notes algorithm for mo pso

Upload: prateikmenon

Post on 13-Apr-2018

216 views

Category:

Documents


0 download

TRANSCRIPT

  • 7/26/2019 Notes Algorithm for MO PSO

    1/6

    Algorithm for SO-PSO (Single Objective Particle Swarm Optimization)

    Step 0)INITIAI!"

    Bounds on mdecision variables (positionsof particles):xiL,xiU; i= 1, 2, . . ., m

    Number of particles (solutions): Nomputational parameters: !ma", !min, ma"iter, c1, c2

    #nitial velocitiesof t$e N particles%solutions:

    (& )

    1,jV

    ,

    (& )

    2, jV

    , . . . ,

    (&)

    ,m jV

    ; for eac$j= 1, 2, . . . , N

    (all 'ero; or an ot$er set of values); see able 1 (for m= 2,N= *)

    Step #)$"N"%AT" T&" INITIA POSITIONS (

    (& )

    ,i jx

    ; i= 1, 2, . . . m;j= 1, 2, . . . , N) O'

    EACH O' T&" N PA%TI" AN* T&" O+,"TI" '.NTIONS +Ij(&)(

    (& )

    1,jx

    ,

    (& )

    2, jx

    , . . . ,

    (& )

    ,m jx

    );j= 1, 2,..., N,O' "A& PA%TI"

    -enerate t$e initial positions of t$e N points !it$in t$e bounds of xi, usinseveralrandom

    numbers (in se/uence),R:

    (& ),i jx

    =xi,L0R(xiUxiL); & R 1; i= 1, 2, . . . , m;j= 1, 2, . . . ,N

    (t$is !ill satisf all t$e bounds). 3nter in able 1 (for m= 2,N= *)

    #f constraints are present, use penalt functions

    3valuate t$e ob4ective function for all t$e particles:Ij(&)(

    (& )

    1,jx

    ,

    (& )

    2, jx

    , . . . ,

    (& )

    ,m jx

    );j= 1, 2,...,N

    3nter in able 1 (for m= 2,N= *)

    Step /)ANA!" T&" %"S.TS TILL NOW

    567 eachof t$ej= 1, 2, . . . ,Nparticles, identif t$epersonalbest,

    (&)

    ,i jPbest

    , value ofI

    (lo!est, obviousl, it isIj(&)) and put t$e correspondinxivalues as

    (&)

    ,i jPbest

    =

    (& )

    ,i jx

    ; i= 1, 2, . . ., m;j= 1, 2, . . . ,N. 3nter in able 1 (for m= 2,N= *)

    1

  • 7/26/2019 Notes Algorithm for MO PSO

    2/6

    839Ij(&)for eachof t$ej= 1, 2, . . . , Nparticles, and identif t$e best in t$e s!arm (i.e.,

    identif t$egroupbest,

    (&)

    ,i jGbest

    ; lo!est amon t$eNvalues ofIj(&);j= 1, 2, . . . , N. a, it is

    particle, j= k. ut t$e correspondinxivalues as

    (&)

    ,i jGbest

    =

    (& ),i kx

    ; i= 1, 2, . . . , m. learl,

    (&)

    ,i jGbestis t$e same for allj. 3nter in able 1 (for m= 2,N= *)

    Step 1).P*AT" T&" "OITI"S

    (1)

    1,jV

    (1)

    2, jV

    2 2 2

    (1)

    ,m jV

    'O% "A&j3 # / 2 2 2 N

    ! = !ma" , iteration No. = &

    (1)

    ,i jV= w

    (1)

    ,i jV0 c1R+

    (&)

    ,i jPbest

    (& )

    ,i jx 0 c2R+

    (&)

    ,i jGbest

    (& )

    ,i jx;

    i= 1, 2, . . . , m;j= 1, 2, . . . ,N

    3nter in able 1 (for m= 2,N= *)

    Note t$at t$e Vs involve t$e old Vs, t$e particle best and t$e s!arm best (a

  • 7/26/2019 Notes Algorithm for MO PSO

    3/6

    #f t$e

    (1)

    ,i jx

    violate t$e bounds, c$ane t$em to satisf t$e bounds

    alculate

    (1)

    jI

    . 3nter in able 2 (for m= 2,N= *)

    ain, calculate

    (1),i jPbest

    and

    (1),i jGbest

    Update iteration number b unit and wusin

    w= wma"> +(wma"> wmin)(iteration number)%ma"iter

    alculate

    (2 )

    ,i jV

    , usin t$e e/uation before

    3nter in able 2 (for m= 2,N= *)

    Table /4 (some entries belo! are copies of t$ose in able1)

    Particle

    No2j

    (& )1,jx

    (&)

    2, jx (1)

    1,jV (1)

    2, jV (1)

    1,jx (1)

    2, jx (1)

    jI

    #

    /

    1

    5

    Particl

    e No2j

    (1)

    1,jV (1)

    2, jV (1)

    1,jx (1)

    2, jx (1)

    1,jPbest (1)

    2,jPbest (1)

    1,jGbest (1)

    1,jGbest (2 )

    1,jV (2 )

    2, jV

    #

    /

    1

    5

    ontinue t$e iterations till converence or !$en iteration number e"ceeds ma"iter

    ??????????????????????????????????????

