RBF神经网络自适应控制MATLAB仿真示例程序

function [sys,x0,str,ts] = spacemodel(t,x,u,flag) switch flag, case 0, [sys,x0,str,ts]=mdlInitializeSizes; case 1, sys=mdlDerivatives(t,x,u); case 3, sys=mdlOutputs(t,x,u); case {2,4,9} sys=[]; otherwise error(['Unhandled flag = ',num2str(flag)]); end function [sys,x0,str,ts]=mdlInitializeSizes global node c_M c_C c_G b ce node=7; c_M=zeros(2,node); c_G=zeros(2,node); c_C=zeros(4,node); c_M=[0.25 0.5 0.75 1 1.25 1.5 1.75; 0.25 0.5 0.75 1 1.25 1.5 1.75]; c_G=[0.25 0.5 0.75 1 1.25 1.5 1.75; 0.25 0.5 0.75 1 1.25 1.5 1.75]; c_C=[0.25 0.5 0.75 1 1.25 1.5 1.75; 0.25 0.5 0.75 1 1.25 1.5 1.75; -1.5 -1 -0.5 0 0.5 1.0 1.50; -1.5 -1 -0.5 0 0.5 1.0 1.50]; b=10; ce=15.0; sizes = simsizes; sizes.NumContStates = 10*node; sizes.NumDiscStates = 0; sizes.NumOutputs = 5; sizes.NumInputs = 10; sizes.DirFeedthrough = 1; sizes.NumSampleTimes = 0; sys = simsizes(sizes); x0 = zeros(1,10*node); str = []; ts = []; function sys=mdlDerivatives(t,x,u) global node c_M c_C c_G b ce xd1=u(1); d_xd1=u(2); dd_xd1=u(3); xd2=u(4); d_xd2=u(5); dd_xd2=u(6); x1=u(7); d_x1=u(8); x2=u(9); d_x2=u(10); xx=[x1;x2]; for j=1:1:node h_M11(j)=exp(-norm(xx-c_M(:,j))^2/(b*b)); h_M12(j)=exp(-norm(xx-c_M(:,j))^2/(b*b)); h_M21(j)=exp(-norm(xx-c_M(:,j))^2/(b*b)); h_M22(j)=exp(-norm(xx-c_M(:,j))^2/(b*b)); end for j=1:1:node h_G1(j)=exp(-norm(xx-c_G(:,j))^2/(b*b)); h_G2(j)=exp(-norm(xx-c_G(:,j))^2/(b*b)); end z=[x1;x2;d_x1;d_x2]; for j=1:1:node h_C11(j)=exp(-norm(z-c_C(:,j))^2/(b*b)); h_C12(j)=exp(-norm(z-c_C(:,j))^2/(b*b)); h_C21(j)=exp(-norm(z-c_C(:,j))^2/(b*b)); h_C22(j)=exp(-norm(z-c_C(:,j))^2/(b*b)); end W_M11=[x(1:node)]'; W_M12=[x(node+1:node*2)]'; W_M21=[x(node*2+1:node*3)]'; W_M22=[x(node*3+1:node*4)]'; e1=xd1-x1; e2=xd2-x2; de1=d_xd1-d_x1; de2=d_xd2-d_x2; e=[e1;e2]; de=[de1;de2]; Hur=ce*eye(2); r=de+Hur*e; dxd=[d_xd1;d_xd2]; dxr=dxd+Hur*e; ddxd=[dd_xd1;dd_xd2]; ddxr=ddxd+Hur*de; T_M11=2*eye(node); T_M12=2*eye(node); T_M21=2*eye(node); T_M22=2*eye(node); for i=1:1:node sys(i)=T_M11(i,i)*h_M11(i)*ddxr(1)*r(1); sys(i+node)=T_M12(i,i)*h_M12(i)*ddxr(2)*r(1); sys(i+node*2)=T_M21(i,i)*h_M21(i)*ddxr(1)*r(2); sys(i+node*3)=T_M22(i,i)*h_M22(i)*ddxr(2)*r(2); end