1 // Copyright (C) 2005, 2006 International Business Machines and others.
2 // All Rights Reserved.
3 // This code is published under the Eclipse Public License.
4 //
5 // $Id$
6 //
7 // Authors: Andreas Waechter IBM 2005-10-18
8
9 #ifndef __MITTELMANNPARACNTRL_HPP__
10 #define __MITTELMANNPARACNTRL_HPP__
11
12 #include "RegisteredTNLP.hpp"
13
14 #ifdef HAVE_CONFIG_H
15 #include "config.h"
16 #else
17 #include "configall_system.h"
18 #endif
19
20 #ifdef HAVE_CMATH
21 # include <cmath>
22 #else
23 # ifdef HAVE_MATH_H
24 # include <math.h>
25 # else
26 # error "don't have header file for math"
27 # endif
28 #endif
29
30 using namespace Ipopt;
31
32 /** Base class for parabolic and elliptic control problems, as
33 * formulated by Hans Mittelmann as problem (P) in "Sufficient
34 * Optimality for Discretized Parabolic and Elliptic Control
35 * Problems"
36 */
37 template<class T>
38 class MittelmannParaCntrlBase : public RegisteredTNLP
39 {
40 public:
41 /** Constructor. */
42 MittelmannParaCntrlBase();
43
44 /** Default destructor */
45 virtual ~MittelmannParaCntrlBase();
46
47 /**@name Overloaded from TNLP */
48 //@{
49 /** Method to return some info about the nlp */
50 virtual bool get_nlp_info(Index& n, Index& m, Index& nnz_jac_g,
51 Index& nnz_h_lag, IndexStyleEnum& index_style);
52
53 /** Method to return the bounds for my problem */
54 virtual bool get_bounds_info(Index n, Number* x_l, Number* x_u,
55 Index m, Number* g_l, Number* g_u);
56
57 /** Method to return the starting point for the algorithm */
58 virtual bool get_starting_point(Index n, bool init_x, Number* x,
59 bool init_z, Number* z_L, Number* z_U,
60 Index m, bool init_lambda,
61 Number* lambda);
62
63 /** Method to return the objective value */
64 virtual bool eval_f(Index n, const Number* x, bool new_x, Number& obj_value);
65
66 /** Method to return the gradient of the objective */
67 virtual bool eval_grad_f(Index n, const Number* x, bool new_x, Number* grad_f);
68
69 /** Method to return the constraint residuals */
70 virtual bool eval_g(Index n, const Number* x, bool new_x, Index m, Number* g);
71
72 /** Method to return:
73 * 1) The structure of the jacobian (if "values" is NULL)
74 * 2) The values of the jacobian (if "values" is not NULL)
75 */
76 virtual bool eval_jac_g(Index n, const Number* x, bool new_x,
77 Index m, Index nele_jac, Index* iRow, Index *jCol,
78 Number* values);
79
80 /** Method to return:
81 * 1) The structure of the hessian of the lagrangian (if "values" is NULL)
82 * 2) The values of the hessian of the lagrangian (if "values" is not NULL)
83 */
84 virtual bool eval_h(Index n, const Number* x, bool new_x,
85 Number obj_factor, Index m, const Number* lambda,
86 bool new_lambda, Index nele_hess, Index* iRow,
87 Index* jCol, Number* values);
88
89 //@}
90
91 /** Method for returning scaling parameters */
92 virtual bool get_scaling_parameters(Number& obj_scaling,
93 bool& use_x_scaling, Index n,
94 Number* x_scaling,
95 bool& use_g_scaling, Index m,
96 Number* g_scaling);
97
98 /** @name Solution Methods */
99 //@{
100 /** This method is called after the optimization, and could write an
101 * output file with the optimal profiles */
102 virtual void finalize_solution(SolverReturn status,
103 Index n, const Number* x, const Number* z_L, const Number* z_U,
104 Index m, const Number* g, const Number* lambda,
105 Number obj_value,
106 const IpoptData* ip_data,
107 IpoptCalculatedQuantities* ip_cq);
108 //@}
109
110 virtual bool InitializeProblem(Index N);
111
112 private:
113 /**@name Methods to block default compiler methods.
114 * The compiler automatically generates the following three methods.
