Home | History | Annotate | Download | only in ceres
      1 // Ceres Solver - A fast non-linear least squares minimizer
      2 // Copyright 2013 Google Inc. All rights reserved.
      3 // http://code.google.com/p/ceres-solver/
      4 //
      5 // Redistribution and use in source and binary forms, with or without
      6 // modification, are permitted provided that the following conditions are met:
      7 //
      8 // * Redistributions of source code must retain the above copyright notice,
      9 //   this list of conditions and the following disclaimer.
     10 // * Redistributions in binary form must reproduce the above copyright notice,
     11 //   this list of conditions and the following disclaimer in the documentation
     12 //   and/or other materials provided with the distribution.
     13 // * Neither the name of Google Inc. nor the names of its contributors may be
     14 //   used to endorse or promote products derived from this software without
     15 //   specific prior written permission.
     16 //
     17 // THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
     18 // AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
     19 // IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
     20 // ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE
     21 // LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
     22 // CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
     23 // SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
     24 // INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
     25 // CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
     26 // ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
     27 // POSSIBILITY OF SUCH DAMAGE.
     28 //
     29 // Author: sameeragarwal (at) google.com (Sameer Agarwal)
     30 
     31 #include <cmath>
     32 #include "ceres/autodiff_local_parameterization.h"
     33 #include "ceres/fpclassify.h"
     34 #include "ceres/local_parameterization.h"
     35 #include "ceres/rotation.h"
     36 #include "gtest/gtest.h"
     37 
     38 namespace ceres {
     39 namespace internal {
     40 
     41 struct IdentityPlus {
     42   template <typename T>
     43   bool operator()(const T* x, const T* delta, T* x_plus_delta) const {
     44     for (int i = 0; i < 3; ++i) {
     45       x_plus_delta[i] = x[i] + delta[i];
     46     }
     47     return true;
     48   }
     49 };
     50 
     51 TEST(AutoDiffLocalParameterizationTest, IdentityParameterization) {
     52   AutoDiffLocalParameterization<IdentityPlus, 3, 3>
     53       parameterization;
     54 
     55   double x[3] = {1.0, 2.0, 3.0};
     56   double delta[3] = {0.0, 1.0, 2.0};
     57   double x_plus_delta[3] = {0.0, 0.0, 0.0};
     58   parameterization.Plus(x, delta, x_plus_delta);
     59 
     60   EXPECT_EQ(x_plus_delta[0], 1.0);
     61   EXPECT_EQ(x_plus_delta[1], 3.0);
     62   EXPECT_EQ(x_plus_delta[2], 5.0);
     63 
     64   double jacobian[9];
     65   parameterization.ComputeJacobian(x, jacobian);
     66   int k = 0;
     67   for (int i = 0; i < 3; ++i) {
     68     for (int j = 0; j < 3; ++j, ++k) {
     69       EXPECT_EQ(jacobian[k], (i == j) ? 1.0 : 0.0);
     70     }
     71   }
     72 }
     73 
     74 struct ScaledPlus {
     75   ScaledPlus(const double &scale_factor)
     76      : scale_factor_(scale_factor)
     77   {}
     78 
     79   template <typename T>
     80   bool operator()(const T* x, const T* delta, T* x_plus_delta) const {
     81     for (int i = 0; i < 3; ++i) {
     82       x_plus_delta[i] = x[i] + T(scale_factor_) * delta[i];
     83     }
     84     return true;
     85   }
     86 
     87   const double scale_factor_;
     88 };
     89 
     90 TEST(AutoDiffLocalParameterizationTest, ScaledParameterization) {
     91   const double kTolerance = 1e-14;
     92 
     93   AutoDiffLocalParameterization<ScaledPlus, 3, 3>
     94       parameterization(new ScaledPlus(1.2345));
     95 
     96   double x[3] = {1.0, 2.0, 3.0};
     97   double delta[3] = {0.0, 1.0, 2.0};
     98   double x_plus_delta[3] = {0.0, 0.0, 0.0};
     99   parameterization.Plus(x, delta, x_plus_delta);
    100 
    101   EXPECT_NEAR(x_plus_delta[0], 1.0, kTolerance);
    102   EXPECT_NEAR(x_plus_delta[1], 3.2345, kTolerance);
    103   EXPECT_NEAR(x_plus_delta[2], 5.469, kTolerance);
    104 
    105   double jacobian[9];
    106   parameterization.ComputeJacobian(x, jacobian);
    107   int k = 0;
    108   for (int i = 0; i < 3; ++i) {
    109     for (int j = 0; j < 3; ++j, ++k) {
    110       EXPECT_NEAR(jacobian[k], (i == j) ? 1.2345 : 0.0, kTolerance);
    111     }
    112   }
    113 }
    114 
    115 struct QuaternionPlus {
    116   template<typename T>
    117   bool operator()(const T* x, const T* delta, T* x_plus_delta) const {
    118     const T squared_norm_delta =
    119         delta[0] * delta[0] + delta[1] * delta[1] + delta[2] * delta[2];
    120 
    121     T q_delta[4];
    122     if (squared_norm_delta > T(0.0)) {
    123       T norm_delta = sqrt(squared_norm_delta);
    124       const T sin_delta_by_delta = sin(norm_delta) / norm_delta;
    125       q_delta[0] = cos(norm_delta);
    126       q_delta[1] = sin_delta_by_delta * delta[0];
    127       q_delta[2] = sin_delta_by_delta * delta[1];
    128       q_delta[3] = sin_delta_by_delta * delta[2];
    129     } else {
    130       // We do not just use q_delta = [1,0,0,0] here because that is a
    131       // constant and when used for automatic differentiation will
    132       // lead to a zero derivative. Instead we take a first order
    133       // approximation and evaluate it at zero.
