nlp.hpp

Go to the documentation of this file.

// Copyright (C) 2022 Petter Nilsson. MIT License.
 
#pragma once
 
#include <limits>
#include <optional>
 
#include <Eigen/Core>
#include <Eigen/Sparse>
 
namespace smooth::feedback {
 
template<typename T>
concept NLP = requires(
  std::decay_t<T> & nlp, const Eigen::Ref<const Eigen::VectorXd> x, const Eigen::Ref<const Eigen::VectorXd> lambda)
{
  // clang-format off
  {nlp.n()} -> std::convertible_to<std::size_t>;
  {nlp.m()} -> std::convertible_to<std::size_t>;
 
  // variable bounds
  {nlp.xl()} -> std::convertible_to<Eigen::VectorXd>;
  {nlp.xu()} -> std::convertible_to<Eigen::VectorXd>;
 
  // objective and its derivatives
  {nlp.f(x)}     -> std::convertible_to<double>;
  {nlp.df_dx(x)} -> std::convertible_to<Eigen::SparseMatrix<double>>;
 
  // constraint, its bounds, and its derivatives
  {nlp.g(x)}     -> std::convertible_to<Eigen::VectorXd>;
  {nlp.gl()}     -> std::convertible_to<Eigen::VectorXd>;
  {nlp.gu()}     -> std::convertible_to<Eigen::VectorXd>;
  {nlp.dg_dx(x)} -> std::convertible_to<Eigen::SparseMatrix<double>>;
  // clang-format on
};
 
template<typename T>
concept HessianNLP = NLP<T> && requires(std::decay_t<T> & nlp, Eigen::VectorXd x, Eigen::VectorXd lambda)
{
  // clang-format off
  {nlp.d2f_dx2(x)}         -> std::convertible_to<Eigen::SparseMatrix<double>>;
  {nlp.d2g_dx2(x, lambda)} -> std::convertible_to<Eigen::SparseMatrix<double>>;
  // clang-format on
};
 
struct NLPSolution
{
  enum class Status {
    Optimal,
    PrimalInfeasible,
    DualInfeasible,
    MaxIterations,
    MaxTime,
    Unknown,
  };
 
  Status status;
 
  std::size_t iter{0};
 
  Eigen::VectorXd x;
 
  Eigen::VectorXd zl, zu;
 
  Eigen::VectorXd lambda;
 
  double objective{0};
};
 
}  // namespace smooth::feedback