningangle.