基于matlab求解二维离散动态规划问题_longer3fp_matlab_动态规划程序_离散规划matlab_二维_

function [p_opt,fval]=mindynprog1(x1,x2,DecisFun,StageObjFun,Stage_TransFun,ObjFun) [k1,k]=size(x1);[k2,k]=size(x2);%k为阶段数,size函数获取矩阵大小 x1_isnan=~isnan(x1);%判断数组的元素是否是NaN x2_isnan=~isnan(x2); f_opt=nan*ones(k1,k2,k);%f_opt(i,j,m)为第m阶段初状态值为（x1（i,m),x2(j,m))下的最优值，初值为非数 d_opt1=f_opt;d_opt2=f_opt;%(d_opt1((i,j,m),d_opt2((i,j,m))为第m阶段初状态值为x1（i,m),x2(j,m))下的最优决策值，初值为非数 tmp11=find(x1_isnan(:,k)); tmp21=find(x2_isnan(:,k));%找出第k阶段值（不是非数）的下标 tmp12=length(tmp11); tmp22=length(tmp21); for i=1:tmp12 for t=1:tmp22 [u1,u2]=feval(DecisFun,k,x1(tmp11(i),k),x2(tmp21(t),k)); tmp13=length(u1);tmp14=length(u2);t_vubm=inf; for j=1:tmp13 for l=1:tmp14 tmp=feval(StageObjFun,k,x1(tmp11(i),k),x2(tmp21(t),k),u1(j),u2(l)); if tmp<=t_vubm f_opt(tmp11(i),tmp21(t),k)=tmp; d_opt1(tmp11(i),tmp21(t),k)=u1(j); d_opt2(tmp11(i),tmp21(t),k)=u2(l); t_vubm=tmp; end end end end end for ii=k-1:-1:1 tmp011 =find(x1_isnan(:,ii));tmp021=find(x2_isnan(:,ii));tmp012=length(tmp011);tmp022 =length(tmp021); for i=1:tmp012 for t=1:tmp022 [u1,u2] =feval(DecisFun,ii,x1(tmp011(i),ii), x2(tmp021(t),ii)); tmp013=length(u1);tmp014=length(u2);t_vubm=inf; for j=1:tmp013 for l=1:tmp014 tmp000 =feval(StageObjFun,ii,x1(tmp011(i),ii),x2(tmp021(t),ii),u1(j),u2(l)); tmp100=feval(Stage_TransFun,ii,x1(tmp011(i),ii),x2(tmp021(t),ii),u1(j),u2(l)); %由该阶段的状态值及相应的决策值求出下一阶段的状态值 tmp200=x1(:,ii+1)-tmp100(1); tmp300=x2(:,ii+1)-tmp100(2); tmp400=find(tmp200==0);%找出下一阶段的状态值在x1(:,ii+1)的下标 tmp500=find(tmp300==0);%找出下一阶段的状态值在x2(:,ii+1)的下标 if ~isempty(tmp400)&&~isempty(tmp500) if nargin<6 tmp000=tmp000+f_opt(tmp400(1),tmp500(1),ii+1); else tmp000=feval(ObjFun,tmp000,f_opt(tmp400(1),tmp500(1),ii+1)); end if tmp000<t_vubm f_opt(tmp011(i),tmp021(t),ii)=tmp000; d_opt1(tmp011(i),tmp021(t),ii)=u1(j); d_opt2(tmp011(i),tmp021(t),ii)=u2(l); t_vubm=tmp000; end end end end end end end fval=f_opt(x1_isnan(:,1),x2_isnan(:,1),1); p_opt=[ ];tmpx1=[ ];tmpx2=[ ];tmpd1=[ ];tmpf=[ ];tmpd2=[ ]; tmp11=find(x1_isnan(:,1));tmp01=length(tmp11); tmp12=find(x2_isnan(:,1));tmp02=length(tmp12); for i=1:tmp01 q=(i-1)*k*tmp02; for j=1:tmp02 t=k*(j-1);t=q+t;tmpd1(i)=d_opt1(tmp11(i),tmp12(j),1); tmpd2(j)=d_opt2(tmp11(i),tmp12(j),1); %求出第一阶段的决策值 tmpx1(i)=x1(tmp11(i),1); tmpx2(j)=x2(tmp12(j),1); %求出第一阶段的状态值 tmpf(i,j)=feval(StageObjFun,1,tmpx1(i),tmpx2(j),tmpd1(i),tmpd2(j)); %求出第一阶段的指标函数值 p_opt(t+1,[1 2 3 4 5 6])=[1,tmpx1(i),tmpx2(j),tmpd1(i),tmpd2(j),tmpf(i,j)]; for ii=2:k %按顺序求出各阶段的决策值、状态值以及指标函数值 u=feval(Stage_TransFun,ii-1,tmpx1(i),tmpx2(j),tmpd1(i),tmpd2(j)); tmpx1(i)=u(1);tmpx2(j)=u(2); tmp1=x1(:,ii)-tmpx1(i);tmp2=x2(:,ii)-tmpx2(j); tmp3=find(tmp1==0);tmp4=find(tmp2==0); if ~isempty(tmp3)&&~isempty(tmp4) tmpd1(i)=d_opt1(tmp3(1),tmp4(1),ii); tmpd2(j)=d_opt2(tmp3(1),tmp4(1),ii); end tmpf(i,j)=feval(StageObjFun,ii,tmpx1(i),tmpx2(j),tmpd1(i),tmpd2(j)); p_opt(t+ii,[1 2 3 4 5 6])=[ii,tmpx1(i),tmpx2(j),tmpd1(i),tmpd2(j),tmpf(i,j)]; end end end end