220 lines
7.6 KiB
Mathematica
220 lines
7.6 KiB
Mathematica
|
function output = calllindo_nlp(interfacedata)
|
||
|
|
|
||
|
|
global MY_LICENSE_FILE
|
||
|
|
|
||
|
|
persistent iEnv
|
||
|
|
|
||
|
|
% Instead of calling lindo, we define the parameters we need. This is
|
||
|
|
% needed to speed up repeated calls
|
||
|
|
%lindo
|
||
|
|
LSERR_NO_ERROR = 0000;
|
||
|
|
LS_IPARAM_NLP_PRINTLEVEL = 203;
|
||
|
|
LS_IPARAM_NLP_SOLVER = 201;
|
||
|
|
LS_IPARAM_NLP_MAXLOCALSEARCH = 221;
|
||
|
|
LS_STATUS_OPTIMAL = 1;
|
||
|
|
LS_STATUS_BASIC_OPTIMAL = 2;
|
||
|
|
LS_STATUS_INFEASIBLE = 3;
|
||
|
|
LS_STATUS_UNBOUNDED = 4;
|
||
|
|
LS_STATUS_FEASIBLE = 5;
|
||
|
|
LS_STATUS_INFORUNB = 6;
|
||
|
|
LS_STATUS_NEAR_OPTIMAL = 7;
|
||
|
|
LS_STATUS_LOCAL_OPTIMAL = 8;
|
||
|
|
LS_STATUS_LOCAL_INFEASIBLE = 9;
|
||
|
|
LS_STATUS_CUTOFF = 10;
|
||
|
|
LS_STATUS_NUMERICAL_ERROR = 11;
|
||
|
|
LS_STATUS_UNKNOWN = 12;
|
||
|
|
LS_STATUS_UNLOADED = 13;
|
||
|
|
LS_STATUS_LOADED = 14;
|
||
|
|
LS_METHOD_FREE = 0;
|
||
|
|
LS_METHOD_PSIMPLEX = 1;
|
||
|
|
LS_METHOD_DSIMPLEX = 2;
|
||
|
|
LS_METHOD_BARRIER = 3;
|
||
|
|
LS_METHOD_NLP = 4;
|
||
|
|
LS_NMETHOD_FREE = 4;
|
||
|
|
LS_NMETHOD_CONOPT = 7;
|
||
|
|
LS_NMETHOD_MSW_GRG = 9;
|
||
|
|
|
||
|
|
if isempty(iEnv)
|
||
|
|
% This call is mighty slow, so we do it only once, unless uses clears
|
||
|
|
% everything
|
||
|
|
[MY_LICENSE_KEY,Err] = mxlindo('LSloadLicenseString',MY_LICENSE_FILE);
|
||
|
|
[iEnv,nErr]=mxlindo('LScreateEnv',MY_LICENSE_KEY);
|
||
|
|
if nErr ~= LSERR_NO_ERROR;output = returnempty(-5); return; end;
|
||
|
|
end
|
||
|
|
|
||
|
|
% Retrieve needed data
|
||
|
|
options = interfacedata.options;
|
||
|
|
F_struc = interfacedata.F_struc;
|
||
|
|
c = interfacedata.c;
|
||
|
|
K = interfacedata.K;
|
||
|
|
x0 = interfacedata.x0;
|
||
|
|
Q = interfacedata.Q;
|
||
|
|
lb = interfacedata.lb;
|
||
|
|
ub = interfacedata.ub;
|
||
|
|
monomtable = interfacedata.monomtable;
|
||
|
|
|
||
|
|
% Do some pre-calc to be used in callbacks
|
||
|
|
nonlinearindicies = find(interfacedata.variabletype~=0);
|
||
|
|
nonlinearindicies = unionstripped(nonlinearindicies,interfacedata.evalVariables);
|
||
|
|
linearindicies = find(interfacedata.variabletype==0);
|
||
|
|
linearindicies = setdiff1D(linearindicies,nonlinearindicies);
|
||
|
|
interfacedata.nonlinearindicies = nonlinearindicies;
|
||
|
|
interfacedata.linearindicies = linearindicies;
|
||
|
|
|
||
|
|
% Init model size
|
||
|
|
m = K.l + K.f;
|
||
|
|
n = length(c);
|
||
|
|
|
||
|
|
% Specifying variable types...
