| Line |
Branch |
Exec |
Source |
| 1 |
|
|
/////////////////////////////////////////////////////////////////////////////// |
| 2 |
|
|
// BSD 3-Clause License |
| 3 |
|
|
// |
| 4 |
|
|
// Copyright (C) 2019-2025, LAAS-CNRS, University of Edinburgh, |
| 5 |
|
|
// Heriot-Watt University |
| 6 |
|
|
// Copyright note valid unless otherwise stated in individual files. |
| 7 |
|
|
// All rights reserved. |
| 8 |
|
|
/////////////////////////////////////////////////////////////////////////////// |
| 9 |
|
|
|
| 10 |
|
|
#ifndef CROCODDYL_CORE_SOLVER_BASE_HPP_ |
| 11 |
|
|
#define CROCODDYL_CORE_SOLVER_BASE_HPP_ |
| 12 |
|
|
|
| 13 |
|
|
#include <vector> |
| 14 |
|
|
|
| 15 |
|
|
#include "crocoddyl/core/fwd.hpp" |
| 16 |
|
|
#include "crocoddyl/core/optctrl/shooting.hpp" |
| 17 |
|
|
#include "crocoddyl/core/utils/stop-watch.hpp" |
| 18 |
|
|
|
| 19 |
|
|
namespace crocoddyl { |
| 20 |
|
|
|
| 21 |
|
|
class CallbackAbstract; // forward declaration |
| 22 |
|
|
static std::vector<Eigen::VectorXd> DEFAULT_VECTOR; |
| 23 |
|
|
|
| 24 |
|
|
enum FeasibilityNorm { LInf = 0, L1 }; |
| 25 |
|
|
|
| 26 |
|
|
/** |
| 27 |
|
|
* @brief Abstract class for optimal control solvers |
| 28 |
|
|
* |
| 29 |
|
|
* A solver resolves an optimal control solver of the form |
| 30 |
|
|
* \f{eqnarray*}{ |
| 31 |
|
|
* \begin{Bmatrix} |
| 32 |
|
|
* \mathbf{x}^*_0,\cdots,\mathbf{x}^*_{T} \\ |
| 33 |
|
|
* \mathbf{u}^*_0,\cdots,\mathbf{u}^*_{T-1} |
| 34 |
|
|
* \end{Bmatrix} = |
| 35 |
|
|
* \arg\min_{\mathbf{x}_s,\mathbf{u}_s} && l_T (\mathbf{x}_T) + \sum_{k=0}^{T-1} |
| 36 |
|
|
* l_k(\mathbf{x}_t,\mathbf{u}_t) \\ |
| 37 |
|
|
* \operatorname{subject}\,\operatorname{to} && \mathbf{x}_0 = |
| 38 |
|
|
* \mathbf{\tilde{x}}_0\\ |
| 39 |
|
|
* && \mathbf{x}_{k+1} = \mathbf{f}_k(\mathbf{x}_k,\mathbf{u}_k)\\ |
| 40 |
|
|
* && \mathbf{x}_k\in\mathcal{X}, \mathbf{u}_k\in\mathcal{U} |
| 41 |
|
|
* \f} |
| 42 |
|
|
* where \f$l_T(\mathbf{x}_T)\f$, \f$l_k(\mathbf{x}_t,\mathbf{u}_t)\f$ are the |
| 43 |
|
|
* terminal and running cost functions, respectively, |
| 44 |
|
|
* \f$\mathbf{f}_k(\mathbf{x}_k,\mathbf{u}_k)\f$ describes evolution of the |
| 45 |
|
|
* system, and state and control admissible sets are defined by |
| 46 |
|
|
* \f$\mathbf{x}_k\in\mathcal{X}\f$, \f$\mathbf{u}_k\in\mathcal{U}\f$. An action |
| 47 |
|
|
* model, defined in the shooting problem, describes each node \f$k\f$. Inside |
| 48 |
|
|
* the action model, we specialize the cost functions, the system evolution and |
| 49 |
|
|
* the admissible sets. |
| 50 |
|
|
* |
| 51 |
|
|
* The main routines are `computeDirection()` and `tryStep()`. The former finds |
| 52 |
|
|
* a search direction and typically computes the derivatives of each action |
| 53 |
|
|
* model. The latter rollout the dynamics and cost (i.e., the action) to try the |
| 54 |
|
|
* search direction found by `computeDirection`. Both functions used the current |
| 55 |
