123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290291292293294295296297298299300301302303304305306307308309310311312313314315316317318319320321322323324325326327328329330331332333334335336337 |
- // Copyright (C) 2005, 2006 International Business Machines and others.
- // All Rights Reserved.
- // This code is published under the Eclipse Public License.
- //
- // $Id: LuksanVlcek3.cpp 2005 2011-06-06 12:55:16Z stefan $
- //
- // Authors: Andreas Waechter IBM 2005-10-127
- #include "LuksanVlcek3.hpp"
- #ifdef HAVE_CONFIG_H
- #include "config.h"
- #else
- #include "configall_system.h"
- #endif
- #ifdef HAVE_CMATH
- # include <cmath>
- #else
- # ifdef HAVE_MATH_H
- # include <math.h>
- # else
- # error "don't have header file for math"
- # endif
- #endif
- #ifdef HAVE_CSTDIO
- # include <cstdio>
- #else
- # ifdef HAVE_STDIO_H
- # include <stdio.h>
- # else
- # error "don't have header file for stdio"
- # endif
- #endif
- using namespace Ipopt;
- LuksanVlcek3::LuksanVlcek3(Number g_l, Number g_u)
- {
- g_l_ = g_l;
- g_u_ = g_u;
- }
- bool LuksanVlcek3::InitializeProblem(Index N)
- {
- N_=N;
- if (N_<=5 || 4*((N_+2)/4)!=N_+2) {
- printf("N needs to be at least 6 and N+2 needs to be a multiple of 4.\n");
- return false;
- }
- return true;
- }
- // returns the size of the problem
- bool LuksanVlcek3::get_nlp_info(Index& n, Index& m, Index& nnz_jac_g,
- Index& nnz_h_lag, IndexStyleEnum& index_style)
- {
- // The problem described in LuksanVlcek3.hpp has 4 variables, x[0] through x[3]
- n = N_+2;
- m = 2;
- nnz_jac_g = 4;
- nnz_h_lag = 5*N_/2 + 3;
- // use the C style numbering of matrix indices (starting at 0)
- index_style = TNLP::C_STYLE;
- return true;
- }
- // returns the variable bounds
- bool LuksanVlcek3::get_bounds_info(Index n, Number* x_l, Number* x_u,
- Index m, Number* g_l, Number* g_u)
- {
- // none of the variables have bounds
- for (Index i=0; i<n; i++) {
- x_l[i] = -1e20;
- x_u[i] = 1e20;
- }
- // Set the bounds for the constraints
- for (Index i=0; i<m; i++) {
- g_l[i] = g_l_;
- g_u[i] = g_u_;
- }
- return true;
- }
- // returns the initial point for the problem
- bool LuksanVlcek3::get_starting_point(Index n, bool init_x, Number* x,
- bool init_z, Number* z_L, Number* z_U,
- Index m, bool init_lambda,
- Number* lambda)
- {
- if (!init_x || init_z || init_lambda) {
- return false;
- }
- // set the starting point
- for (Index i=0; i<n/4; i++) {
- x[4*i] = 3.;
- x[4*i+1] = -1.;
- x[4*i+2] = 0.;
- x[4*i+3] = 1.;
- }
- /*
- // DELETEME
- for (Index i=0; i<n; i++) {
- x[i] += 0.1*((Number) i);
- }
- */
- return true;
- }