115 * Since the default compiler implementation is generally not what
116 * you want (for all but the most simple classes), we usually
117 * put the declarations of these methods in the private section
118 * and never implement them. This prevents the compiler from
119 * implementing an incorrect "default" behavior without us
120 * knowing. (See Scott Meyers book, "Effective C++")
121 *
122 */
123 //@{
124 MittelmannParaCntrlBase(const MittelmannParaCntrlBase<T>&);
125 MittelmannParaCntrlBase& operator=(const MittelmannParaCntrlBase<T>&);
126 //@}
127
128 /**@name Problem specification */
129 //@{
130 /** Upper bound on t */
131 Number T_;
132 /** Upper bound on x */
133 Number l_;
134 /** Number of mesh points in t direction */
135 Index Nt_;
136 /** Number of mesh points in x direction */
137 Index Nx_;
138 /** Step size in t direction*/
139 Number dt_;
140 /** Step size in x direction*/
141 Number dx_;
142 /** overall lower bound on y */
143 Number lb_y_;
144 /** overall upper bound on y */
145 Number ub_y_;
146 /** overall lower bound on u */
147 Number lb_u_;
148 /** overall upper bound on u */
149 Number ub_u_;
150 /** Weighting parameter for the control target deviation functional
151 * in the objective */
152 Number alpha_;
153 /** Weighting parameter in PDE */
154 Number beta_;
155 /** Array for the target profile for y in objective */
156 Number* y_T_;
157 /** Array for weighting function a_y in objective */
158 Number* a_y_;
159 /** Array for weighting function a_u in objective */
160 Number* a_u_;
161 //@}
162
163 /**@name Auxilliary methods */
164 //@{
165 /** Translation of mesh point indices to NLP variable indices for
166 * y(x_ij) */
y_index(Index jx,Index it) const167 inline Index y_index(Index jx, Index it) const
168 {
169 return jx + (Nx_+1)*it;
170 }
u_index(Index it) const171 inline Index u_index(Index it) const
172 {
173 return (Nt_+1)*(Nx_+1) + it - 1;
174 }
175 /** Compute the grid coordinate for given index in t direction */
t_grid(Index i) const176 inline Number t_grid(Index i) const
177 {
178 return dt_*(Number)i;
179 }
180 /** Compute the grid coordinate for given index in x direction */
x_grid(Index j) const181 inline Number x_grid(Index j) const
182 {
183 return dx_*(Number)j;
184 }
185 //@}
186 };
187
188 template <class T>
MittelmannParaCntrlBase()189 MittelmannParaCntrlBase<T>::MittelmannParaCntrlBase()
190 :
191 y_T_(NULL),
192 a_y_(NULL),
193 a_u_(NULL)
194 {}
195
196 template <class T>
~MittelmannParaCntrlBase()197 MittelmannParaCntrlBase<T>::~MittelmannParaCntrlBase()
198 {
199 delete [] y_T_;
200 delete [] a_y_;
201 delete [] a_u_;
202 }
203
204 template <class T>
205 bool MittelmannParaCntrlBase<T>::
InitializeProblem(Index N)206 InitializeProblem(Index N)
207 {
208 typename T::ProblemSpecs p;
209
210 if (N<1) {
211 printf("N has to be at least 1.");
212 return false;
213 }
214
215 T_ = p.T();
216 l_ = p.l();
217 Nt_ = N;
218 Nx_ = N;
219 dt_ = T_/Nt_;
220 dx_ = l_/Nx_;
221 lb_y_ = p.lb_y();