|
|
* guess defined by `setCandidate()`. Finally, `solve()` function is used to |
| 56 |
|
|
* define when the search direction and length are computed in each iterate. It |
| 57 |
|
|
* also describes the globalization strategy (i.e., regularization) of the |
| 58 |
|
|
* numerical optimization. |
| 59 |
|
|
* |
| 60 |
|
|
* \sa `solve()`, `computeDirection()`, `tryStep()`, `stoppingCriteria()` |
| 61 |
|
|
*/ |
| 62 |
|
|
class SolverAbstract { |
| 63 |
|
|
public: |
| 64 |
|
|
EIGEN_MAKE_ALIGNED_OPERATOR_NEW |
| 65 |
|
|
|
| 66 |
|
|
/** |
| 67 |
|
|
* @brief Initialize the solver |
| 68 |
|
|
* |
| 69 |
|
|
* @param[in] problem shooting problem |
| 70 |
|
|
*/ |
| 71 |
|
|
explicit SolverAbstract(std::shared_ptr<ShootingProblem> problem); |
| 72 |
|
|
virtual ~SolverAbstract(); |
| 73 |
|
|
|
| 74 |
|
|
/** |
| 75 |
|
|
* @brief Compute the optimal trajectory \f$\mathbf{x}^*_s,\mathbf{u}^*_s\f$ |
| 76 |
|
|
* as lists of \f$T+1\f$ and \f$T\f$ terms |
| 77 |
|
|
* |
| 78 |
|
|
* From an initial guess \p init_xs, \p init_us (feasible or not), iterate |
| 79 |
|
|
* over `computeDirection()` and `tryStep()` until `stoppingCriteria()` is |
| 80 |
|
|
* below threshold. It also describes the globalization strategy used during |
| 81 |
|
|
* the numerical optimization. |
| 82 |
|
|
* |
| 83 |
|
|
* @param[in] init_xs initial guess for state trajectory with \f$T+1\f$ |
| 84 |
|
|
* elements (default []) |
| 85 |
|
|
* @param[in] init_us initial guess for control trajectory with \f$T\f$ |
| 86 |
|
|
* elements (default []) |
| 87 |
|
|
* @param[in] maxiter maximum allowed number of iterations (default 100) |
| 88 |
|
|
* @param[in] is_feasible true if the \p init_xs are obtained from |
| 89 |
|
|
* integrating the \p init_us (rollout) (default false) |
| 90 |
|
|
* @param[in] init_reg initial guess for the regularization value. Very |
| 91 |
|
|
* low values are typical used with very good guess points (default 1e-9). |
| 92 |
|
|
* @return A boolean that describes if convergence was reached. |
| 93 |
|
|
*/ |
| 94 |
|
|
virtual bool solve( |
| 95 |
|
|
const std::vector<Eigen::VectorXd>& init_xs = DEFAULT_VECTOR, |
| 96 |
|
|
const std::vector<Eigen::VectorXd>& init_us = DEFAULT_VECTOR, |
| 97 |
|
|
const std::size_t maxiter = 100, const bool is_feasible = false, |
| 98 |
|
|
const double reg_init = NAN) = 0; |
| 99 |
|
|
|
| 100 |
|
|
/** |
| 101 |
|
|
* @brief Compute the search direction |
| 102 |
|
|
* \f$(\delta\mathbf{x}^k,\delta\mathbf{u}^k)\f$ for the current guess |
| 103 |
|
|
* \f$(\mathbf{x}^k_s,\mathbf{u}^k_s)\f$. |
| 104 |
|
|
* |
| 105 |
|
|
* You must call `setCandidate()` first in order to define the current guess. |
| 106 |
|
|
* A current guess defines a state and control trajectory |
| 107 |
|
|
* \f$(\mathbf{x}^k_s,\mathbf{u}^k_s)\f$ of \f$T+1\f$ and \f$T\f$ elements, |
| 108 |
|
|
* respectively. |
| 109 |
|
|
* |
| 110 |
|
|