    @

  • 7/26/2019 Notes Algorithm for MO PSO

    4/6

    *aterial belo! on *O+,-O is not correcte$ et

    Step 5)A%&IIN$ O' NON-*O6INAT"* POINTS

    ompare 7(< 0 1)!it$ everynon?dominated member present in 5ile 2, one b one

    #f 7(< 0 1)is a non?dominated point, cop it in 5ile 2 (as !ell as in 5ile 1)

    #f 7(< 0 1)is dominated over b a member alread present in 5ile 2, do not include it in 5ile 2

    but c$ec< if ou can cop it in 5ile 1 (tep above)

    #f an earlier member is dominated over b t$is ne! point, delete t$e former from 5ile 2 (it is

    alread in 5ile 1)

    Step 8)%"T.%N TO +AS"

    7eturn to an uncro!ded base point ($avin a lare cro!din distance from 5ile 2)

    periodically (so as to enerate a more continuous areto set, b locall e"plorin around

    uncro!ded points in t$e non?dominated set)

    5irst return to base is done after NB,1(= 2N,1) iterations (after t$e first N,1)

    Number of iterations before 4t$return to base = NB,4= rBNB,4 ? 1; 4 = 2, @, . . .

    & rB 1 (enerall, &.A); s.t.: NB,4 1&

    -enerate 5ile @ (of uncro!ded members in 5ile 2)

    Normali'e fi(usin ma" and min values) for all points in 5ile 2, so t$at & fi 1

    5ind t$e cro!din distances (#C#) of allpoints in 5ile 2 (as in N-?##)

    op 4of t$e most uncro!ded points (includin t$e boundar points) into 5ile @, !$ere

    4= D4; D401= r D4; D1= 1; r = &.A; 4 E#N= F

    Use a random number (diiti'ed appropriatel) to select t$e uncro!ded base point to be

    returned to, from 5ile @

    IST O' %AN*O6 N.6+"%S $I"N +"O9 (.S" IN T&" S":."N" ;n $I"N) "S" T&" 9ONT +" %AN*O6)

