?? predopt.m
字號:
function [sys,x0,str,ts] = predopt(t,x,u,flag,N2,Ts,Nu,maxiter,csrchfun,rho,alpha,S1,IW,LW1_2,LW2_1,B1,B2,Ni,Nj,min_i,max_i,minp,maxp,mint,maxt,Normalize)
%PREDOPT Executes the Predictive Controller Approximation based on Gauss Newton.
%
% Copyright 1992-2003 The MathWorks, Inc.
% Orlando De Jesus, Martin Hagan, 1-25-00
% $Revision: 1.6.2.2 $ $Date: 2004/04/10 23:45:58 $
switch flag,
%%%%%%%%%%%%%%%%%%
% Initialization %
%%%%%%%%%%%%%%%%%%
case 0,
load_system('ptest3sim2');
if Normalize
IW_gU=((maxt-mint)/(maxp-minp))*IW;
else
IW_gU=IW;
end
set_param('ptest3sim2/Subsystem','B2',num2str(B2,20),'B1',mat2str(B1,20),'LW2_1',mat2str(LW2_1,20), ...
'LW1_2',mat2str(LW1_2,20),'IW',mat2str(IW,20),'IW_gU',mat2str(IW_gU,20), ...
'Ts',num2str(Ts),'S1',num2str(S1),'Ni',num2str(Ni), ...
'Nj',num2str(Nj),'minp',num2str(minp,20),'maxp',num2str(maxp,20), ...
'minp',num2str(minp,20),'mint',num2str(mint,20),'maxt',num2str(maxt,20), ...
'Normalize',num2str(Normalize),'Nu',num2str(Nu));
assignin('base','t_init',cputime);
assignin('base','cont_u',0);
[sys,x0,str,ts]=mdlInitializeSizes(N2,Ts,Nu,alpha,S1,Ni,Nj,min_i,max_i);
%%%%%%%%%%
% Update %
%%%%%%%%%%
case 2,
sys = mdlUpdate(t,x,u,N2,Ts,Nu,maxiter,csrchfun,rho,alpha,S1,Ni,Nj,min_i,max_i,minp,maxp,mint,maxt,Normalize);
%%%%%%%%%%
% Output %
%%%%%%%%%%
case 3,
sys = mdlOutputs(t,x,u,Nu,Ni);
%%%%%%%%%%%%%
% Terminate %
%%%%%%%%%%%%%
case 9,
close_system('ptest3sim2',0);
assignin('base','t_end',cputime);
sys = [];
otherwise
error(['unhandled flag = ',num2str(flag)]);
end
%end sfundsc1
%
%=============================================================================
% mdlInitializeSizes
% Return the sizes, initial conditions, and sample times for the S-function.
%=============================================================================
%
function [sys,x0,str,ts]=mdlInitializeSizes(N2,Ts,Nu,alpha,S1,Ni,Nj,min_i,max_i)
global tiu dUtilde_dU
global N1 d alpha2 upi uvi
sizes = simsizes;
sizes.NumContStates = 0;
sizes.NumDiscStates = Ni+Nu-1+(S1+1)*(Nj-1);
sizes.NumOutputs = 1;
sizes.NumInputs = -1;
sizes.DirFeedthrough = 0;
sizes.NumSampleTimes = 1;
sys = simsizes(sizes);
% State Index:
%
% x(1:Ni-1) = Previous Plant input u - Controller output (Size Ni-1).
% x(Ni) = Actual Plant input u - Controller output (Size 1).
% x(Ni+1:Nu+Ni-1) = Next Plant input u - Controller output (Size Nu-1).
% x(Nu+Ni) = Previous NN 2nd layer output (Size 1).
% x(Nu+Ni+1:Nu+Ni+S1) = Previuos NN 1st layer output (Size S1).
%
% Last two variables will repeat in case of multiple outputs. Not tested yet.
%
x0 = zeros(Ni+Nu-1+(S1+1)*(Nj-1),1);
% ODJ 1-31-00 We place initial Plant input u - Controller output at mid range.
x0(Ni:Nu+Ni-1) = (max_i-min_i)/2;
str = [];
ts = [Ts 0]; % Inherited sample time
tiu=Ni;
dUtilde_dU = eye(Nu);
dUtilde_dU(1:Nu-1,2:Nu)=dUtilde_dU(1:Nu-1,2:Nu)-eye(Nu-1);
N1=1;
d=1;
alpha2 = alpha*alpha;
upi = [1:Nu-1 Nu(ones(1,N2-d-Nu+2))];
uvi = [tiu:N2-N1+Ni];
% end mdlInitializeSizes
%
%=======================================================================
% mdlUpdate
% Handle discrete state updates, sample time hits, and major time step
% requirements.