222 ub_y_ = p.ub_y();
223 lb_u_ = p.lb_u();
224 ub_u_ = p.ub_u();
225 alpha_ = p.alpha();
226 beta_ = p.beta();
227
228 y_T_ = new Number[Nx_+1];
229 for (Index j=0; j<=Nx_; j++) {
230 y_T_[j] = p.y_T(x_grid(j));
231 }
232 a_y_ = new Number[Nt_];
233 for (Index i=1; i<=Nt_; i++) {
234 a_y_[i-1] = p.a_y(t_grid(i));
235 }
236 a_u_ = new Number[Nt_];
237 for (Index i=1; i<=Nt_; i++) {
238 a_u_[i-1] = p.a_u(t_grid(i));
239 }
240
241 return true;
242 }
243
244 template <class T>
245 bool MittelmannParaCntrlBase<T>::
get_nlp_info(Index & n,Index & m,Index & nnz_jac_g,Index & nnz_h_lag,IndexStyleEnum & index_style)246 get_nlp_info(Index& n, Index& m, Index& nnz_jac_g,
247 Index& nnz_h_lag, IndexStyleEnum& index_style)
248 {
249 typename T::ProblemSpecs p;
250
251 n = (Nt_+1)*(Nx_+1) + Nt_;
252
253 m = Nt_*(Nx_-1) + Nt_ + Nt_;
254
255 nnz_jac_g = 6*Nt_*(Nx_-1) + 3*Nt_ + 4*Nt_;
256
257 nnz_h_lag = Nx_+1 + Nt_;
258 if (!p.phi_dydy_always_zero()) {
259 nnz_h_lag += Nt_;
260 }
261
262 index_style = C_STYLE;
263
264 return true;
265 }
266
267 template <class T>
268 bool MittelmannParaCntrlBase<T>::
get_bounds_info(Index n,Number * x_l,Number * x_u,Index m,Number * g_l,Number * g_u)269 get_bounds_info(Index n, Number* x_l, Number* x_u,
270 Index m, Number* g_l, Number* g_u)
271 {
272 typename T::ProblemSpecs p;
273
274 // Set overall bounds on the variables
275 for (Index jx=0; jx<=Nx_; jx++) {
276 for (Index it=1; it<=Nt_; it++) {
277 Index iy = y_index(jx,it);
278 x_l[iy] = lb_y_;
279 x_u[iy] = ub_y_;
280 }
281 }
282 for (Index i=1; i<=Nt_; i++) {
283 Index iu = u_index(i);
284 x_l[iu] = lb_u_;
285 x_u[iu] = ub_u_;
286 }
287
288 /*
289 // Boundary condition for t=0
290 for (Index it=0; it<=Nt_; it++) {
291 Index iy = y_index(0,it);
292 x_u[iy] = x_l[iy] = p.a(t_grid(it));
293 }
294 */
295 // Boundary condition for t=0
296 for (Index jx=0; jx<=Nx_; jx++) {
297 Index iy = y_index(jx,0);
298 x_u[iy] = x_l[iy] = p.a(x_grid(jx));
299 }
300
301 // all discretized PDE constraints have right hand side zero
302 for (Index i=0; i<Nt_*(Nx_-1) + Nt_; i++) {
303 g_l[i] = 0.;
304 g_u[i] = 0.;
305 }
306
307 // but we put b on the right hand side for the x=L boundary condition
308 for (Index i=0; i<Nt_; i++) {
309 g_l[Nt_*(Nx_-1) + Nt_ + i]
310 = g_u[Nt_*(Nx_-1) + Nt_ + i]
311 = p.b(t_grid(i+1));
312 }
313
314 return true;
315 }
316
317 template <class T>
318 bool MittelmannParaCntrlBase<T>::
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)319 get_starting_point(Index n, bool init_x, Number* x,
320 bool init_z, Number* z_L, Number* z_U,
321 Index m, bool init_lambda,
322 Number* lambda)
323 {
324 DBG_ASSERT(init_x==true && init_z==false && init_lambda==false);
325
326 // Set starting point for y
327 for (Index jx=0; jx<=Nx_; jx++) {
328 for (Index it=0; it<=Nt_; it++) {
329 x[y_index(jx,it)] = 0.;
330 }
331 }
332 // for u
333 for (Index i=1; i<=Nt_; i++) {
334 x[u_index(i)] = (ub_u_+lb_u_)/2.;
335 }
336
337 /*
338 // DELETEME
339 for (Index i=0; i<n; i++) {