    *

  • 7/26/2019 Notes Algorithm for MO PSO

    5/6

    1 G.2H&@I23&1 @& A.IFFA@&3&1 HA *.1@**FH3&1 II @.AF2F1F3&1

    2 @.*@GF1G3&1 @1 F.AA1@*23&1 F& G.I*GI1@3&1 IA G.21&1AH3&1

    @ @.@HI&&13&1 @2 H.G*F2IF3&1 F1 A.1&11FA3&1 A& *.FF&HAG3&1

    * *.*&I2*23&1 @@ G.1GA&GG3&1 F2 @.IG1II23&1 A1 G.I1G2@F3&1

    H 1.1FG&@*3&1 @* 1.G&*HIG3&1 F@ I.1&F21*3&1 A2 @.&G&2AH3&1

    F F.*HI*A13&1 @H 2.G@GGF13&1 F* @.21F*F13&1 A@ A.GI&AAF3&1

    G A.@&&IA*3&1 @F @.@1A2213&1 FH A.I@HFG23&1 A* 2.2G12&H3&1

    I @.11A*GH3&1 @G G.@HG@223&1 FF H.F1@1GA3&1 AH 2.*FGF@@3&1

    A 1.@A&F&*3&1 @I I.A1@G&H3&1 FG H.H&1AI*3&1 AF F.2F*HG&3&1

    1& G.AGIHHH3&1 @A H.&1@2F13&1 FI *.2IG&2G3&1 AG 1.2A1HHF3&1

    11 2.HH@2I23&1 *& I.1G*1*@3&1 FA A.1@2G1&3&1 AI G.**AIF23&1

    12 I.*G*I@A3&1 *1 F.AA*AII3&1 G& 2.G*2&HF3&1 AA @.H@G@AH3&1

    1@ G.&AIA*A3&1 *2 F.A*&1&@3&1 G1 H.G&FH&A3&1 1&& I.*H2*&H3&1

    1* 2.IGHF@&3&1 *@ 2.AG2IH*3&1 G2 G.@&*1I&3&1 1&1 1.H@AAAG3&1

    1H 1.2GFAG23&1 ** 2.I11A223&1 G@ H.F*IH*F3&1 1&2 @.&H@I&H3&1

    1F I.FF2AGH3&1 *H @.G*2A2&3&1 G* H.H22AA@3&1 1&@ I.GFI2FI3&1

    1G A.&@1A&&3&1 *F I.H@*A1G3&1 GH A.2G*H2G3&1 1&* F.H*@F*I3&1

    1I *.*@AH1H3&1 *G H.1HAGII3&1 GF H.2A&H1H3&1 1&H *.1*HI1I3&1

    1A 2.HH211I3&1 *I I.@&1*I23&1 GG F.AA*A2&3&1 1&F *.I2@@I23&1

    2& G.2H*G1H3&1 *A I.GA&AFA3&1 GI 2.1&HAI@3&1 1&G 1.&IGHGF3&1

    21 @.IIG@2I3&1 H& G.&2A12F3&1 GA A.H@@21H3&1 1&I @.1FAIF13&1

    22 G.2&A&GF3&1 H1 G.AI&1II3&1 I& 2.2*G@2&3&1 1&A *.2&&HFI3&1

    2@ G.&G2@1@3&1 H2 I.HAFI123&1 I1 2.II@@&I3&1 11& 1.F1G*&@3&1

    2* G.GH*GG23&1 H@ 2.1*1HHA3&1 I2 @.1HFI&F3&1 111 @.IF*&2*3&1

    2H @.GH1*1F3&1 H* *.AFF@2A3&1 I@ G.H&A2@F3&1 112 2.2AHA2*3&1

    2F A.1*1I*I3&1 HH I.221HHG3&1 I* G.&**2AH3&1 11@ A.@2@A*@3&1

    2G *.I1&22&3&1 HF 1.H&*&AF3&1 IH G.FHGF113&1 11* A.*G@IFA3&1

    2I 1.F*@G&23&1 HG F.2HIH*&3&1 IF F.F&*1H*3&1 11H @.&HGG1I3&1

    2A *.IA21&I3&1 HI 1.FH@*1*3&1 IG G.F*&&G23&1 11F F.&@&IG*3&1

    H

  • 7/26/2019 Notes Algorithm for MO PSO

    6/6

    11G F.&F1A2*3&1 1*H @.@IAG*&3&1 1G@ *.GFHIF23&1

    11I H.I@H&&A3&1 1*F A.GFH11*3&1 1G* A.@@IAIG3&1

    11A G.*II2GH3&1 1*G 2.@2&FF@3&1 1GH *.1FHGI13&1

    12& *.G1IAAG3&1 1*I A.1&&FA13&1 1GF 2.1@*AF*3&1

    121 *.@21F2@3&1 1*A 1.&FA1*F3&1 1GG 1.AGA*@F3&1

    122 I.&H212@3&1 1H& A.I1*A2A3&1 1GI 1.FF@@I23&1

    12@ G.*AF@213&1 1H1 *.@&2*1A3&1 1GA F.*@&I@A3&1

    12* F.I&@&*I3&1 1H2 2.1IFFHI3&1 1I& @.*IG1213&1

    12H 1.2&2&HA3&1 1H@ F.&G2FIG3&1 1I1 G.H&*@1H3&1

    12F 2.FH2G&H3&1 1H* A.&IF2H&3&1 1I2 1.1&AFA*3&1

    12G 1.AF21H&3&1 1HH *.2*A*FH3&1 1I@ A.**H*FH3&1

    12I A.1&*IAI3&1 1HF G.G1*A@I3&1 1I* G.I1IG*13&1

    12A 1.@GG1GH3&1 1HG H.1*I@IA3&1 1IH 1.IH21GH3&1

    1@& *.*H11*H3&1 1HI I.H2*HF13&1 1IF 1.1H&G@13&1

    1@1 *.2&2A@A3&1 1HA I.1F&@H23&1 1IG G.FFIA1F3&1

    1@2 @.I2H&HA3&1 1F& @.HHI*@*3&1 1II I.*@AG@*3&1

    1@@ G.A&H*1H3&1 1F1 2.HAIAG&3&1 1IA 2.2AF@G&3&1

    1@* G.A1HI123&1 1F2 @.&H22&I3&1 1A& F.@2*@@G3&1

    1@H I.@I@2AG3&1 1F@ F.1HF1*@3&1 1A1 1.H&H2*G3&1

    1@F H.IG@I@H3&1 1F* A.*FHFI13&1 1A2 G.2A*&*&3&1

    1@G 2.22**123&1 1FH H.F1A**13&1 1A@ I.A&&*G@3&1

    1@I I.&@H@FI3&1 1FF 2.&HIAF&3&1 1A* F.G*I2H*3&1

    1@A G.&1@G1I3&1 1FG A.&I21@F3&1 1AH 1.&2*FGH3&1

    1*& *.G@*IIG3&1 1FI *.A@&@G23&1 1AF *.*A2*F&3&1

    1*1 H.AAGAF23&1 1FA A.1@*IAH3&1 1AG G.&@A2&13&1

    1*2 1.I&H&F@3&1 1G& 2.FH*F@F3&1 1AI 2.2GAIA&3&1

    1*@ @.21@GII3&1 1G1 @.G1&2123&1 1AA H.HIH***3&1

    1** F.&&2&223&1 1G2 1.H&HH&13&1 2&& I.2&*&1&3&1

    Notes lorit$m E6 6

    F