|
||
|
|
vtype = repmat('C',1,length(c(linearindicies)));
|
||
|
|
vtype(interfacedata.integer_variables) = 'I';
|
||
|
|
|
||
|
|
oshift = interfacedata.f;
|
||
|
|
|
||
|
|
if m == 0
|
||
|
|
interfacedata.F_struc = [1e6 -ones(1,length(c))];
|
||
|
|
K.l = 1;
|
||
|
|
F_struc = [1e6 -ones(1,length(c))];
|
||
|
|
m = 1;
|
||
|
|
csense = [repmat('E',1,K.f) repmat('L',1,K.l)];
|
||
|
|
end
|
||
|
|
|
||
|
|
[Nbegcol,Nlencol,Nrowndx] = lindosparse(ones(m,length(linearindicies)));
|
||
|
|
oJacobian = ones(length(linearindicies),1);
|
||
|
|
Nobjndx = find(oJacobian) - 1;
|
||
|
|
Nobjcnt = length(Nobjndx);
|
||
|
|
if isempty(Nobjndx)
|
||
|
|
Nobjndx = [];
|
||
|
|
end
|
||
|
|
|
||
|
|
%[Nbegcol,Nlencol,Nrowndx,Nobjcnt,Nobjndx,Apatt] = jacSparsityGeometric(interfacedata);
|
||
|
|
% A = -F_struc(:,1+linearindicies);
|
||
|
|
% b = full(F_struc(:,1));
|
||
|
|
csense = [repmat('E',1,K.f) repmat('L',1,K.l)];
|
||
|
|
% A = A.*(~Apatt);
|
||
|
|
% b(any(Apatt,2)) = 0;
|
||
|
|
|
||
|
|
[iModel,nErr]=mxlindo('LScreateModel',iEnv);
|
||
|
|
if nErr ~= LSERR_NO_ERROR;output = returnempty(11); return; end;
|
||
|
|
|
||
|
|
% SETUP data for callbacks
|
||
|
|
mt = interfacedata.monomtable;
|
||
|
|
linear_variables = find((sum(abs(mt),2)==1) & (any(mt==1,2)));
|
||
|
|
nonlinear_variables = setdiff((1:size(mt,1))',linear_variables);
|
||
|
|
sigmonial_variables = find(any(0>mt,2) | any(mt-fix(mt),2));
|
||
|
|
extended_variables = interfacedata.extended_variables;
|
||
|
|
|
||
|
|
[prob,problem] = yalmip2geometric(options,F_struc,c,Q,K,ub,lb,mt,linear_variables,extended_variables);
|
||
|
|
prob.interfacedata = interfacedata;
|
||
|
|
if problem
|
||
|
|
output = createoutput([],[],[],problem,'LINDO',[],[],0);
|
||
|
|
return
|
||
|
|
end
|
||
|
|
lindo_fungp([],[],[],[],[],[],prob);
|
||
|
|
|
||
|
|
[nErr] = mxlindo('LSsetFuncalc', iModel, 'lindo_fungp',prob);
|
||
|
|
if nErr ~= LSERR_NO_ERROR;output = returnempty(11); return; end;
|
||
|
|
|
||
|
|
[nErr] = mxlindo('LSsetModelIntParameter', iModel, LS_IPARAM_NLP_PRINTLEVEL, options.verbose+1);
|
||
|
|
if nErr ~= LSERR_NO_ERROR;output = returnempty(11); return; end;
|
||
|
|
|
||
|
|
% Set NLP solver
|
||
|
|
[nErr] = mxlindo('LSsetModelIntParameter', iModel, LS_IPARAM_NLP_SOLVER, eval(options.lindo.LS_IPARAM_NLP_SOLVER));
|
||
|
|
[nErr] = mxlindo('LSsetModelIntParameter', iModel, LS_IPARAM_NLP_MAXLOCALSEARCH,options.lindo.LS_IPARAM_NLP_MAXLOCALSEARCH);
|
||
|
|
|
||
|
|
% Load the LP portion of model
|
||
|
|
if ~isempty(lb)
|
||
|
|
lb = lb(linear_variables);
|
||
|
|
ind = find(lb<1e-2);
|
||
|
|
lb(ind) = exp(log(1e-2)+(lb(ind)-1e-2)/1e-2);
|
||
|
|
lb = log(lb+sqrt(eps));
|
||
|
|
end
|
||
|
|
if ~isempty(ub)
|
||
|
|
ub = ub(linear_variables);
|
||
|
|
ind = find(ub<1e-2);
|
||
|
|
ub(ind) = exp(log(1e-2)+(ub(ind)-1e-2)/1e-2);
|
||
|
|
ub = log(ub+sqrt(eps));
|
||