340 x[i] += 0.01*i;
341 }
342 */
343
344 return true;
345 }
346
347 template <class T>
348 bool MittelmannParaCntrlBase<T>::
get_scaling_parameters(Number & obj_scaling,bool & use_x_scaling,Index n,Number * x_scaling,bool & use_g_scaling,Index m,Number * g_scaling)349 get_scaling_parameters(Number& obj_scaling,
350 bool& use_x_scaling,
351 Index n, Number* x_scaling,
352 bool& use_g_scaling,
353 Index m, Number* g_scaling)
354 {
355 obj_scaling = 1./Min(dx_,dt_);
356 use_x_scaling = false;
357 use_g_scaling = false;
358 return true;
359 }
360
361 template <class T>
362 bool MittelmannParaCntrlBase<T>::
eval_f(Index n,const Number * x,bool new_x,Number & obj_value)363 eval_f(Index n, const Number* x,
364 bool new_x, Number& obj_value)
365 {
366 // Deviation of y from target
367 Number a = x[y_index(0,Nt_)] - y_T_[0];
368 Number sum = 0.5*a*a;
369 for (Index jx=1; jx<Nx_; jx++) {
370 a = x[y_index(jx,Nt_)] - y_T_[jx];
371 sum += a*a;
372 }
373 a = x[y_index(Nx_,Nt_)] - y_T_[Nx_];
374 sum += 0.5*a*a;
375 obj_value = 0.5*dx_*sum;
376
377 // smoothing for control
378 if (alpha_!=.0) {
379 sum = 0.5*x[u_index(Nt_)]*x[u_index(Nt_)];
380 for (Index it=1; it < Nt_; it++) {
381 sum += x[u_index(it)]*x[u_index(it)];
382 }
383 obj_value += 0.5*alpha_*dt_*sum;
384 }
385
386 // third term
387 sum = 0.;
388 for (Index it=1; it<Nt_; it++) {
389 sum += a_y_[it-1]*x[y_index(Nx_,it)] + a_u_[it-1]*x[u_index(it)];
390 }
391 sum += 0.5*(a_y_[Nt_-1]*x[y_index(Nx_,Nt_)] + a_u_[Nt_-1]*x[u_index(Nt_)]);
392 obj_value += dt_*sum;
393
394 return true;
395 }
396
397 template <class T>
398 bool MittelmannParaCntrlBase<T>::
eval_grad_f(Index n,const Number * x,bool new_x,Number * grad_f)399 eval_grad_f(Index n, const Number* x, bool new_x, Number* grad_f)
400 {
401 // First set all y entries to zero
402 for (Index jx=0; jx<=Nx_; jx++) {
403 for (Index it=0; it<=Nt_; it++) {
404 grad_f[y_index(jx,it)] = 0.;
405 }
406 }
407
408 // y entries from y target
409 grad_f[y_index(0,Nt_)] = 0.5*dx_*(x[y_index(0,Nt_)] - y_T_[0]);
410 for (Index jx=1; jx<Nx_; jx++) {
411 grad_f[y_index(jx,Nt_)] = dx_*(x[y_index(jx,Nt_)] - y_T_[jx]);
412 }
413 grad_f[y_index(Nx_,Nt_)] = 0.5*dx_*(x[y_index(Nx_,Nt_)] - y_T_[Nx_]);
414
415 // y entries from thrid term
416 for (Index it=1; it<Nt_; it++) {
417 grad_f[y_index(Nx_,it)] = dt_*a_y_[it-1];
418 }
419 grad_f[y_index(Nx_,Nt_)] += 0.5*dt_*a_y_[Nt_-1];
420
421 // u entries from smoothing and third term
422 for (Index it=1; it<Nt_; it++) {
423 grad_f[u_index(it)] = alpha_*dt_*x[u_index(it)] + dt_*a_u_[it-1];
424 }
425 grad_f[u_index(Nt_)] = 0.5*dt_*(alpha_*x[u_index(Nt_)] + a_u_[Nt_-1]);
426
427 return true;
428 }
429
430 template <class T>
431 bool MittelmannParaCntrlBase<T>::
eval_g(Index n,const Number * x,bool new_x,Index m,Number * g)432 eval_g(Index n, const Number* x, bool new_x, Index m, Number* g)
433 {
434 typename T::ProblemSpecs p;
435
436 Index ig=0;
437
438 Number f = 1./(2.*dx_*dx_);
439 for (Index jx=1; jx<Nx_; jx++) {