* @param[in] recalc true for recalculating the derivatives at current state |
| 111 |
|
|
* and control |
| 112 |
|
|
* @return The search direction \f$(\delta\mathbf{x},\delta\mathbf{u})\f$ and |
| 113 |
|
|
* the dual lambdas as lists of \f$T+1\f$, \f$T\f$ and \f$T+1\f$ lengths, |
| 114 |
|
|
* respectively |
| 115 |
|
|
*/ |
| 116 |
|
|
virtual void computeDirection(const bool recalc) = 0; |
| 117 |
|
|
|
| 118 |
|
|
/** |
| 119 |
|
|
* @brief Try a predefined step length \f$\alpha\f$ and compute its cost |
| 120 |
|
|
* improvement \f$dV\f$. |
| 121 |
|
|
* |
| 122 |
|
|
* It uses the search direction found by `computeDirection()` to try a |
| 123 |
|
|
* determined step length \f$\alpha\f$. Therefore, it assumes that we have run |
| 124 |
|
|
* `computeDirection()` first. Additionally, it returns the cost improvement |
| 125 |
|
|
* \f$dV\f$ along the predefined step length \f$\alpha\f$. |
| 126 |
|
|
* |
| 127 |
|
|
* @param[in] steplength applied step length (\f$0\leq\alpha\leq1\f$) |
| 128 |
|
|
* @return the cost improvement |
| 129 |
|
|
*/ |
| 130 |
|
|
virtual double tryStep(const double steplength = 1) = 0; |
| 131 |
|
|
|
| 132 |
|
|
/** |
| 133 |
|
|
* @brief Return a positive value that quantifies the algorithm termination |
| 134 |
|
|
* |
| 135 |
|
|
* These values typically represents the gradient norm which tell us that it's |
| 136 |
|
|
* been reached the local minima. The stopping criteria strictly speaking |
| 137 |
|
|
* depends on the search direction (calculated by `computeDirection()`) but |
| 138 |
|
|
* it could also depend on the chosen step length, tested by `tryStep()`. |
| 139 |
|
|
*/ |
| 140 |
|
|
virtual double stoppingCriteria() = 0; |
| 141 |
|
|
|
| 142 |
|
|
/** |
| 143 |
|
|
* @brief Return the expected improvement \f$dV_{exp}\f$ from a given current |
| 144 |
|
|
* search direction \f$(\delta\mathbf{x}^k,\delta\mathbf{u}^k)\f$ |
| 145 |
|
|
* |
| 146 |
|
|
* For computing the expected improvement, you need to compute the search |
| 147 |
|
|
* direction first via `computeDirection()`. |
| 148 |
|
|
*/ |
| 149 |
|
|
virtual const Eigen::Vector2d& expectedImprovement() = 0; |
| 150 |
|
|
|
| 151 |
|
|
/** |
| 152 |
|
|
* @brief Resizing the solver data |
| 153 |
|
|
* |
| 154 |
|
|
* If the shooting problem has changed after construction, then this function |
| 155 |
|
|
* resizes all the data before starting resolve the problem. |
| 156 |
|
|
*/ |
| 157 |
|
|
virtual void resizeData(); |
| 158 |
|
|
|
| 159 |
|
|
/** |
| 160 |
|
|
* @brief Compute the dynamic feasibility |
| 161 |
|
|
* \f$\|\mathbf{f}_{\mathbf{s}}\|_{\infty,1}\f$ for the current guess |
| 162 |
|
|
* \f$(\mathbf{x}^k,\mathbf{u}^k)\f$ |
| 163 |
|
|
* |
| 164 |
|
|
* The feasibility can be computed using different norms (e.g, |
| 165 |
|
|
* \f$\ell_\infty\f$ or \f$\ell_1\f$ norms). By default we use the |
| 166 |
|
|
* \f$\ell_\infty\f$ norm, however, we can change the type of norm using |
