| Line |
Branch |
Exec |
Source |
| 1 |
|
|
// |
| 2 |
|
|
// Copyright (c) 2022 INRIA |
| 3 |
|
|
// |
| 4 |
|
|
|
| 5 |
|
|
#include <iostream> |
| 6 |
|
|
|
| 7 |
|
|
#include <pinocchio/math/eigenvalues.hpp> |
| 8 |
|
|
|
| 9 |
|
|
#include <boost/variant.hpp> // to avoid C99 warnings |
| 10 |
|
|
|
| 11 |
|
|
#include <boost/test/unit_test.hpp> |
| 12 |
|
|
#include <boost/utility/binary.hpp> |
| 13 |
|
|
|
| 14 |
|
|
#include <Eigen/Eigenvalues> |
| 15 |
|
|
#include <algorithm> |
| 16 |
|
|
|
| 17 |
|
|
BOOST_AUTO_TEST_SUITE(BOOST_TEST_MODULE) |
| 18 |
|
|
|
| 19 |
|
|
using namespace pinocchio; |
| 20 |
|
|
|
| 21 |
|
✗ |
BOOST_AUTO_TEST_CASE(test_identity) |
| 22 |
|
|
{ |
| 23 |
|
✗ |
const Eigen::DenseIndex mat_size = 20; |
| 24 |
|
✗ |
const Eigen::MatrixXd identity_mat = Eigen::MatrixXd::Identity(mat_size, mat_size); |
| 25 |
|
|
|
| 26 |
|
✗ |
auto eigen_vec = computeLargestEigenvector(identity_mat); |
| 27 |
|
✗ |
auto eigen_val = retrieveLargestEigenvalue(eigen_vec); |
| 28 |
|
✗ |
BOOST_CHECK(std::fabs(eigen_val - 1.) <= 1e-4); |
| 29 |
|
|
} |
| 30 |
|
|
|
| 31 |
|
✗ |
BOOST_AUTO_TEST_CASE(test_random_matrix) |
| 32 |
|
|
{ |
| 33 |
|
✗ |
const Eigen::DenseIndex mat_size = 10; |
| 34 |
|
✗ |
const int num_tests = 1000; |
| 35 |
|
✗ |
const int num_it_max = 10; |
| 36 |
|
✗ |
const int num_it_max_finer = 100 * num_it_max; |
| 37 |
|
|
|
| 38 |
|
✗ |
for (int k = 0; k < num_tests; ++k) |
| 39 |
|
|
{ |
| 40 |
|
✗ |
const Eigen::MatrixXd A = Eigen::MatrixXd::Random(mat_size, mat_size); |
| 41 |
|
✗ |
const Eigen::MatrixXd sym_mat = (A * A.transpose()); |
| 42 |
|
|
|
| 43 |
|
✗ |
const Eigen::EigenSolver<Eigen::MatrixXd> eigen_solver(sym_mat, true); |
| 44 |
|
✗ |
BOOST_CHECK(eigen_solver.info() == Eigen::Success); |
| 45 |
|
|
|
| 46 |
|
|
// Sort eigenvalues |
| 47 |
|
✗ |
Eigen::VectorXd sorted_eigen_values = eigen_solver.eigenvalues().real(); |
| 48 |
|
✗ |
std::sort(sorted_eigen_values.data(), sorted_eigen_values.data() + sorted_eigen_values.size()); |
| 49 |
|
|
|
| 50 |
|
✗ |
const double eigen_val_ref = sorted_eigen_values[sorted_eigen_values.size() - 1]; |
| 51 |
|
|
|
| 52 |
|
✗ |
auto eigen_vec = computeLargestEigenvector(sym_mat, num_it_max); |
| 53 |
|
✗ |
auto eigen_val = retrieveLargestEigenvalue(eigen_vec); |
| 54 |
|
|
|
| 55 |
|
|
bool test_relative_eigen_val = |
| 56 |
|
✗ |
std::fabs(eigen_val - eigen_val_ref) / (std::max)(eigen_val, eigen_val_ref) <= 1e-2; |
| 57 |
|
✗ |
bool test_eigen_vec = (sym_mat * eigen_vec).isApprox(eigen_val * eigen_vec, 1e-2); |
| 58 |
|
|
|
| 59 |
|
✗ |
if (!test_relative_eigen_val || !test_eigen_vec) |
| 60 |
|
|
{ |
| 61 |
|
✗ |
auto eigen_vec_finer = computeLargestEigenvector(sym_mat, num_it_max_finer); |
| 62 |
|
✗ |
auto eigen_val_finer = retrieveLargestEigenvalue(eigen_vec_finer); |
| 63 |
|
|
|
| 64 |
|
✗ |
test_relative_eigen_val = |
| 65 |
|
✗ |
std::fabs(eigen_val_finer - eigen_val_ref) / (std::max)(eigen_val_finer, eigen_val_ref) |
| 66 |
|
|
<= 1e-2; |
| 67 |
|
|
test_eigen_vec = |
| 68 |
|
✗ |
(sym_mat * eigen_vec_finer).isApprox(eigen_val_finer * eigen_vec_finer, 1e-2); |
| 69 |
|
✗ |
std::cout << "res: " << (sym_mat * eigen_vec - eigen_val * eigen_vec).norm() << std::endl; |
| 70 |
|
✗ |
std::cout << "eigen_val: " << eigen_val << std::endl; |
| 71 |
|
✗ |
std::cout << "eigen_val_finer: " << eigen_val_finer << std::endl; |
| 72 |
|
✗ |
std::cout << "eigen_val_ref: " << eigen_val_ref << std::endl; |
| 73 |
|
✗ |
std::cout << "res finer: " |
| 74 |
|
✗ |
<< (sym_mat * eigen_vec_finer - eigen_val_finer * eigen_vec_finer).norm() |
| 75 |
|
✗ |
<< std::endl; |
| 76 |
|
|
} |
| 77 |
|
✗ |
BOOST_CHECK(test_relative_eigen_val); |
| 78 |
|
✗ |
BOOST_CHECK(test_eigen_vec); |
| 79 |
|
|
} |
| 80 |
|
|
} |
| 81 |
|
|
|
| 82 |
|
|
BOOST_AUTO_TEST_SUITE_END() |
| 83 |
|
|
|