%=======================================================================
%
function sys = mdlOutputs(t,x,u,Nu,Ni)
sys = x(Ni);
%end mdlUpdate
%
%=======================================================================
% mdlOutputs
% Return the output vector for the S-function
%=======================================================================
%
function sys = mdlUpdate(t,x,u,N2,Ts,Nu,maxiter,csrchfun,rho,alpha,S1,Ni,Nj,min_i,max_i,minp,maxp,mint,maxt,Normalize)
global tiu dUtilde_dU
global N1 d alpha2 upi uvi
Ai=num2cell(zeros(2,Nj));
for k=1:Nj-1
Ai{1,k}=x(Nu+Ni+1+(k-1)*(S1+1):Nu+Ni+S1+(k-1)*(S1+1));
Ai{2,k}=x(Nu+Ni+(k-1)*(S1+1)); % delayed plant output
end
Ai{1,Nj}=u(4:3+S1);
ref(1:N2,1)=u(1);
initval = '[upmin(Nu)]';
upmin=[x(Ni+1:Nu+Ni-1);x(Nu+Ni-1)];
u_vec(1:Ni-1,1)=x(2:Ni);
if Normalize
ref=((ref-mint)*2/(maxt-mint)-1);
Ai{2,Nj}=((u(3)-mint)*2/(maxt-mint)-1); % Actual NN output
upmin=((upmin-minp)*2/(maxp-minp)-1);
u_vec=((u_vec-minp)*2/(maxp-minp)-1);
else
Ai{2,Nj}=u(3);
end
upmin0 = upmin;
einitval = eval(initval); % Evaluate inival string
for tr=1:length(einitval),
up=upmin0; % Initial value for numerical search for a new u
up(Nu)=einitval(tr);
u_vec(uvi,1) = up(upi);
dw = 1; % Flag specifying that up is new
lambda = 0.1; % Initialize Levenberg-Marquardt parameter
%>>>>>>>>>>>>>>> COMPUTE PREDICTIONS FROM TIME t+N1 TO t+N2 <<<<<<<<<<<<<<<<
assignin('base','cont_u',evalin('base','cont_u')+1);
set_param('ptest3sim2/Subsystem','u_init',mat2str(u_vec(Ni),20),'ud_init',mat2str(u_vec(Ni-1:-1:1),20), ...
'y_init',mat2str(Ai{2,Nj},20),'yd_init',mat2str(cat(1,Ai{2,Nj-1:-1:1}),20));
[time,xx0,Ac1,Ac2,E,gU,gUd,dY_dU] = sim('ptest3sim2',[0 N2*Ts],[],[(0:Ts:(N2-2)*Ts)' u_vec(1:N2-1) ref(1:N2-1)]);
yhat_vec=Ac1(1:N2+1,1)';
E=E(2:N2+1,:);
gU=gU(1:N2,:)';
gUd=gUd(1:N2,:)';
evec=E;
if tiu==1
duvec = [0; u_vec(tiu+1:tiu+Nu-1)-u_vec(tiu:tiu+Nu-2)];
else
duvec = u_vec(tiu:tiu+Nu-1)-u_vec(tiu-1:tiu+Nu-2);
end
JJ = evec'*evec + rho*(duvec'*duvec);
% Forward Perturbation
dY_dU=dY_dU(2:N2+1,:)';
dJJ = 2*(-dY_dU*evec + rho*(dUtilde_dU*duvec));
if Normalize
dJJ=dJJ/(maxp-minp);
end
%>>>>>>>>>>>>>>>>>>>>>> EVALUATE CRITERION <<<<<<<<<<<<<<<<<<<<<<
J = JJ;
%>>>>>>>>>>>>>>>>>>>>>>>> DETERMINE dyhat/du <<<<<<<<<<<<<<<<<<<<<<<<<
%>>>>>>>>>>>>>>>>>>>>>>>>>>>> DETERMINE dJ/du <<<<<<<<<<<<<<<<<<<<<<<<<<<<
dJdu = dJJ;
%>>>>>>>>>>>>>>>>>>>>>> DETERMINE INVERSE HESSIAN <<<<<<<<<<<<<<<<<<<<<<<<<
B = eye(Nu); % Initialize Hessian to I
delta=1;
tol=1/delta;
ch_perf = J; % for first iteration.