| 167 |
|
|
* `set_feasnorm`. Note that \f$\mathbf{f}_{\mathbf{s}}\f$ are the gaps on the |
| 168 |
|
|
* dynamics, which are computed at each node as |
| 169 |
|
|
* \f$\mathbf{x}^{'}-\mathbf{f}(\mathbf{x},\mathbf{u})\f$. |
| 170 |
|
|
*/ |
| 171 |
|
|
double computeDynamicFeasibility(); |
| 172 |
|
|
|
| 173 |
|
|
/** |
| 174 |
|
|
* @brief Compute the feasibility of the inequality constraints for the |
| 175 |
|
|
* current guess |
| 176 |
|
|
* |
| 177 |
|
|
* The feasibility can be computed using different norms (e.g, |
| 178 |
|
|
* \f$\ell_\infty\f$ or \f$\ell_1\f$ norms). By default we use the |
| 179 |
|
|
* \f$\ell_\infty\f$ norm, however, we can change the type of norm using |
| 180 |
|
|
* `set_feasnorm`. |
| 181 |
|
|
*/ |
| 182 |
|
|
double computeInequalityFeasibility(); |
| 183 |
|
|
|
| 184 |
|
|
/** |
| 185 |
|
|
* @brief Compute the feasibility of the equality constraints for the current |
| 186 |
|
|
* guess |
| 187 |
|
|
* |
| 188 |
|
|
* The feasibility can be computed using different norms (e.g, |
| 189 |
|
|
* \f$\ell_\infty\f$ or \f$\ell_1\f$ norms). By default we use the |
| 190 |
|
|
* \f$\ell_\infty\f$ norm, however, we can change the type of norm using |
| 191 |
|
|
* `set_feasnorm`. |
| 192 |
|
|
*/ |
| 193 |
|
|
double computeEqualityFeasibility(); |
| 194 |
|
|
|
| 195 |
|
|
/** |
| 196 |
|
|
* @brief Set the solver candidate trajectories |
| 197 |
|
|
* \f$(\mathbf{x}_s,\mathbf{u}_s)\f$ |
| 198 |
|
|
* |
| 199 |
|
|
* The solver candidates are defined as a state and control trajectories |
| 200 |
|
|
* \f$(\mathbf{x}_s,\mathbf{u}_s)\f$ of \f$T+1\f$ and \f$T\f$ elements, |
| 201 |
|
|
* respectively. Additionally, we need to define the dynamic feasibility of |
| 202 |
|
|
* the \f$(\mathbf{x}_s,\mathbf{u}_s)\f$ pair. Note that the trajectories are |
| 203 |
|
|
* feasible if \f$\mathbf{x}_s\f$ is the resulting trajectory from the system |
| 204 |
|
|
* rollout with \f$\mathbf{u}_s\f$ inputs. |
| 205 |
|
|
* |
| 206 |
|
|
* @param[in] xs state trajectory of \f$T+1\f$ elements (default []) |
| 207 |
|
|
* @param[in] us control trajectory of \f$T\f$ elements (default []) |
| 208 |
|
|
* @param[in] isFeasible true if the \p xs are obtained from integrating the |
| 209 |
|
|
* \p us (rollout) |
| 210 |
|
|
*/ |
| 211 |
|
|
void setCandidate( |
| 212 |
|
|
const std::vector<Eigen::VectorXd>& xs_warm = DEFAULT_VECTOR, |
| 213 |
|
|
const std::vector<Eigen::VectorXd>& us_warm = DEFAULT_VECTOR, |
| 214 |
|
|
const bool is_feasible = false); |
| 215 |
|
|
|
| 216 |
|
|
/** |
| 217 |
|
|
* @brief Set a list of callback functions using for the solver diagnostic |
| 218 |
|
|
* |
| 219 |
|
|
* Each iteration, the solver calls these set of functions in order to allowed |
| 220 |
|
|
* user the diagnostic of its performance. |
| 221 |
|
|
* |
| 222 |
|
|
* @param callbacks set of callback functions |
| 223 |
|
|
*/ |
| 224 |
|
|
void setCallbacks( |
| 225 |