W_G1=[x(node*4+1:node*5)]'; W_G2=[x(node*5+1:node*6)]'; T_G1=5*eye(node); T_G2=5*eye(node); for i=1:1:node sys(i+node*4)=T_G1(i,i)*h_G1(i)*r(1); sys(i+node*5)=T_G2(i,i)*h_G2(i)*r(2); end W_C11=[x(node*6+1:node*7)]'; W_C12=[x(node*7+1:node*8)]'; W_C21=[x(node*8+1:node*9)]'; W_C22=[x(node*9+1:node*10)]'; T_C11=0.5*eye(node); T_C12=0.5*eye(node); T_C21=0.5*eye(node); T_C22=0.5*eye(node); for i=1:1:node sys(i+node*6)=T_C11(i,i)*h_C11(i)*dxr(1)*r(1); sys(i+node*7)=T_C12(i,i)*h_C12(i)*ddxr(2)*r(1); sys(i+node*8)=T_C21(i,i)*h_C21(i)*dxr(1)*r(2); sys(i+node*9)=T_C22(i,i)*h_C22(i)*ddxr(2)*r(2); end function sys=mdlOutputs(t,x,u) global node c_M c_C c_G b ce xd1=u(1); d_xd1=u(2); dd_xd1=u(3); xd2=u(4); d_xd2=u(5); dd_xd2=u(6); x1=u(7); d_x1=u(8); x2=u(9); d_x2=u(10); xx=[x1;x2]; for j=1:1:node h_M11(j)=exp(-norm(xx-c_M(:,j))^2/(b*b)); h_M12(j)=exp(-norm(xx-c_M(:,j))^2/(b*b)); h_M21(j)=exp(-norm(xx-c_M(:,j))^2/(b*b)); h_M22(j)=exp(-norm(xx-c_M(:,j))^2/(b*b)); end for j=1:1:node h_G1(j)=exp(-norm(xx-c_G(:,j))^2/(b*b)); h_G2(j)=exp(-norm(xx-c_G(:,j))^2/(b*b)); end z=[x1;x2;d_x1;d_x2]; for j=1:1:node h_C11(j)=exp(-norm(z-c_C(:,j))^2/(b*b)); h_C12(j)=exp(-norm(z-c_C(:,j))^2/(b*b)); h_C21(j)=exp(-norm(z-c_C(:,j))^2/(b*b)); h_C22(j)=exp(-norm(z-c_C(:,j))^2/(b*b)); end W_M11=[x(1:node)]'; W_M12=[x(node+1:node*2)]'; W_M21=[x(node*2+1:node*3)]'; W_M22=[x(node*3+1:node*4)]'; MSNN_g=[W_M11*h_M11' W_M12*h_M12'; W_M21*h_M21' W_M22*h_M22']; norm_Mp=norm(MSNN_g); W_G1=[x(node*4+1:node*5)]'; W_G2=[x(node*5+1:node*6)]'; GSNN_g=[W_G1*h_G1'; W_G2*h_G2']; norm_Gp=norm(GSNN_g); W_C11=[x(node*6+1:node*7)]'; W_C12=[x(node*7+1:node*8)]'; W_C21=[x(node*8+1:node*9)]'; W_C22=[x(node*9+1:node*10)]'; CDNN_g=[W_C11*h_C11' W_C12*h_C12'; W_C21*h_C21' W_C22*h_C22']; norm_Cp=norm(CDNN_g); e1=xd1-x1; e2=xd2-x2; de1=d_xd1-d_x1; de2=d_xd2-d_x2; e=[e1;e2]; de=[de1;de2]; Hur=ce*eye(2); r=de+Hur*e; dxd=[d_xd1;d_xd2]; dxr=dxd+Hur*e; ddxd=[dd_xd1;dd_xd2]; ddxr=ddxd+Hur*de; Ks=0.5; K=30*eye(2); Fx=MSNN_g*ddxr+CDNN_g*dxr+GSNN_g+K*r+Ks*sign(r); sys(1)=Fx(1); sys(2)=Fx(2); sys(3)=norm_Mp; sys(4)=norm_Cp; sys(5)=norm_Gp;