|
|
end
|
||
|
|
b = zeros(m,1);
|
||
|
|
A = spalloc(m,length(linearindicies),0);
|
||
|
|
[nErr] = mxlindo('LSXloadLPData', iModel, 1, 0, zeros(length(linearindicies),1),b, csense,A, lb, ub);
|
||
|
|
if nErr ~= LSERR_NO_ERROR;output = createoutput(11); return; end;
|
||
|
|
|
||
|
|
nErr = mxlindo('LSloadVarType',iModel,vtype);
|
||
|
|
if nErr ~= LSERR_NO_ERROR;output = createoutput(11); return; end;
|
||
|
|
|
||
|
|
% Load the NLP portion of the model
|
||
|
|
[nErr] = mxlindo('LSloadNLPData', iModel, Nbegcol, Nlencol,[], Nrowndx, Nobjcnt,Nobjndx,[]);
|
||
|
|
if nErr ~= LSERR_NO_ERROR;output = createoutput(11); return; end;
|
||
|
|
|
||
|
|
% Optimize model
|
||
|
|
solvertime = clock;
|
||
|
|
|
||
|
|
if isempty(interfacedata.integer_variables)
|
||
|
|
solver = 2;
|
||
|
|
else
|
||
|
|
solver = 1;
|
||
|
|
end
|
||
|
|
solvertime = tic;
|
||
|
|
switch solver
|
||
|
|
case 1
|
||
|
|
[solstat,nErr] = mxlindo('LSsolveMIP', iModel);
|
||
|
|
if ~ismember(solstat,[2009 LS_STATUS_INFEASIBLE])
|
||
|
|
[x,nErr] = mxlindo('LSgetMIPPrimalSolution',iModel);
|
||
|
|
else
|
||
|
|
x = zeros(length(linearindicies),1);
|
||
|
|
end
|
||
|
|
case 2
|
||
|
|
[solstat,nErr] = mxlindo('LSoptimize', iModel, eval(options.lindo.LS_METHOD));
|
||
|
|
if ~ismember(solstat,[2009 LS_STATUS_INFEASIBLE])
|
||
|
|
[x,nErr] = mxlindo('LSgetPrimalSolution',iModel);
|
||
|
|
else
|
||
|
|
x = zeros(length(linearindicies),1);
|
||
|
|
end
|
||
|
|
case 3
|
||
|
|
[solStatus,nErr] = mxlindo('LSsolveGOP', iModel);
|
||
|
|
[x,nErr] = mxlindo('LSgetPrimalSolution',iModel);
|
||
|
|
otherwise
|
||
|
|
end
|
||
|
|
solvertime = toc(solvertime);
|
||
|
|
|
||
|
|
w = zeros(length(c),1);w(linearindicies) = exp(x);
|
||
|
|
y = [];
|
||
|
|
|
||
|
|
%[nErr]=mxlindo('LSdeleteEnv',iEnv);
|
||
|
|
[nErr]=mxlindo('LSdeleteModel',iModel);
|
||
|
|
|
||
|
|
switch solstat
|
||
|
|
case {LS_STATUS_OPTIMAL,LS_STATUS_BASIC_OPTIMAL,7,8}
|
||
|
|
problem = 0;
|
||
|
|
case {LS_STATUS_INFEASIBLE}
|
||
|
|
problem = 1;
|
||
|
|
case {LS_STATUS_UNBOUNDED}
|
||
|
|
problem = 2;
|
||
|
|
otherwise
|
||
|
|
problem = 11;
|
||
|
|
end
|
||
|
|
infostr = yalmiperror(problem,'LINDO-QP');
|
||
|
|
|
||
|
|
% Save all data sent to solver?
|
||
|
|
if options.savesolverinput
|
||
|
|
solverinput.solstat = solstat;
|
||
|
|
solverinput.nErr = nErr;
|
||
|
|
solverinput.x = x;
|
||
|
|
else
|
||
|
|
solverinput = [];
|
||
|
|
end
|
||
|
|
|
||
|
|
% Save all data from the solver?
|
||
|
|
if options.savesolveroutput
|
||
|
|
solveroutput.x = x;
|
||
|
|
solveroutput.fmin = fmin;
|
||
|
|
solveroutput.flag = flag;
|
||
|
|
solveroutput.output=output;
|
||
|
|
solveroutput.lambda=lambda;
|
||
|
|
else
|
||
|
|
solveroutput = [];
|
||
|
|
end
|
||
|
|
|
||
|
|
% Standard interface
|
||
|
|
output = createoutput(w,y,[],problem,'LINDO',solverinput,solveroutput,solvertime);
|