|
|
const std::vector<std::shared_ptr<CallbackAbstract> >& callbacks); |
| 226 |
|
|
|
| 227 |
|
|
/** |
| 228 |
|
|
* @brief Return the list of callback functions using for diagnostic |
| 229 |
|
|
*/ |
| 230 |
|
|
const std::vector<std::shared_ptr<CallbackAbstract> >& getCallbacks() const; |
| 231 |
|
|
|
| 232 |
|
|
/** |
| 233 |
|
|
* @brief Return the shooting problem |
| 234 |
|
|
*/ |
| 235 |
|
|
const std::shared_ptr<ShootingProblem>& get_problem() const; |
| 236 |
|
|
|
| 237 |
|
|
/** |
| 238 |
|
|
* @brief Return the state trajectory \f$\mathbf{x}_s\f$ |
| 239 |
|
|
*/ |
| 240 |
|
|
const std::vector<Eigen::VectorXd>& get_xs() const; |
| 241 |
|
|
|
| 242 |
|
|
/** |
| 243 |
|
|
* @brief Return the control trajectory \f$\mathbf{u}_s\f$ |
| 244 |
|
|
*/ |
| 245 |
|
|
const std::vector<Eigen::VectorXd>& get_us() const; |
| 246 |
|
|
|
| 247 |
|
|
/** |
| 248 |
|
|
* @brief Return the dynamic infeasibility \f$\mathbf{f}_{s}\f$ |
| 249 |
|
|
*/ |
| 250 |
|
|
const std::vector<Eigen::VectorXd>& get_fs() const; |
| 251 |
|
|
|
| 252 |
|
|
/** |
| 253 |
|
|
* @brief Return the feasibility status of the |
| 254 |
|
|
* \f$(\mathbf{x}_s,\mathbf{u}_s)\f$ trajectory |
| 255 |
|
|
*/ |
| 256 |
|
|
bool get_is_feasible() const; |
| 257 |
|
|
|
| 258 |
|
|
/** |
| 259 |
|
|
* @brief Return the cost for the current guess |
| 260 |
|
|
*/ |
| 261 |
|
|
double get_cost() const; |
| 262 |
|
|
|
| 263 |
|
|
/** |
| 264 |
|
|
* @brief Return the merit for the current guess |
| 265 |
|
|
*/ |
| 266 |
|
|
double get_merit() const; |
| 267 |
|
|
|
| 268 |
|
|
/** |
| 269 |
|
|
* @brief Return the stopping-criteria value computed by `stoppingCriteria()` |
| 270 |
|
|
*/ |
| 271 |
|
|
double get_stop() const; |
| 272 |
|
|
|
| 273 |
|
|
/** |
| 274 |
|
|
* @brief Return the linear and quadratic terms of the expected improvement |
| 275 |
|
|
*/ |
| 276 |
|
|
const Eigen::Vector2d& get_d() const; |
| 277 |
|
|
|
| 278 |
|
|
/** |
| 279 |
|
|
* @brief Return the reduction in the cost function \f$\Delta V\f$ |
| 280 |
|
|
*/ |
| 281 |
|
|
double get_dV() const; |
| 282 |
|
|
|
| 283 |
|
|
/** |
| 284 |
|
|
* @brief Return the reduction in the merit function \f$\Delta\Phi\f$ |
| 285 |
|
|
*/ |
| 286 |
|
|
double get_dPhi() const; |
| 287 |
|
|
|
| 288 |
|
|
/** |
| 289 |
|
|
* @brief Return the expected reduction in the cost function \f$\Delta |
| 290 |
|
|
* V_{exp}\f$ |
| 291 |
|
|
*/ |
| 292 |
|
|
double get_dVexp() const; |
| 293 |
|
|
|
| 294 |
|
|
/** |
| 295 |
|
|
* @brief Return the expected reduction in the merit function |
| 296 |
|
|
* \f$\Delta\Phi_{exp}\f$ |
| 297 |
|
|
*/ |
| 298 |
|
|
double get_dPhiexp() const; |
| 299 |
|
|
|
| 300 |
|
|
/** |
| 301 |
|
|
* @brief Return the reduction in the feasibility |
| 302 |
|
|
*/ |
| 303 |
|
|
double get_dfeas() const; |
| 304 |
|
|
|
| 305 |
|
|
/** |
| 306 |
|
|
* @brief Return the total feasibility for the current guess |