440 for (Index it=0; it<Nt_; it++) {
441 g[ig] = (x[y_index(jx,it)]-x[y_index(jx,it+1)])/dt_
442 + f*(x[y_index(jx-1,it)] - 2.*x[y_index(jx,it)]
443 + x[y_index(jx+1,it)] + x[y_index(jx-1,it+1)]
444 - 2.*x[y_index(jx,it+1)] + x[y_index(jx+1,it+1)]);
445 ig++;
446 }
447 }
448
449 for (Index it=1; it<=Nt_; it++) {
450 g[ig] = (x[y_index(2,it)] - 4.*x[y_index(1,it)] + 3.*x[y_index(0,it)]);
451 ig++;
452 }
453
454 f = 1./(2.*dx_);
455 for (Index it=1; it<=Nt_; it++) {
456 g[ig] =
457 f*(x[y_index(Nx_-2,it)] - 4.*x[y_index(Nx_-1,it)]
458 + 3.*x[y_index(Nx_,it)]) + beta_*x[y_index(Nx_,it)]
459 - x[u_index(it)] + p.phi(x[y_index(Nx_,it)]);
460 ig++;
461 }
462
463 DBG_ASSERT(ig == m);
464
465 return true;
466 }
467
468 template <class T>
469 bool MittelmannParaCntrlBase<T>::
eval_jac_g(Index n,const Number * x,bool new_x,Index m,Index nele_jac,Index * iRow,Index * jCol,Number * values)470 eval_jac_g(Index n, const Number* x, bool new_x,
471 Index m, Index nele_jac, Index* iRow, Index *jCol,
472 Number* values)
473 {
474 typename T::ProblemSpecs p;
475
476 Index ijac = 0;
477
478 if (values == NULL) {
479 Index ig = 0;
480 for (Index jx=1; jx<Nx_; jx++) {
481 for (Index it=0; it<Nt_; it++) {
482 iRow[ijac] = ig;
483 jCol[ijac] = y_index(jx-1,it);
484 ijac++;
485 iRow[ijac] = ig;
486 jCol[ijac] = y_index(jx,it);
487 ijac++;
488 iRow[ijac] = ig;
489 jCol[ijac] = y_index(jx+1,it);
490 ijac++;
491 iRow[ijac] = ig;
492 jCol[ijac] = y_index(jx-1,it+1);
493 ijac++;
494 iRow[ijac] = ig;
495 jCol[ijac] = y_index(jx,it+1);
496 ijac++;
497 iRow[ijac] = ig;
498 jCol[ijac] = y_index(jx+1,it+1);
499 ijac++;
500
501 ig++;
502 }
503 }
504
505 for (Index it=1; it<=Nt_; it++) {
506 iRow[ijac] = ig;
507 jCol[ijac] = y_index(0,it);
508 ijac++;
509 iRow[ijac] = ig;
510 jCol[ijac] = y_index(1,it);
511 ijac++;
512 iRow[ijac] = ig;
513 jCol[ijac] = y_index(2,it);
514 ijac++;
515
516 ig++;
517 }
518
519 for (Index it=1; it<=Nt_; it++) {
520 iRow[ijac] = ig;
521 jCol[ijac] = y_index(Nx_-2,it);
522 ijac++;
523 iRow[ijac] = ig;
524 jCol[ijac] = y_index(Nx_-1,it);
525 ijac++;
526 iRow[ijac] = ig;
527 jCol[ijac] = y_index(Nx_,it);
528 ijac++;
529 iRow[ijac] = ig;
530 jCol[ijac] = u_index(it);
531 ijac++;
532
533 ig++;
534 }
535 DBG_ASSERT(ig == m);
536 }
537 else {
538 Number f = 1./(2.*dx_*dx_);
539 Number f2 = 1./dt_;
540 for (Index jx=1; jx<Nx_; jx++) {
541 for (Index it=0; it<Nt_; it++) {
542 values[ijac++] = f;
543 values[ijac++] = f2 - 2.*f;
544 values[ijac++] = f;
545 values[ijac++] = f;
546 values[ijac++] = -f2 - 2.*f;
547 values[ijac++] = f;
548 }
549 }
550
551 for (Index it=1; it<=Nt_; it++) {
552 values[ijac++] = 3.;
553 values[ijac++] = -4.;
554 values[ijac++] = 1.;
555 }
556
557 f = 1./(2.*dx_);
558 for (Index it=1; it<=Nt_; it++) {
559 values[ijac++] = f;
560 values[ijac++] = -4.*f;
561 values[ijac++] = 3.*f + beta_ + p.phi_dy(x[y_index(Nx_,it)]);
562 values[ijac++] = -1.;
563 }
564
565 }
566
567 return true;
568 }
569
570 template <class T>