- // returns the value of the objective function
- bool LuksanVlcek3::eval_f(Index n, const Number* x, bool new_x, Number& obj_value)
- {
- obj_value = 0.;
- for (Index i=0; i<N_/2; i++) {
- Number a1 = x[2*i]+10.*x[2*i+1];
- Number a2 = x[2*i+2] - x[2*i+3];
- Number a3 = x[2*i+1] - 2.*x[2*i+2];
- Number a4 = x[2*i] - x[2*i+3];
- obj_value += a1*a1 + 5.*a2*a2 + pow(a3,4)+ 10.*pow(a4,4);
- }
- return true;
- }
- // return the gradient of the objective function grad_{x} f(x)
- bool LuksanVlcek3::eval_grad_f(Index n, const Number* x, bool new_x, Number* grad_f)
- {
- grad_f[0] = 0.;
- grad_f[1] = 0.;
- for (Index i=0; i<N_/2; i++) {
- Number a1 = x[2*i]+10.*x[2*i+1];
- Number a2 = x[2*i+2] - x[2*i+3];
- Number a3 = x[2*i+1] - 2.*x[2*i+2];
- Number a4 = x[2*i] - x[2*i+3];
- grad_f[2*i] += 2.*a1 + 40.*pow(a4,3);
- grad_f[2*i+1] += 20.*a1 + 4.*pow(a3,3);
- grad_f[2*i+2] = 10.*a2 - 8.*pow(a3,3);
- grad_f[2*i+3] = -10.*a2 - 40.*pow(a4,3);
- }
- return true;
- }
- // return the value of the constraints: g(x)
- bool LuksanVlcek3::eval_g(Index n, const Number* x, bool new_x, Index m, Number* g)
- {
- g[0] = 3.*pow(x[0],3) + 2.*x[1] - 5. + sin(x[0]-x[1])*sin(x[0]+x[1]);
- g[1] = 4.*x[n-3] - x[n-4]*exp(x[n-4]-x[n-3]) - 3;
- ;
- return true;
- }
- // return the structure or values of the jacobian
- bool LuksanVlcek3::eval_jac_g(Index n, const Number* x, bool new_x,
- Index m, Index nele_jac, Index* iRow, Index *jCol,
- Number* values)
- {
- if (values == NULL) {
- // return the structure of the jacobian
- Index ijac = 0;
- iRow[ijac] = 0;
- jCol[ijac] = 0;
- ijac++;
- iRow[ijac] = 0;
- jCol[ijac] = 1;
- ijac++;
- iRow[ijac] = 1;
- jCol[ijac] = n-4;
- ijac++;
- iRow[ijac] = 1;
- jCol[ijac] = n-3;
- ijac++;
- DBG_ASSERT(ijac == nele_jac);
- }
- else {
- // return the values of the jacobian of the constraints
- Index ijac = 0;
- values[ijac] = 9.*x[0]*x[0]
- + cos(x[0]-x[1])*sin(x[0]+x[1])
- + sin(x[0]-x[1])*cos(x[0]+x[1]);
- ijac++;
- values[ijac] = 2.
- - cos(x[0]-x[1])*sin(x[0]+x[1])
- + sin(x[0]-x[1])*cos(x[0]+x[1]);
- ijac++;
- values[ijac] = -(1.+x[n-4])*exp(x[n-4]-x[n-3]);
- ijac++;
- values[ijac] = 4. + x[n-4]*exp(x[n-4]-x[n-3]);
- ijac++;
- }
- return true;
- }
- //return the structure or values of the hessian
- bool LuksanVlcek3::eval_h(Index n, const Number* x, bool new_x,
- Number obj_factor, Index m, const Number* lambda,
- bool new_lambda, Index nele_hess, Index* iRow,
- Index* jCol, Number* values)
- {
- if (values == NULL) {
- Index ihes=0;
- for (Index i=0; i<N_/2; i++) {
- iRow[ihes] = 2*i;
- jCol[ihes] = 2*i;
- ihes++;
- iRow[ihes] = 2*i;
- jCol[ihes] = 2*i+1;
- ihes++;