%>>>>>>>>>>>>>>>>>>>>>>> BEGIN SEARCH FOR MINIMUM <<<<<<<<<<<<<<<<<<<<<<
for m = 1:maxiter,
%>>>>>>>>>>>>>>>>>>>>>>> DETERMINE SEARCH DIRECTION <<<<<<<<<<<<<<<<<<<<<<<
dX = -B*dJdu;
if dX'*dJdu>0 % We reset the gradient if positive.
%>>>>>>>>>>>>>>>>>>>>>> DETERMINE INVERSE HESSIAN <<<<<<<<<<<<<<<<<<<<<<<<<
B = eye(Nu); % Initialize Hessian to I
delta=1;
tol=1/delta;
ch_perf = J; % for first iteration.
%>>>>>>>>>>>>>>>>>>>>>>> DETERMINE SEARCH DIRECTION <<<<<<<<<<<<<<<<<<<<<<<
dX = -B*dJdu;
end
if Normalize
switch csrchfun,
case 1, %'csrchgol',
[up_delta,J,dJdu_old,dJdu,retcode,delta,tol]=csrchgol(up,u_vec,ref,Ai,Nu,N1,N2,d,Ni,Nj,dX,dJdu,J,dX'*dJdu,delta,rho,dUtilde_dU,alpha,tol,Ts,-1,1,Normalize,minp,maxp);
case 2 %'csrchbac',
[up_delta,J,dJdu_old,dJdu,retcode,delta,tol]=csrchbac(up,u_vec,ref,Ai,Nu,N1,N2,d,Ni,Nj,dX,dJdu,J,dX'*dJdu,delta,rho,dUtilde_dU,alpha,tol,Ts,-1,1,Normalize,minp,maxp);
case 3 %'csrchhyb'
[up_delta,J,dJdu_old,dJdu,retcode,delta,tol]=csrchhyb(up,u_vec,ref,Ai,Nu,N1,N2,d,Ni,Nj,dX,dJdu,J,dX'*dJdu,delta,rho,dUtilde_dU,alpha,tol,Ts,-1,1,Normalize,minp,maxp);
case 4 %'csrchbre'
[up_delta,J,dJdu_old,dJdu,retcode,delta,tol]=csrchbre(up,u_vec,ref,Ai,Nu,N1,N2,d,Ni,Nj,dX,dJdu,J,dX'*dJdu,delta,rho,dUtilde_dU,alpha,tol,Ts,-1,1,Normalize,minp,maxp);
case 5 %'csrchcha'
J_old=J;
[up_delta,J,dJdu_old,dJdu,retcode,delta,tol]=csrchcha(up,u_vec,ref,Ai,Nu,N1,N2,d,Ni,Nj,dX,dJdu,J,dX'*dJdu,delta,rho,dUtilde_dU,alpha,tol,Ts,-1,1,Normalize,minp,maxp,ch_perf);
ch_perf = J - J_old;
otherwise
J_old=J;
[up_delta,J,dJdu_old,dJdu,retcode,delta,tol]=feval(csrchfun,up,u_vec,ref,Ai,Nu,N1,N2,d,Ni,Nj,dX,dJdu,J,dX'*dJdu,delta,rho,dUtilde_dU,alpha,tol,Ts,-1,1,Normalize,minp,maxp,ch_perf);
ch_perf = J - J_old;
end
else
switch csrchfun,
case 1, %'csrchgol',
[up_delta,J,dJdu_old,dJdu,retcode,delta,tol]=csrchgol(up,u_vec,ref,Ai,Nu,N1,N2,d,Ni,Nj,dX,dJdu,J,dX'*dJdu,delta,rho,dUtilde_dU,alpha,tol,Ts,min_i,max_i,Normalize,minp,maxp);
case 2 %'csrchbac',
[up_delta,J,dJdu_old,dJdu,retcode,delta,tol]=csrchbac(up,u_vec,ref,Ai,Nu,N1,N2,d,Ni,Nj,dX,dJdu,J,dX'*dJdu,delta,rho,dUtilde_dU,alpha,tol,Ts,min_i,max_i,Normalize,minp,maxp);
case 3 %'csrchhyb'
[up_delta,J,dJdu_old,dJdu,retcode,delta,tol]=csrchhyb(up,u_vec,ref,Ai,Nu,N1,N2,d,Ni,Nj,dX,dJdu,J,dX'*dJdu,delta,rho,dUtilde_dU,alpha,tol,Ts,min_i,max_i,Normalize,minp,maxp);