| 307 |
|
|
*/ |
| 308 |
|
|
double get_feas() const; |
| 309 |
|
|
|
| 310 |
|
|
/** |
| 311 |
|
|
* @brief Return the dynamic feasibility for the current guess |
| 312 |
|
|
*/ |
| 313 |
|
|
double get_ffeas() const; |
| 314 |
|
|
|
| 315 |
|
|
/** |
| 316 |
|
|
* @brief Return the inequality feasibility for the current guess |
| 317 |
|
|
*/ |
| 318 |
|
|
double get_gfeas() const; |
| 319 |
|
|
|
| 320 |
|
|
/** |
| 321 |
|
|
* @brief Return the equality feasibility for the current guess |
| 322 |
|
|
*/ |
| 323 |
|
|
double get_hfeas() const; |
| 324 |
|
|
|
| 325 |
|
|
/** |
| 326 |
|
|
* @brief Return the dynamic feasibility for the current step length |
| 327 |
|
|
*/ |
| 328 |
|
|
double get_ffeas_try() const; |
| 329 |
|
|
|
| 330 |
|
|
/** |
| 331 |
|
|
* @brief Return the inequality feasibility for the current step length |
| 332 |
|
|
*/ |
| 333 |
|
|
double get_gfeas_try() const; |
| 334 |
|
|
|
| 335 |
|
|
/** |
| 336 |
|
|
* @brief Return the equality feasibility for the current step length |
| 337 |
|
|
*/ |
| 338 |
|
|
double get_hfeas_try() const; |
| 339 |
|
|
|
| 340 |
|
|
/** |
| 341 |
|
|
* @brief Return the primal-variable regularization |
| 342 |
|
|
*/ |
| 343 |
|
|
double get_preg() const; |
| 344 |
|
|
|
| 345 |
|
|
/** |
| 346 |
|
|
* @brief Return the dual-variable regularization |
| 347 |
|
|
*/ |
| 348 |
|
|
double get_dreg() const; |
| 349 |
|
|
|
| 350 |
|
|
DEPRECATED("Use get_preg for primal-variable regularization", |
| 351 |
|
|
double get_xreg() const;) |
| 352 |
|
|
DEPRECATED("Use get_preg for primal-variable regularization", |
| 353 |
|
|
double get_ureg() const;) |
| 354 |
|
|
|
| 355 |
|
|
/** |
| 356 |
|
|
* @brief Return the step length \f$\alpha\f$ |
| 357 |
|
|
*/ |
| 358 |
|
|
double get_steplength() const; |
| 359 |
|
|
|
| 360 |
|
|
/** |
| 361 |
|
|
* @brief Return the threshold used for accepting a step |
| 362 |
|
|
*/ |
| 363 |
|
|
double get_th_acceptstep() const; |
| 364 |
|
|
|
| 365 |
|
|
/** |
| 366 |
|
|
* @brief Return the tolerance for stopping the algorithm |
| 367 |
|
|
*/ |
| 368 |
|
|
double get_th_stop() const; |
| 369 |
|
|
|
| 370 |
|
|
/** |
| 371 |
|
|
* @brief Return the threshold for accepting a gap as non-zero |
| 372 |
|
|
*/ |
| 373 |
|
|
double get_th_gaptol() const; |
| 374 |
|
|
|
| 375 |
|
|
/** |
| 376 |
|
|
* @brief Return the type of norm used to evaluate the dynamic and constraints |
| 377 |
|
|
* feasibility |
| 378 |
|
|
*/ |
| 379 |
|
|
FeasibilityNorm get_feasnorm() const; |
| 380 |
|
|
|
| 381 |
|
|
/** |
| 382 |
|
|
* @brief Return the number of iterations performed by the solver |
| 383 |
|
|
*/ |
| 384 |
|
|
std::size_t get_iter() const; |
| 385 |
|
|
|
| 386 |
|
|
/** |
| 387 |
|
|
* @brief Modify the state trajectory \f$\mathbf{x}_s\f$ |
| 388 |
|
|
*/ |
| 389 |
|
|
void set_xs(const std::vector<Eigen::VectorXd>& xs); |
| 390 |
|
|
|
| 391 |
|
|
/** |
| 392 |
|
|