571 bool MittelmannParaCntrlBase<T>::
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)572 eval_h(Index n, const Number* x, bool new_x,
573 Number obj_factor, Index m, const Number* lambda,
574 bool new_lambda, Index nele_hess, Index* iRow,
575 Index* jCol, Number* values)
576 {
577 typename T::ProblemSpecs p;
578
579 Index ihes = 0;
580
581 if (values == NULL) {
582 // y values from objective
583 for (Index jx=0; jx<= Nx_; jx++) {
584 iRow[ihes] = y_index(jx,Nt_);
585 jCol[ihes] = y_index(jx,Nt_);
586 ihes++;
587 }
588 // u from objective
589 for (Index it=1; it<=Nt_; it++) {
590 iRow[ihes] = u_index(it);
591 jCol[ihes] = u_index(it);
592 ihes++;
593 }
594
595 // constraint
596 if (!p.phi_dydy_always_zero()) {
597 for (Index it=1; it<=Nt_; it++) {
598 iRow[ihes] = y_index(Nx_,it);
599 jCol[ihes] = y_index(Nx_,it);
600 ihes++;
601 }
602 }
603 }
604 else {
605 // y values from objective
606 values[ihes++] = obj_factor*0.5*dx_;
607 for (Index jx=1; jx<Nx_; jx++) {
608 values[ihes++] = obj_factor*dx_;
609 }
610 values[ihes++] = obj_factor*0.5*dx_;
611 // u from objective
612 for (Index it=1; it<Nt_; it++) {
613 values[ihes++] = obj_factor*alpha_*dt_;
614 }
615 values[ihes++] = obj_factor*0.5*alpha_*dt_;
616
617 // constrainT
618 if (!p.phi_dydy_always_zero()) {
619 Index ig = (Nx_-1)*Nt_ + Nt_;
620 for (Index it=1; it<=Nt_; it++) {
621 values[ihes++] = lambda[ig++]*p.phi_dydy(x[y_index(Nx_,it)]);
622 }
623 }
624 }
625
626 DBG_ASSERT(ihes==nele_hess);
627
628 return true;
629 }
630
631 template <class T>
632 void MittelmannParaCntrlBase<T>::
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)633 finalize_solution(SolverReturn status,
634 Index n, const Number* x, const Number* z_L,
635 const Number* z_U,
636 Index m, const Number* g, const Number* lambda,
637 Number obj_value,
638 const IpoptData* ip_data,
639 IpoptCalculatedQuantities* ip_cq)
640 {}
641
642 class MittelmannParaCntrl5_1
643 {
644 public:
645 class ProblemSpecs
646 {
647 public:
ProblemSpecs()648 ProblemSpecs ()
649 :
650 pi_(4.*atan(1.)),
651 exp13_(exp(1./3.)),
652 exp23_(exp(2./3.)),
653 exp1_(exp(1.)),
654 expm1_(exp(-1.)),
655 sqrt2_(sqrt(2.))
656 {}
T()657 Number T()
658 {
659 return 1.;
660 }
l()661 Number l()
662 {
663 return pi_/4.;
664 }
lb_y()665 Number lb_y()
666 {
667 return -1e20;
668 }
ub_y()669 Number ub_y()
670 {
671 return 1e20;
672 }
lb_u()673 Number lb_u()
674 {
675 return 0.;
676 }
ub_u()677 Number ub_u()
678 {
679 return 1.;
680 }
alpha()681 Number alpha()
682 {
683 return sqrt2_/2.*(exp23_-exp13_);
684 }
beta()685 Number beta()
686 {
687 return 1.;
688 }
y_T(Number x)689 inline Number y_T(Number x)
690 {
691 return (exp1_ + expm1_)*cos(x);
692 }
a(Number x)693 inline Number a(Number x)
694 {
695 return cos(x);
696 }
a_y(Number t)697 inline Number a_y(Number t)
698 {
699 return -exp(-2.*t);
700 }
a_u(Number t)701 inline Number a_u(Number t)
702 {
703 return sqrt2_/2.*exp13_;
704 }
b(Number t)705 inline Number b(Number t)
706 {
707 return exp(-4.*t)/4.
