void Sqpmethod::init() {
// Call the init method of the base class
NlpSolverInternal::init();
// Read options
max_iter_ = getOption("max_iter");
max_iter_ls_ = getOption("max_iter_ls");
c1_ = getOption("c1");
beta_ = getOption("beta");
merit_memsize_ = getOption("merit_memory");
lbfgs_memory_ = getOption("lbfgs_memory");
tol_pr_ = getOption("tol_pr");
tol_du_ = getOption("tol_du");
regularize_ = getOption("regularize");
exact_hessian_ = getOption("hessian_approximation")=="exact";
min_step_size_ = getOption("min_step_size");
// Get/generate required functions
gradF();
jacG();
if (exact_hessian_) {
hessLag();
}
// Allocate a QP solver
Sparsity H_sparsity = exact_hessian_ ? hessLag().output().sparsity()
: Sparsity::dense(nx_, nx_);
H_sparsity = H_sparsity + Sparsity::diag(nx_);
Sparsity A_sparsity = jacG().isNull() ? Sparsity(0, nx_)
: jacG().output().sparsity();
// QP solver options
Dict qp_solver_options;
if (hasSetOption("qp_solver_options")) {
qp_solver_options = getOption("qp_solver_options");
}
// Allocate a QP solver
qp_solver_ = QpSolver("qp_solver", getOption("qp_solver"),
make_map("h", H_sparsity, "a", A_sparsity),
qp_solver_options);
// Lagrange multipliers of the NLP
mu_.resize(ng_);
mu_x_.resize(nx_);
// Lagrange gradient in the next iterate
gLag_.resize(nx_);
gLag_old_.resize(nx_);
// Current linearization point
x_.resize(nx_);
x_cand_.resize(nx_);
x_old_.resize(nx_);
// Constraint function value
gk_.resize(ng_);
gk_cand_.resize(ng_);
// Hessian approximation
Bk_ = DMatrix::zeros(H_sparsity);
// Jacobian
Jk_ = DMatrix::zeros(A_sparsity);
// Bounds of the QP
qp_LBA_.resize(ng_);
qp_UBA_.resize(ng_);
qp_LBX_.resize(nx_);
qp_UBX_.resize(nx_);
// QP solution
dx_.resize(nx_);
qp_DUAL_X_.resize(nx_);
qp_DUAL_A_.resize(ng_);
// Gradient of the objective
gf_.resize(nx_);
// Create Hessian update function
if (!exact_hessian_) {
// Create expressions corresponding to Bk, x, x_old, gLag and gLag_old
SX Bk = SX::sym("Bk", H_sparsity);
SX x = SX::sym("x", input(NLP_SOLVER_X0).sparsity());
SX x_old = SX::sym("x", x.sparsity());
SX gLag = SX::sym("gLag", x.sparsity());
SX gLag_old = SX::sym("gLag_old", x.sparsity());
SX sk = x - x_old;
SX yk = gLag - gLag_old;
SX qk = mul(Bk, sk);
// Calculating theta
SX skBksk = inner_prod(sk, qk);
SX omega = if_else(inner_prod(yk, sk) < 0.2 * inner_prod(sk, qk),
0.8 * skBksk / (skBksk - inner_prod(sk, yk)),
1);
yk = omega * yk + (1 - omega) * qk;
SX theta = 1. / inner_prod(sk, yk);
SX phi = 1. / inner_prod(qk, sk);
SX Bk_new = Bk + theta * mul(yk, yk.T()) - phi * mul(qk, qk.T());
// Inputs of the BFGS update function
vector<SX> bfgs_in(BFGS_NUM_IN);
bfgs_in[BFGS_BK] = Bk;
bfgs_in[BFGS_X] = x;
bfgs_in[BFGS_X_OLD] = x_old;
bfgs_in[BFGS_GLAG] = gLag;
bfgs_in[BFGS_GLAG_OLD] = gLag_old;
bfgs_ = SXFunction("bfgs", bfgs_in, make_vector(Bk_new));
// Initial Hessian approximation
B_init_ = DMatrix::eye(nx_);
}
// Header
if (static_cast<bool>(getOption("print_header"))) {
userOut()
<< "-------------------------------------------" << endl
<< "This is casadi::SQPMethod." << endl;
if (exact_hessian_) {
userOut() << "Using exact Hessian" << endl;
} else {
userOut() << "Using limited memory BFGS Hessian approximation" << endl;
}
userOut()
<< endl
<< "Number of variables: " << setw(9) << nx_ << endl
<< "Number of constraints: " << setw(9) << ng_ << endl
<< "Number of nonzeros in constraint Jacobian: " << setw(9) << A_sparsity.nnz() << endl
<< "Number of nonzeros in Lagrangian Hessian: " << setw(9) << H_sparsity.nnz() << endl
<< endl;
}
}
void SQPInternal::init(){
// Call the init method of the base class
NLPSolverInternal::init();
// Read options
maxiter_ = getOption("maxiter");
maxiter_ls_ = getOption("maxiter_ls");
c1_ = getOption("c1");
beta_ = getOption("beta");
merit_memsize_ = getOption("merit_memory");
