多目标粒子群优化算法matlab_【优化求解】优化的离散二进制粒子群算法

% NBPSO code v1.0% Generated by Majid Rostami Shahrbabaki, 2010. %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%5% Particle swarm optimization (PSO) is one of the modern heuristic algorithms that can be applied to continuous and discrete optimization problems.% The original binary PSO (BPSO) has got some disadvantages that make the algorithm not to converge well.% To deal with these disadvantages, a new BPSO (NBPSO) is introduced. The results provided in the following papers show the superiority of the NBPSO.% % [1]. M.Rostami Shahrbabak and H.Nezamabadi-pour, " A New Approach to Binary PSO Algorithm" 14th Iranian Conference on Electrical Engineering, may 2006.% [2]. H.Nezamabadi-pour, M.Rostami Shahrbabaki and M.Maghfoori Farsangi "Binary Particle Swarm Optimization: Challenges and new% Solutions"The CSI Journal on Computer Science and Engineering Vol. 6, No. 1 (a), pp. 21-32, 2008.%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% Main function for NBPSO algorithm. The function called here are :% initialize : gives the initial parameters% range_func : depend on the Num_func, determines ranges for position and velocity% evaluate : depend on the Num_func, evaluates the function% renewp_best : finds the Particle_best% renewg_best : finds the Global_best% update_v_p  : updates position and velocities of particles% display_result : displays the resultsclear,close all,clc  %#okfor j = 1:50    [N,K,D,L,var,w_max,w_min,c1,c2,position,p_best,g_best,fitness,p_best_fit,...        Num_func,Min_Max_flag,Gl_Lo_flag] = initialize;    [v_max,x_max,velocity,] = range_func(Num_func,N,D) ;    for k=1:K        w = w_max+(((w_min-w_max)*(k-1))/(K-1));        fitness = evaluate(position,k,N,D,L,var,x_max,fitness,Num_func);        [p_best,p_best_fit] = renewp_best(D,fitness,p_best,N,k,position,p_best_fit,Min_Max_flag);        g_best=renewg_best(p_best,p_best_fit,N,Min_Max_flag,Gl_Lo_flag);        [position,velocity]=update_v_p(D,N,c1,c2,w,p_best,g_best,position,velocity,v_max,Gl_Lo_flag);    end    if Min_Max_flag == 1        best_fit(j,:) = min(fitness);   %#ok    else        best_fit(j,:) = min(fitness); %#ok    end    mean_fit(j,:) = mean(fitness); %#ok    disp([' End of run ' , num2str(j) , '   ==> Total run is 50'])enddisplay_result(best_fit,mean_fit,K,Min_Max_flag)

function X=decode(temp,L,x_max)x=0;for i=1:L    x=x+temp(i)*2^(i-1);endx=x/(2^L-1);X=((2*x_max)/(1-0))*(x-0)-x_max;return

function fitness=evaluate(position,k,N,D,L,var,x_max,fitness,Num_func)for i=1:N    for j=1:var        temp=position(i,(j-1)*L+1:j*L);        X(j)=decode(temp,L,x_max); %#ok    end    switch Num_func        case 1            result = sum (X.^2);        case 2            result = sum(abs(X)) + prod(abs(X)) ;        case 3            result = 0 ;            for ii=1:var                result = result + sum(X(1:ii)).^2;            end        case 4            result = max (abs(X));        case 5            result = 0;            for ii=1:var-1                result = result + 100*((X(ii+1)-X(ii)^2)^2+(X(ii)-1)^2);            end    end    fitness(i,k) = result ; endreturn

function [N,K,D,L,var,w_max,w_min,c1,c2,position,p_best,g_best,fitness,p_best_fit,...    Num_func,Min_Max_flag,Gl_Lo_flag]=initializeN = 50;    % N is the number of the particlesK = 1000;  %K is the number of iterationvar = 5; % var is number of variablesL = 15 ; % L is the lenght for each variableD = L*var; % D is the dimension of each particlew_min=0.1;w_max=0.6;c1=2;c2=2;% w is the inertia factor and c1 & c2 are learning factorsposition = rand(N,D)>0.5; % Generates initial populationfitness=0;p_best = rand(N,D)>0.5;g_best = rand(N,D)>0.5;p_best_fit = ones(N,1);Num_func = 1 ; % Select the number of function to be evaluatedMin_Max_flag = 1 ;  % 1 if the function must be minimized  ....  2 if the function must be maximizedGl_Lo_flag = 1  ;   % 1 if the search is global ...............  2 if the search is localreturn

function [v_max,x_max,velocity] = range_func(Num_func,N,D)switch Num_func    case 1        x_max = 100 ;        % x_min=-x_max  (here)    case 2        x_max = 10 ;        % x_min=-x_max  (here)    case 3        x_max = 100 ;        % x_min=-x_max  (here)    case 4        x_max = 100 ;        % x_min=-x_max  (here)    case 5        x_max = 30 ;        % x_min=-x_max  (here)end    v_max = 0.5 * x_max; % it can be tunedvelocity = -v_max + 2*v_max*rand(N,D);return