- iRow[ihes] = 2*i;
- jCol[ihes] = 2*i+3;
- ihes++;
- iRow[ihes] = 2*i+1;
- jCol[ihes] = 2*i+1;
- ihes++;
- iRow[ihes] = 2*i+1;
- jCol[ihes] = 2*i+2;
- ihes++;
- }
- iRow[ihes] = N_;
- jCol[ihes] = N_;
- ihes++;
- iRow[ihes] = N_;
- jCol[ihes] = N_+1;
- ihes++;
- iRow[ihes] = N_+1;
- jCol[ihes] = N_+1;
- ihes++;
- DBG_ASSERT(ihes == nele_hess);
- }
- else {
- Index ihes=0;
- values[0] = 0.;
- values[1] = 0.;
- values[3] = 0.;
- for (Index i=0; i<N_/2-1; i++) {
- Number a3 = x[2*i+1] - 2.*x[2*i+2];
- Number a4 = x[2*i] - x[2*i+3];
- // x[2*i] x[2*i]
- values[ihes] += obj_factor*(2. + 120.*a4*a4);
- ihes++;
- // x[2*i] x[2*i+1]
- values[ihes] += obj_factor*20.;
- ihes++;
- // x[2*i] x[2*i+3]
- values[ihes] = -obj_factor*120.*a4*a4;
- ihes++;
- // x[2*i+1] x[2*i+1]
- values[ihes] += obj_factor*(200. + 12.*a3*a3);
- ihes++;
- // x[2*i+1] x[2*i+2]
- values[ihes] = -obj_factor*24.*a3*a3;
- ihes++;
- // x[2*i+2] x[2*i+2]
- values[ihes] = obj_factor*(10. + 48.*a3*a3);
- // x[2*i+2] x[2*i+3]
- values[ihes+1] = -obj_factor*10.;
- // x[2*i+3] x[2*i+3]
- values[ihes+3] = obj_factor*(10. + 120.*a4*a4);
- }
- {
- Index i = N_/2-1;
- Number a3 = x[2*i+1] - 2.*x[2*i+2];
- Number a4 = x[2*i] - x[2*i+3];
- // x[2*i] x[2*i]
- values[ihes] += obj_factor*(2. + 120.*a4*a4);
- ihes++;
- // x[2*i] x[2*i+1]
- values[ihes] += obj_factor*20.;
- ihes++;
- // x[2*i] x[2*i+3]
- values[ihes] = -obj_factor*120.*a4*a4;
- ihes++;
- // x[2*i+1] x[2*i+1]
- values[ihes] += obj_factor*(200. + 12.*a3*a3);
- ihes++;
- // x[2*i+1] x[2*i+2]
- values[ihes] = -obj_factor*24.*a3*a3;
- ihes++;
- // x[2*i+2] x[2*i+2]
- values[ihes] = obj_factor*(10. + 48.*a3*a3);
- // x[2*i+2] x[2*i+3]
- values[ihes+1] = -obj_factor*10.;
- // x[2*i+3] x[2*i+3]
- values[ihes+2] = obj_factor*(10. + 120.*a4*a4);
- }
- // Now the constraints
- ihes = 0;
- Number d1 = x[0] - x[1];
- Number d2 = x[0] + x[1];
- values[ihes] += lambda[0]*(18.*x[0]
- -2.*sin(d1)*sin(d2)
- +2.*cos(d1)*cos(d2));
- ihes+=3;
- values[ihes] += lambda[0]*(-2.*sin(d1)*sin(d2)
- -2.*cos(d1)*cos(d2));
- d1 = x[n-4]-x[n-3];
- // x[n-4] x[n-4]
- ihes = nele_hess - 8;
- values[ihes] += -lambda[1]*(2.+x[n-4])*exp(d1);
- // x[n-4] x[n-3]
- ihes++;
- values[ihes] += lambda[1]*(1.+x[n-4])*exp(d1);
- // x[n-3] x[n-3]
- ihes += 2;
- values[ihes] += -lambda[1]*x[n-4]*exp(d1);
- }
- return true;
- }
- void LuksanVlcek3::finalize_solution(SolverReturn status,
- Index n, const Number* x, const Number* z_L, const Number* z_U,
- Index m, const Number* g, const Number* lambda,
- Number obj_value,
- const IpoptData* ip_data,
- IpoptCalculatedQuantities* ip_cq)
- {}
|