    134       q_delta[0] = T(1.0);
    135       q_delta[1] = delta[0];
    136       q_delta[2] = delta[1];
    137       q_delta[3] = delta[2];
    138     }
    139 
    140     QuaternionProduct(q_delta, x, x_plus_delta);
    141     return true;
    142   }
    143 };
    144 
    145 void QuaternionParameterizationTestHelper(const double* x,
    146                                           const double* delta) {
    147   const double kTolerance = 1e-14;
    148   double x_plus_delta_ref[4] = {0.0, 0.0, 0.0, 0.0};
    149   double jacobian_ref[12];
    150 
    151 
    152   QuaternionParameterization ref_parameterization;
    153   ref_parameterization.Plus(x, delta, x_plus_delta_ref);
    154   ref_parameterization.ComputeJacobian(x, jacobian_ref);
    155 
    156   double x_plus_delta[4] = {0.0, 0.0, 0.0, 0.0};
    157   double jacobian[12];
    158   AutoDiffLocalParameterization<QuaternionPlus, 4, 3> parameterization;
    159   parameterization.Plus(x, delta, x_plus_delta);
    160   parameterization.ComputeJacobian(x, jacobian);
    161 
    162   for (int i = 0; i < 4; ++i) {
    163     EXPECT_NEAR(x_plus_delta[i], x_plus_delta_ref[i], kTolerance);
    164   }
    165 
    166   const double x_plus_delta_norm =
    167       sqrt(x_plus_delta[0] * x_plus_delta[0] +
    168            x_plus_delta[1] * x_plus_delta[1] +
    169            x_plus_delta[2] * x_plus_delta[2] +
    170            x_plus_delta[3] * x_plus_delta[3]);
    171 
    172   EXPECT_NEAR(x_plus_delta_norm, 1.0, kTolerance);
    173 
    174   for (int i = 0; i < 12; ++i) {
    175     EXPECT_TRUE(IsFinite(jacobian[i]));
    176     EXPECT_NEAR(jacobian[i], jacobian_ref[i], kTolerance)
    177         << "Jacobian mismatch: i = " << i
    178         << "\n Expected \n" << ConstMatrixRef(jacobian_ref, 4, 3)
    179         << "\n Actual \n" << ConstMatrixRef(jacobian, 4, 3);
    180   }
    181 }
    182 
    183 TEST(AutoDiffLocalParameterization, QuaternionParameterizationZeroTest) {
    184   double x[4] = {0.5, 0.5, 0.5, 0.5};
    185   double delta[3] = {0.0, 0.0, 0.0};
    186   QuaternionParameterizationTestHelper(x, delta);
    187 }
    188 
    189 
    190 TEST(AutoDiffLocalParameterization, QuaternionParameterizationNearZeroTest) {
    191   double x[4] = {0.52, 0.25, 0.15, 0.45};
    192   double norm_x = sqrt(x[0] * x[0] +
    193                        x[1] * x[1] +
    194                        x[2] * x[2] +
    195                        x[3] * x[3]);
    196   for (int i = 0; i < 4; ++i) {
    197     x[i] = x[i] / norm_x;
    198   }
    199 
    200   double delta[3] = {0.24, 0.15, 0.10};
    201   for (int i = 0; i < 3; ++i) {
    202     delta[i] = delta[i] * 1e-14;
    203   }
    204 
    205   QuaternionParameterizationTestHelper(x, delta);
    206 }
    207 
    208 TEST(AutoDiffLocalParameterization, QuaternionParameterizationNonZeroTest) {
    209   double x[4] = {0.52, 0.25, 0.15, 0.45};
    210   double norm_x = sqrt(x[0] * x[0] +
    211                        x[1] * x[1] +
    212                        x[2] * x[2] +
    213                        x[3] * x[3]);
    214 
    215   for (int i = 0; i < 4; ++i) {
    216     x[i] = x[i] / norm_x;
    217   }
    218 
    219   double delta[3] = {0.24, 0.15, 0.10};
    220   QuaternionParameterizationTestHelper(x, delta);
    221 }
    222 
    223 }  // namespace internal
    224 }  // namespace ceres
    225