case 4 %'csrchbre'
[up_delta,J,dJdu_old,dJdu,retcode,delta,tol]=csrchbre(up,u_vec,ref,Ai,Nu,N1,N2,d,Ni,Nj,dX,dJdu,J,dX'*dJdu,delta,rho,dUtilde_dU,alpha,tol,Ts,min_i,max_i,Normalize,minp,maxp);
case 5 %'csrchcha'
J_old=J;
[up_delta,J,dJdu_old,dJdu,retcode,delta,tol]=csrchcha(up,u_vec,ref,Ai,Nu,N1,N2,d,Ni,Nj,dX,dJdu,J,dX'*dJdu,delta,rho,dUtilde_dU,alpha,tol,Ts,min_i,max_i,Normalize,minp,maxp,ch_perf);
ch_perf = J - J_old;
otherwise
J_old=J;
[up_delta,J,dJdu_old,dJdu,retcode,delta,tol]=feval(csrchfun,up,u_vec,ref,Ai,Nu,N1,N2,d,Ni,Nj,dX,dJdu,J,dX'*dJdu,delta,rho,dUtilde_dU,alpha,tol,Ts,min_i,max_i,Normalize,minp,maxp,ch_perf);
ch_perf = J - J_old;
end
end
%>>>>>>>>>>>>>>>>>>>>>>>> UPDATE FUTURE CONTROLS <<<<<<<<<<<<<<<<<<<<<<<<<
up_old = up;
up = up_delta;
%>>>>>>>>>>>>>>>>>>>>>>>> CHECK STOP CONDITION <<<<<<<<<<<<<<<<<<<<<<<
dup = up-up_old;
if (dup'*dup < alpha2) | (ch_perf==0),
break;
end
%>>>>>>>>>>>>>>>>>>> BFGS UPDATE OF INVERSE HESSIAN <<<<<<<<<<<<<<<<<<
dG = dJdu - dJdu_old;
BdG = B*dG;
dupdG = dup'*dG;
fac = 1/dupdG;
diff = dup - BdG;
dupfac=dup*fac;
diffdup = diff*(dupfac');
B = B + diffdup + diffdup' - (diff'*dG)*(dupfac*dupfac');
end
%>>>>>>>>>>>>>>>>>>>>>>> SELECT BEST MINIMUM <<<<<<<<<<<<<<<<<<<<<<<<<
if tr==1,
Jmin_old = J;
upmin = up;
else
if J<Jmin_old,
upmin = up;
end
end
end
x(1:Ni-1)=x(2:Ni); % State 1 to Nu = actual controls
if upmin(1)>1 | upmin(1)<-1
upmin(1)=upmin(1);
end
if Normalize
upmin=(upmin+1)*(maxp-minp)/2+minp;
end
x(Ni:Nu+Ni-1)=upmin; % State 1 to Nu = actual controls
for k=1:Nj-2
x(Nu+Ni+1+(k-1)*(S1+1):Nu+Ni+S1+(k-1)*(S1+1))=x(Nu+Ni+1+(k)*(S1+1):Nu+Ni+S1+(k)*(S1+1));
x(Nu+Ni+(k-1)*(S1+1))=x(Nu+Ni+(k)*(S1+1)); % delayed plant output
end
if Nj>=2
if Normalize
x(Nu+Ni+(Nj-2)*(S1+1))=((u(3)-mint)*2/(maxt-mint)-1); % state Nu+1 = NN output
else
x(Nu+Ni+(Nj-2)*(S1+1))=u(2);
end
x(Nu+Ni+1+(Nj-2)*(S1+1):Nu+Ni+S1+(Nj-2)*(S1+1))=Ai{1,Nj}; % State Nu+2... = delayed layer 1 output.
end
sys=x;
%end mdlUpdate
x(Nu+Ni+1+(Nj-2)*(S1+1):Nu+Ni+S1+(Nj-2)*(S1+1))=Ai{1,Nj}; % State Nu+2... = delayed layer 1 output.
sys=x;
%end mdlUpdate
?? 快捷鍵說明
復(fù)制代碼
Ctrl + C
搜索代碼
Ctrl + F
全屏模式
F11
切換主題
Ctrl + Shift + D
顯示快捷鍵
?
增大字號
Ctrl + =
減小字號
Ctrl + -