lbfgs_memory_ = getOption("lbfgs_memory");
tol_pr_ = getOption("tol_pr");
tol_du_ = getOption("tol_du");
regularize_ = getOption("regularize");
if(getOption("hessian_approximation")=="exact")
hess_mode_ = HESS_EXACT;
else if(getOption("hessian_approximation")=="limited-memory")
hess_mode_ = HESS_BFGS;
if (hess_mode_== HESS_EXACT && H_.isNull()) {
if (!getOption("generate_hessian")){
casadi_error("SQPInternal::evaluate: you set option 'hessian_approximation' to 'exact', but no hessian was supplied. Try with option \"generate_hessian\".");
}
}
// If the Hessian is generated, we use exact approximation by default
if (bool(getOption("generate_hessian"))){
setOption("hessian_approximation", "exact");
}
// Allocate a QP solver
CRSSparsity H_sparsity = hess_mode_==HESS_EXACT ? H_.output().sparsity() : sp_dense(n_,n_);
H_sparsity = H_sparsity + DMatrix::eye(n_).sparsity();
CRSSparsity A_sparsity = J_.isNull() ? CRSSparsity(0,n_,false) : J_.output().sparsity();
QPSolverCreator qp_solver_creator = getOption("qp_solver");
qp_solver_ = qp_solver_creator(H_sparsity,A_sparsity);
// Set options if provided
if(hasSetOption("qp_solver_options")){
Dictionary qp_solver_options = getOption("qp_solver_options");
qp_solver_.setOption(qp_solver_options);
}
qp_solver_.init();
// Lagrange multipliers of the NLP
mu_.resize(m_);
mu_x_.resize(n_);
// Lagrange gradient in the next iterate
gLag_.resize(n_);
gLag_old_.resize(n_);
// Current linearization point
x_.resize(n_);
x_cand_.resize(n_);
x_old_.resize(n_);
// Constraint function value
gk_.resize(m_);
gk_cand_.resize(m_);
// Hessian approximation
Bk_ = DMatrix(H_sparsity);
// Jacobian
Jk_ = DMatrix(A_sparsity);
// Bounds of the QP
qp_LBA_.resize(m_);
qp_UBA_.resize(m_);
qp_LBX_.resize(n_);
qp_UBX_.resize(n_);
// QP solution
dx_.resize(n_);
qp_DUAL_X_.resize(n_);
qp_DUAL_A_.resize(m_);
// Gradient of the objective
gf_.resize(n_);
// Create Hessian update function
if(hess_mode_ == HESS_BFGS){
// Create expressions corresponding to Bk, x, x_old, gLag and gLag_old
SXMatrix Bk = ssym("Bk",H_sparsity);
SXMatrix x = ssym("x",input(NLP_X_INIT).sparsity());
SXMatrix x_old = ssym("x",x.sparsity());
SXMatrix gLag = ssym("gLag",x.sparsity());
SXMatrix gLag_old = ssym("gLag_old",x.sparsity());
SXMatrix sk = x - x_old;
SXMatrix yk = gLag - gLag_old;
SXMatrix qk = mul(Bk, sk);
// Calculating theta
SXMatrix skBksk = inner_prod(sk, qk);
SXMatrix omega = if_else(inner_prod(yk, sk) < 0.2 * inner_prod(sk, qk),
0.8 * skBksk / (skBksk - inner_prod(sk, yk)),
1);
yk = omega * yk + (1 - omega) * qk;
SXMatrix theta = 1. / inner_prod(sk, yk);
SXMatrix phi = 1. / inner_prod(qk, sk);
SXMatrix Bk_new = Bk + theta * mul(yk, trans(yk)) - phi * mul(qk, trans(qk));
// Inputs of the BFGS update function
vector<SXMatrix> bfgs_in(BFGS_NUM_IN);
bfgs_in[BFGS_BK] = Bk;
bfgs_in[BFGS_X] = x;
bfgs_in[BFGS_X_OLD] = x_old;
bfgs_in[BFGS_GLAG] = gLag;
bfgs_in[BFGS_GLAG_OLD] = gLag_old;
bfgs_ = SXFunction(bfgs_in,Bk_new);
bfgs_.setOption("number_of_fwd_dir",0);
bfgs_.setOption("number_of_adj_dir",0);
bfgs_.init();
// Initial Hessian approximation
B_init_ = DMatrix::eye(n_);
}
// Header
if(bool(getOption("print_header"))){
cout << "-------------------------------------------" << endl;
cout << "This is CasADi::SQPMethod." << endl;
switch (hess_mode_) {
case HESS_EXACT:
cout << "Using exact Hessian" << endl;
break;
case HESS_BFGS:
cout << "Using limited memory BFGS Hessian approximation" << endl;
break;
}
cout << endl;
cout << "Number of variables: " << setw(9) << n_ << endl;
cout << "Number of constraints: " << setw(9) << m_ << endl;
cout << "Number of nonzeros in constraint Jacobian: " << setw(9) << A_sparsity.size() << endl;
cout << "Number of nonzeros in Lagrangian Hessian: " << setw(9) << H_sparsity.size() << endl;
cout << endl;
}
}