* @brief Modify the control trajectory \f$\mathbf{u}_s\f$ |
| 393 |
|
|
*/ |
| 394 |
|
|
void set_us(const std::vector<Eigen::VectorXd>& us); |
| 395 |
|
|
|
| 396 |
|
|
/** |
| 397 |
|
|
* @brief Modify the primal-variable regularization value |
| 398 |
|
|
*/ |
| 399 |
|
|
void set_preg(const double preg); |
| 400 |
|
|
|
| 401 |
|
|
/** |
| 402 |
|
|
* @brief Modify the dual-variable regularization value |
| 403 |
|
|
*/ |
| 404 |
|
|
void set_dreg(const double dreg); |
| 405 |
|
|
|
| 406 |
|
|
DEPRECATED("Use set_preg for primal-variable regularization", |
| 407 |
|
|
void set_xreg(const double xreg);) |
| 408 |
|
|
DEPRECATED("Use set_preg for primal-variable regularization", |
| 409 |
|
|
void set_ureg(const double ureg);) |
| 410 |
|
|
|
| 411 |
|
|
/** |
| 412 |
|
|
* @brief Modify the threshold used for accepting step |
| 413 |
|
|
*/ |
| 414 |
|
|
void set_th_acceptstep(const double th_acceptstep); |
| 415 |
|
|
|
| 416 |
|
|
/** |
| 417 |
|
|
* @brief Modify the tolerance for stopping the algorithm |
| 418 |
|
|
*/ |
| 419 |
|
|
void set_th_stop(const double th_stop); |
| 420 |
|
|
|
| 421 |
|
|
/** |
| 422 |
|
|
* @brief Modify the threshold for accepting a gap as non-zero |
| 423 |
|
|
*/ |
| 424 |
|
|
void set_th_gaptol(const double th_gaptol); |
| 425 |
|
|
|
| 426 |
|
|
/** |
| 427 |
|
|
* @brief Modify the current norm used for computed the dynamic and constraint |
| 428 |
|
|
* feasibility |
| 429 |
|
|
*/ |
| 430 |
|
|
void set_feasnorm(const FeasibilityNorm feas_norm); |
| 431 |
|
|
|
| 432 |
|
|
protected: |
| 433 |
|
|
std::shared_ptr<ShootingProblem> problem_; //!< optimal control problem |
| 434 |
|
|
std::vector<Eigen::VectorXd> xs_; //!< State trajectory |
| 435 |
|
|
std::vector<Eigen::VectorXd> us_; //!< Control trajectory |
| 436 |
|
|
std::vector<Eigen::VectorXd> fs_; //!< Gaps/defects between shooting nodes |
| 437 |
|
|
std::vector<std::shared_ptr<CallbackAbstract> > |
| 438 |
|
|
callbacks_; //!< Callback functions |
| 439 |
|
|
bool is_feasible_; //!< Label that indicates is the iteration is feasible |
| 440 |
|
|
bool was_feasible_; //!< Label that indicates in the previous iterate was |
| 441 |
|
|
//!< feasible |
| 442 |
|
|
double cost_; //!< Cost for the current guess |
| 443 |
|
|
double merit_; //!< Merit for the current guess |
| 444 |
|
|
double stop_; //!< Value computed by `stoppingCriteria()` |
| 445 |
|
|
Eigen::Vector2d d_; //!< LQ approximation of the expected improvement |
| 446 |
|
|
double dV_; //!< Reduction in the cost function computed by `tryStep()` |
| 447 |
|
|
double dPhi_; //!< Reduction in the merit function computed by `tryStep()` |
| 448 |
|
|
double dVexp_; //!< Expected reduction in the cost function |
| 449 |
|
|
double dPhiexp_; //!< Expected reduction in the merit function |
| 450 |
|
|
double dfeas_; //!< Reduction in the feasibility |
| 451 |
|
|
double feas_; //!< Total feasibility for the current guess |
| 452 |
|
|
