blob: 7e90177cfe4e8917361b82518891e1e343b3d99c [file] [log] [blame]
Sergey Sharybineeedd3a2013-04-10 23:58:32 +06001// 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@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
38namespace ceres {
39namespace internal {
40
41struct 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
52TEST(AutoDiffLocalParameterizationTest, IdentityParameterization) {
53 AutoDiffLocalParameterization<IdentityPlus, 3, 3>
54 parameterization;
55
56 double x[3] = {1.0, 2.0, 3.0};
57 double delta[3] = {0.0, 1.0, 2.0};
58 double x_plus_delta[3] = {0.0, 0.0, 0.0};
59 parameterization.Plus(x, delta, x_plus_delta);
60
61 EXPECT_EQ(x_plus_delta[0], 1.0);
62 EXPECT_EQ(x_plus_delta[1], 3.0);
63 EXPECT_EQ(x_plus_delta[2], 5.0);
64
65 double jacobian[9];
66 parameterization.ComputeJacobian(x, jacobian);
67 int k = 0;
68 for (int i = 0; i < 3; ++i) {
69 for (int j = 0; j < 3; ++j, ++k) {
70 EXPECT_EQ(jacobian[k], (i == j) ? 1.0 : 0.0);
71 }
72 }
73}
74
75struct QuaternionPlus {
76 template<typename T>
77 bool operator()(const T* x, const T* delta, T* x_plus_delta) const {
78 const T squared_norm_delta =
79 delta[0] * delta[0] + delta[1] * delta[1] + delta[2] * delta[2];
80
81 T q_delta[4];
82 if (squared_norm_delta > T(0.0)) {
83 T norm_delta = sqrt(squared_norm_delta);
84 const T sin_delta_by_delta = sin(norm_delta) / norm_delta;
85 q_delta[0] = cos(norm_delta);
86 q_delta[1] = sin_delta_by_delta * delta[0];
87 q_delta[2] = sin_delta_by_delta * delta[1];
88 q_delta[3] = sin_delta_by_delta * delta[2];
89 } else {
90 // We do not just use q_delta = [1,0,0,0] here because that is a
91 // constant and when used for automatic differentiation will
92 // lead to a zero derivative. Instead we take a first order
93 // approximation and evaluate it at zero.
94 q_delta[0] = T(1.0);
95 q_delta[1] = delta[0];
96 q_delta[2] = delta[1];
97 q_delta[3] = delta[2];
98 }
99
100 QuaternionProduct(q_delta, x, x_plus_delta);
101 return true;
102 }
103};
104
105void QuaternionParameterizationTestHelper(const double* x,
106 const double* delta) {
107 const double kTolerance = 1e-14;
108 double x_plus_delta_ref[4] = {0.0, 0.0, 0.0, 0.0};
109 double jacobian_ref[12];
110
111
112 QuaternionParameterization ref_parameterization;
113 ref_parameterization.Plus(x, delta, x_plus_delta_ref);
114 ref_parameterization.ComputeJacobian(x, jacobian_ref);
115
116 double x_plus_delta[4] = {0.0, 0.0, 0.0, 0.0};
117 double jacobian[12];
118 AutoDiffLocalParameterization<QuaternionPlus, 4, 3> parameterization;
119 parameterization.Plus(x, delta, x_plus_delta);
120 parameterization.ComputeJacobian(x, jacobian);
121
122 for (int i = 0; i < 4; ++i) {
123 EXPECT_NEAR(x_plus_delta[i], x_plus_delta_ref[i], kTolerance);
124 }
125
126 const double x_plus_delta_norm =
127 sqrt(x_plus_delta[0] * x_plus_delta[0] +
128 x_plus_delta[1] * x_plus_delta[1] +
129 x_plus_delta[2] * x_plus_delta[2] +
130 x_plus_delta[3] * x_plus_delta[3]);
131
132 EXPECT_NEAR(x_plus_delta_norm, 1.0, kTolerance);
133
134 for (int i = 0; i < 12; ++i) {
135 EXPECT_TRUE(IsFinite(jacobian[i]));
136 EXPECT_NEAR(jacobian[i], jacobian_ref[i], kTolerance)
137 << "Jacobian mismatch: i = " << i
138 << "\n Expected \n" << ConstMatrixRef(jacobian_ref, 4, 3)
139 << "\n Actual \n" << ConstMatrixRef(jacobian, 4, 3);
140 }
141}
142
143TEST(AutoDiffLocalParameterization, QuaternionParameterizationZeroTest) {
144 double x[4] = {0.5, 0.5, 0.5, 0.5};
145 double delta[3] = {0.0, 0.0, 0.0};
146 QuaternionParameterizationTestHelper(x, delta);
147}
148
149
150TEST(AutoDiffLocalParameterization, QuaternionParameterizationNearZeroTest) {
151 double x[4] = {0.52, 0.25, 0.15, 0.45};
152 double norm_x = sqrt(x[0] * x[0] +
153 x[1] * x[1] +
154 x[2] * x[2] +
155 x[3] * x[3]);
156 for (int i = 0; i < 4; ++i) {
157 x[i] = x[i] / norm_x;
158 }
159
160 double delta[3] = {0.24, 0.15, 0.10};
161 for (int i = 0; i < 3; ++i) {
162 delta[i] = delta[i] * 1e-14;
163 }
164
165 QuaternionParameterizationTestHelper(x, delta);
166}
167
168TEST(AutoDiffLocalParameterization, QuaternionParameterizationNonZeroTest) {
169 double x[4] = {0.52, 0.25, 0.15, 0.45};
170 double norm_x = sqrt(x[0] * x[0] +
171 x[1] * x[1] +
172 x[2] * x[2] +
173 x[3] * x[3]);
174
175 for (int i = 0; i < 4; ++i) {
176 x[i] = x[i] / norm_x;
177 }
178
179 double delta[3] = {0.24, 0.15, 0.10};
180 QuaternionParameterizationTestHelper(x, delta);
181}
182
183} // namespace internal
184} // namespace ceres