708 - Min(1.,Max(0.,(exp(t)-exp13_)/(exp23_-exp13_)));
709 }
phi(Number y)710 inline Number phi(Number y)
711 {
712 return y*pow(fabs(y),3);
713 }
phi_dy(Number y)714 inline Number phi_dy(Number y)
715 {
716 return 4.*pow(fabs(y),3);
717 }
phi_dydy(Number y)718 inline Number phi_dydy(Number y)
719 {
720 return 12.*y*y;
721 }
phi_dydy_always_zero()722 inline bool phi_dydy_always_zero()
723 {
724 return false;
725 }
726 private:
727 const Number pi_;
728 const Number exp13_;
729 const Number exp23_;
730 const Number exp1_;
731 const Number expm1_;
732 const Number sqrt2_;
733 };
734 };
735
736 class MittelmannParaCntrl5_2_1
737 {
738 public:
739 class ProblemSpecs
740 {
741 public:
ProblemSpecs()742 ProblemSpecs ()
743 {}
T()744 Number T()
745 {
746 return 1.58;
747 }
l()748 Number l()
749 {
750 return 1.;
751 }
lb_y()752 Number lb_y()
753 {
754 return -1e20;
755 }
ub_y()756 Number ub_y()
757 {
758 return 1e20;
759 }
lb_u()760 Number lb_u()
761 {
762 return -1.;
763 }
ub_u()764 Number ub_u()
765 {
766 return 1.;
767 }
alpha()768 Number alpha()
769 {
770 return 0.001;
771 }
beta()772 Number beta()
773 {
774 return 1.;
775 }
y_T(Number x)776 inline Number y_T(Number x)
777 {
778 return .5*(1.-x*x);
779 }
a(Number x)780 inline Number a(Number x)
781 {
782 return 0.;
783 }
a_y(Number t)784 inline Number a_y(Number t)
785 {
786 return 0.;
787 }
a_u(Number t)788 inline Number a_u(Number t)
789 {
790 return 0.;
791 }
b(Number t)792 inline Number b(Number t)
793 {
794 return 0.;
795 }
phi(Number y)796 inline Number phi(Number y)
797 {
798 return 0.;
799 }
phi_dy(Number y)800 inline Number phi_dy(Number y)
801 {
802 return 0.;
803 }
phi_dydy(Number y)804 inline Number phi_dydy(Number y)
805 {
806 DBG_ASSERT(false);
807 return 0.;
808 }
phi_dydy_always_zero()809 inline bool phi_dydy_always_zero()
810 {
811 return true;
812 }
813 };
814 };
815
816 class MittelmannParaCntrl5_2_2
817 {
818 public:
819 class ProblemSpecs
820 {
821 public:
ProblemSpecs()822 ProblemSpecs ()
823 {}
T()824 Number T()
825 {
826 return 1.58;
827 }
l()828 Number l()
829 {
830 return 1.;
831 }
lb_y()832 Number lb_y()
833 {
834 return -1e20;
835 }
ub_y()836 Number ub_y()
837 {
838 return 1e20;
839 }
lb_u()840 Number lb_u()
841 {
842 return -1.;
843 }
ub_u()844 Number ub_u()
845 {
846 return 1.;
847 }
alpha()848 Number alpha()
849 {
850 return 0.001;
851 }
beta()852 Number beta()
853 {
854 return 0.;
855 }
y_T(Number x)856 inline Number y_T(Number x)
857 {
858 return .5*(1.-x*x);
859 }
a(Number x)860 inline Number a(Number x)
861 {
862 return 0.;
863 }
a_y(Number t)864 inline Number a_y(Number t)
865 {
866 return 0.;
867 }
a_u(Number t)868 inline Number a_u(Number t)
869 {
870 return 0.;
871 }
b(Number t)872 inline Number b(Number t)
873 {
874 return 0.;
875 }
phi(Number y)876 inline Number phi(Number y)
877 {
878 return y*y;
879 }
phi_dy(Number y)880 inline Number phi_dy(Number y)
881 {
882 return 2.*y;
883 }
phi_dydy(Number y)884 inline Number phi_dydy(Number y)
885 {
886 return 2.;
887 }
phi_dydy_always_zero()888 inline bool phi_dydy_always_zero()
889 {
890 return false;