double |
| 453 |
|
|
ffeas_; //!< Feasibility of the dynamic constraints for the current guess |
| 454 |
|
|
double gfeas_; //!< Feasibility of the inequality constraints for the current |
| 455 |
|
|
//!< guess |
| 456 |
|
|
double hfeas_; //!< Feasibility of the equality constraints for the current |
| 457 |
|
|
//!< guess |
| 458 |
|
|
double ffeas_try_; //!< Feasibility of the dynamic constraints evaluated for |
| 459 |
|
|
//!< the current step length |
| 460 |
|
|
double gfeas_try_; //!< Feasibility of the inequality constraints evaluated |
| 461 |
|
|
//!< for the current step length |
| 462 |
|
|
double hfeas_try_; //!< Feasibility of the equality constraints evaluated for |
| 463 |
|
|
//!< the current step length |
| 464 |
|
|
double preg_; //!< Current primal-variable regularization value |
| 465 |
|
|
double dreg_; //!< Current dual-variable regularization value |
| 466 |
|
|
DEPRECATED("Use preg_ for primal-variable regularization", |
| 467 |
|
|
double xreg_;) //!< Current state regularization value |
| 468 |
|
|
DEPRECATED("Use dreg_ for primal-variable regularization", |
| 469 |
|
|
double ureg_;) //!< Current control regularization values |
| 470 |
|
|
double steplength_; //!< Current applied step length |
| 471 |
|
|
double th_acceptstep_; //!< Threshold used for accepting step |
| 472 |
|
|
double th_stop_; //!< Tolerance for stopping the algorithm |
| 473 |
|
|
double th_gaptol_; //!< Threshold limit to check non-zero gaps |
| 474 |
|
|
enum FeasibilityNorm feasnorm_; //!< Type of norm used to evaluate the |
| 475 |
|
|
//!< dynamics and constraints feasibility |
| 476 |
|
|
std::size_t iter_; //!< Number of iteration performed by the solver |
| 477 |
|
|
double tmp_feas_; //!< Temporal variables used for computed the feasibility |
| 478 |
|
|
std::vector<Eigen::VectorXd> g_adj_; //!< Adjusted inequality bound |
| 479 |
|
|
}; |
| 480 |
|
|
|
| 481 |
|
|
/** |
| 482 |
|
|
* @brief Abstract class for solver callbacks |
| 483 |
|
|
* |
| 484 |
|
|
* A callback is used to diagnostic the behaviour of our solver in each |
| 485 |
|
|
* iteration of it. For instance, it can be used to print values, record data or |
| 486 |
|
|
* display motions. |
| 487 |
|
|
*/ |
| 488 |
|
|
class CallbackAbstract { |
| 489 |
|
|
public: |
| 490 |
|
|
/** |
| 491 |
|
|
* @brief Initialize the callback function |
| 492 |
|
|
*/ |
| 493 |
|
26 |
CallbackAbstract() {} |
| 494 |
|
56 |
virtual ~CallbackAbstract() {} |
| 495 |
|
|
|
| 496 |
|
|
/** |
| 497 |
|
|
* @brief Run the callback function given a solver |
| 498 |
|
|
* |
| 499 |
|
|
* @param[in] solver solver to be diagnostic |
| 500 |
|
|
*/ |
| 501 |
|
|
virtual void operator()(SolverAbstract& solver) = 0; |
| 502 |
|
|
}; |
| 503 |
|
|
|
| 504 |
|
|
bool raiseIfNaN(const double value); |
| 505 |
|
|
|
| 506 |
|
|
} // namespace crocoddyl |
| 507 |
|
|
|
| 508 |
|
|
#endif // CROCODDYL_CORE_SOLVER_BASE_HPP_ |
| 509 |
|
|
|