function [p_best,p_best_fit]=renewp_best(D,fitness,p_best,N,k,position,p_best_fit,Min_Max_flag)if Min_Max_flag == 1    for n=1:N        if k==1            min_p=fitness(n,k);            p_best(n,:)=position(n,:);            p_best_fit(n,1)=min_p;        elseif p_best_fit(n,1) > fitness(n,k)              p_best_fit(n,1)=fitness(n,k);            p_best(n,:)=position(n,:);        end    endelse    for n=1:N        if k==1            min_p=fitness(n,k);            p_best(n,:)=position(n,:);            p_best_fit(n,1)=min_p;        elseif p_best_fit(n,1) < fitness(n,k)              p_best_fit(n,1)=fitness(n,k);            p_best(n,:)=position(n,:);        end    endendreturn

function g_best=renewg_best(p_best,p_best_fit,N,Min_Max_flag,Gl_Lo_flag)            if Gl_Lo_flag == 1   %%% This part is for Global model              if Min_Max_flag == 1        [m_g place_g]=min(p_best_fit);    % finding minimum        g_best=p_best(place_g,:);    else        [m_g place_g]=max(p_best_fit);    % finding maximum        g_best=p_best(place_g,:);    endelse    if Min_Max_flag == 1 %%% This part is for local model                  for i=1:N            if i==1                temp_g=p_best_fit(N);                temp_g(2:3)=p_best_fit(i:i+1,1);                [min_g,place_g]=min(temp_g);     % finding minimum                switch place_g                    case 1                        g_best(i,:)=p_best(N,:);                    case {2,3}                        g_best(i,:)=p_best(place_g+i-2,:);                end            elseif i==N                temp_g=p_best_fit(1);                temp_g(2:3)=p_best_fit(i-1:i,1);                [min_g,place_g]=min(temp_g);   % finding minimum                switch place_g                    case 1                        g_best(i,:)=p_best(1,:);                    case {2,3}                        g_best(i,:)=p_best(place_g+i-3,:);                end            else                [min_g,place_g]=min(p_best_fit(i-1:i+1,1));  % finding minimum                g_best(i,:)=p_best(place_g+i-2,:);            end        end    else        for i=1:N            if i==1                temp_g=p_best_fit(N);                temp_g(2:3)=p_best_fit(i:i+1,1);                [max_g,place_g]=max(temp_g);     % finding maximum                switch place_g                    case 1                        g_best(i,:)=p_best(N,:);                    case {2,3}                        g_best(i,:)=p_best(place_g+i-2,:);                end            elseif i==N                temp_g=p_best_fit(1);                temp_g(2:3)=p_best_fit(i-1:i,1);                [max_g,place_g]=max(temp_g);   % finding maximum                switch place_g                    case 1                        g_best(i,:)=p_best(1,:);                    case {2,3}                        g_best(i,:)=p_best(place_g+i-3,:);                end            else                [max_g,place_g]=max(p_best_fit(i-1:i+1,1));  % finding maximum                g_best(i,:)=p_best(place_g+i-2,:);            end        end    end    end

function [position,velocity]=update_v_p(D,N,c1,c2,w,p_best,g_best,position,velocity,v_max,Gl_Lo_flag)if Gl_Lo_flag==1    for i=1:N        temp(i,:)=g_best-position(i,:); %#ok    end    velocity=w*velocity+c1*rand*(p_best-position)+c2*rand*temp;   %  For global modelelse    velocity=w*velocity+c1*rand*(p_best-position)+c2*rand*(g_best-position); % For loval modelendfor d=1:D    for n=1:N        if velocity(n,d)>v_max           velocity(n,d)=v_max;        elseif velocity(n,d)           velocity(n,d)=-v_max;        end    endendpro_velocity=abs(2*(logsig(velocity)-0.5));for n=1:N    for d=1:D        if rand            position(n,d)=xor(position(n,d),1);        else            position(n,d)=position(n,d);        end    endendreturn

往期回顾>>>>>>

【模式识别】Matlab指纹识别【优化求解】A*算法解决三维路径规划问题【优化求解】模拟退火遗传实现带时间窗的车辆路径规划问题【数学建模】Matlab实现SEIR模型【优化求解】基于NSGA-2的求解多目标柔性车间调度算法 【优化求解】蚁群算法求最优值  【图像处理】粒子群算法结合模糊聚类分割算法实现图像的分割 【优化算法】基于粒子群算法的充电站最优布局 【优化求解】粒子群算法之电线布局优化 【优化求解】遗传算法解决背包问题

【优化求解】基于粒子群之机器人栅格路径规划