891 }
892 };
893 };
894
895 class MittelmannParaCntrl5_2_3
896 {
897 public:
898 class ProblemSpecs
899 {
900 public:
ProblemSpecs()901 ProblemSpecs ()
902 {}
T()903 Number T()
904 {
905 return 1.58;
906 }
l()907 Number l()
908 {
909 return 1.;
910 }
lb_y()911 Number lb_y()
912 {
913 return 0.;
914 }
ub_y()915 Number ub_y()
916 {
917 return 0.675;
918 }
lb_u()919 Number lb_u()
920 {
921 return -1.;
922 }
ub_u()923 Number ub_u()
924 {
925 return 1.;
926 }
alpha()927 Number alpha()
928 {
929 return 0.001;
930 }
beta()931 Number beta()
932 {
933 return 0.;
934 }
y_T(Number x)935 inline Number y_T(Number x)
936 {
937 return .5*(1.-x*x);
938 }
a(Number x)939 inline Number a(Number x)
940 {
941 return 0.;
942 }
a_y(Number t)943 inline Number a_y(Number t)
944 {
945 return 0.;
946 }
a_u(Number t)947 inline Number a_u(Number t)
948 {
949 return 0.;
950 }
b(Number t)951 inline Number b(Number t)
952 {
953 return 0.;
954 }
phi(Number y)955 inline Number phi(Number y)
956 {
957 return y*y;
958 }
phi_dy(Number y)959 inline Number phi_dy(Number y)
960 {
961 return 2.*y;
962 }
phi_dydy(Number y)963 inline Number phi_dydy(Number y)
964 {
965 return 2.;
966 }
phi_dydy_always_zero()967 inline bool phi_dydy_always_zero()
968 {
969 return false;
970 }
971 };
972 };
973
974 class MittelmannParaCntrl5_try
975 {
976 public:
977 class ProblemSpecs
978 {
979 public:
ProblemSpecs()980 ProblemSpecs ()
981 :
982 pi_(4.*atan(1.)),
983 exp13_(exp(1./3.)),
984 exp23_(exp(2./3.)),
985 exp1_(exp(1.)),
986 expm1_(exp(-1.)),
987 sqrt2_(sqrt(2.))
988 {}
T()989 Number T()
990 {
991 return 1.;
992 }
l()993 Number l()
994 {
995 return pi_/4.;
996 }
lb_y()997 Number lb_y()
998 {
999 return -1e20;
1000 }
ub_y()1001 Number ub_y()
1002 {
1003 return 1e20;
1004 }
lb_u()1005 Number lb_u()
1006 {
1007 return 0.;
1008 }
ub_u()1009 Number ub_u()
1010 {
1011 return 1.;
1012 }
alpha()1013 Number alpha()
1014 {
1015 return sqrt2_/2.*(exp23_-exp13_);
1016 }
beta()1017 Number beta()
1018 {
1019 return 1.;
1020 }
y_T(Number x)1021 inline Number y_T(Number x)
1022 {
1023 return (exp1_ + expm1_)*cos(x);
1024 }
a(Number x)1025 inline Number a(Number x)
1026 {
1027 return cos(x);
1028 }
a_y(Number t)1029 inline Number a_y(Number t)
1030 {
1031 return -exp(-2.*t);
1032 }
a_u(Number t)1033 inline Number a_u(Number t)
1034 {
1035 return sqrt2_/2.*exp13_;
1036 }
b(Number t)1037 inline Number b(Number t)
1038 {
1039 return exp(-4.*t)/4.
1040 - Min(1.,Max(0.,(exp(t)-exp13_)/(exp23_-exp13_)));
1041 }
phi(Number y)1042 inline Number phi(Number y)
1043 {
1044 return -y*sin(y/10.);
1045 }
phi_dy(Number y)1046 inline Number phi_dy(Number y)
1047 {
1048 return -y*cos(y/10.)/10. - sin(y/10.);
1049 }
phi_dydy(Number y)1050 inline Number phi_dydy(Number y)
1051 {
1052 return y*sin(y/10.)/100.;
1053 }
phi_dydy_always_zero()1054 inline bool phi_dydy_always_zero()
1055 {
1056 return false;
1057 }
1058 private:
1059 const Number pi_;
1060 const Number exp13_;
1061 const Number exp23_;
1062 const Number exp1_;
1063 const Number expm1_;
1064 const Number sqrt2_;
1065 };
1066 };
1067
1068 #endif
1069