protot/3rdparty/fcl/test/geometry/shape/test_convex.cpp

533 lines
19 KiB
C++
Raw Permalink Normal View History

2018-12-23 11:20:54 +01:00
/*
* Software License Agreement (BSD License)
*
* Copyright (c) 2018. Toyota Research Institute
* All rights reserved.
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions
* are met:
*
* * Redistributions of source code must retain the above copyright
* notice, this list of conditions and the following disclaimer.
* * Redistributions in binary form must reproduce the above
* copyright notice, this list of conditions and the following
* disclaimer in the documentation and/or other materials provided
* with the distribution.
* * Neither the name of CNRS-LAAS and AIST nor the names of its
* contributors may be used to endorse or promote products derived
* from this software without specific prior written permission.
*
* THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
* "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
* LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS
* FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE
* COPYRIGHT OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,
* INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING,
* BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
* LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
* CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
* LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN
* ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
* POSSIBILITY OF SUCH DAMAGE.
*/
/** @author Sean Curtis (sean@tri.global) (2018) */
// Tests the implementation of a convex polytope geometry.
#include "fcl/geometry/shape/convex.h"
#include <vector>
#include <Eigen/StdVector>
#include <gtest/gtest.h>
#include "eigen_matrix_compare.h"
#include "fcl/common/types.h"
namespace fcl {
namespace {
using std::max;
// Necessary to satisfy Eigen's alignment requirements. See
// http://eigen.tuxfamily.org/dox-devel/group__TopicStlContainers.html#StlContainers_vector
template <typename S>
using PoseVector = std::vector<Transform3<S>,
Eigen::aligned_allocator<Transform3<S>>>;
// Utilities to print scalar type in error messages.
template <typename S>
struct ScalarString {
static std::string value() { return "unknown"; }
};
template <>
struct ScalarString<double> {
static std::string value() { return "double"; }
};
template <>
struct ScalarString<float> {
static std::string value() { return "float"; }
};
// Base definition of a "unit" convex polytope. Specific instances should define
// faces, vertices, and quantities such as volume, center of mass, and moment of
// inertia in terms of a scale factor.
template <typename S>
class Polytope {
public:
explicit Polytope(S scale)
: vertices_(std::make_shared<std::vector<Vector3<S>>>()),
polygons_(std::make_shared<std::vector<int>>()), scale_(scale) {}
Polytope(const Polytope &other)
: vertices_(std::make_shared<std::vector<Vector3<S>>>(*(other.vertices_))),
polygons_(std::make_shared<std::vector<int>>(*(other.polygons_))),
scale_(other.scale_) {}
virtual int face_count() const = 0;
virtual int vertex_count() const = 0;
virtual S volume() const = 0;
virtual Vector3<S> com() const = 0;
virtual Matrix3<S> principal_inertia_tensor() const = 0;
virtual std::string description() const = 0;
// The scale of the polytope to use with test tolerances.
S scale() const { return scale_; }
std::shared_ptr<const std::vector<Vector3<S>>> points() const {
return vertices_;
}
std::shared_ptr<const std::vector<int>> polygons() const {
return polygons_;
}
Convex<S> MakeConvex() const {
// The Polytope class makes the pointers to vertices and faces const access.
// The Convex class calls for non-const pointers. Temporarily const-casting
// them to make it compatible.
return Convex<S>(points(), face_count(), polygons());
}
Vector3<S> min_point() const {
Vector3<S> m;
m.setConstant(std::numeric_limits<S>::max());
for (const auto& v : *vertices_) {
for (int i = 0; i < 3; ++i) {
if (v(i) < m(i)) m(i) = v(i);
}
}
return m;
}
Vector3<S> max_point() const {
Vector3<S> m;
m.setConstant(-std::numeric_limits<S>::max());
for (const auto& v : *vertices_) {
for (int i = 0; i < 3; ++i) {
if (v(i) > m(i)) m(i) = v(i);
}
}
return m;
}
Vector3<S> aabb_center() const {
return (max_point() + min_point()) / 2;
}
S aabb_radius() const { return (min_point() - aabb_center()).norm(); }
void SetPose(const Transform3<S>& X_WP) {
for (auto& v : *vertices_) {
v = X_WP * v;
}
}
protected:
void add_vertex(const Vector3<S>& vertex) { vertices_->push_back(vertex); }
void add_face(std::initializer_list<int> indices) {
polygons_->push_back(static_cast<int>(indices.size()));
polygons_->insert(polygons_->end(), indices);
}
// Confirms the number of vertices and number of polygons matches the counts
// implied by vertex_count() and face_count(), respectively.
void confirm_data() {
// Confirm point count.
GTEST_ASSERT_EQ(vertex_count(), static_cast<int>(vertices_->size()));
// Confirm face count.
// Count the number of faces encoded in polygons_;
int count = 0;
int i = 0;
while (i < static_cast<int>(polygons_->size())) {
++count;
i += (*polygons_)[i] + 1;
}
GTEST_ASSERT_EQ(i, static_cast<int>(polygons_->size()))
<< "Badly defined polygons";
GTEST_ASSERT_EQ(face_count(), count);
}
private:
std::shared_ptr<std::vector<Vector3<S>>> vertices_;
std::shared_ptr<std::vector<int>> polygons_;
S scale_{};
};
// A simple regular tetrahedron with edges of length `scale` centered on the
// origin.
template <typename S>
class EquilateralTetrahedron : public Polytope<S> {
public:
// Constructs the tetrahedron (of edge length `s`).
explicit EquilateralTetrahedron(S scale) : Polytope<S>(scale), scale_(scale) {
// Tetrahedron vertices in the tet's canonical frame T. The tet is
// "centered" on the origin so that it's center of mass is simple [0, 0, 0].
const S z_base = -1 / S(2 * sqrt(6.));
Vector3<S> points_T[] = {{S(0.5), S(-0.5 / sqrt(3.)), z_base},
{S(-0.5), S(-0.5 / sqrt(3.)), z_base},
{S(0), S(1. / sqrt(3.)), z_base},
{S(0), S(0), S(sqrt(3. / 8))}};
for (const auto& v : points_T) {
this->add_vertex(scale * v);
};
// Now add the polygons
this->add_face({0, 1, 2});
this->add_face({1, 0, 3});
this->add_face({0, 2, 3});
this->add_face({2, 1, 3});
this->confirm_data();
}
// Properties of the polytope.
int face_count() const final { return 4; }
int vertex_count() const final { return 4; }
virtual S volume() const final {
// This assumes unit mass.
S s = this->scale();
return S(sqrt(2) / 12) * s * s * s;
}
virtual Vector3<S> com() const final { return Vector3<S>::Zero(); }
virtual Matrix3<S> principal_inertia_tensor() const {
// TODO(SeanCurtis-TRI): Replace this with a legitimate tensor.
throw std::logic_error("Not implemented yet");
};
std::string description() const final {
return "Tetrahedron with scale: " + std::to_string(this->scale());
}
private:
S scale_{0};
};
// A simple cube with sides of length `scale`.
template <typename S>
class Cube : public Polytope<S> {
public:
Cube(S scale) : Polytope<S>(scale) {
// Cube vertices in the cube's canonical frame C.
Vector3<S> points_C[] = {{S(-0.5), S(-0.5), S(-0.5)}, // v0
{S(0.5), S(-0.5), S(-0.5)}, // v1
{S(-0.5), S(0.5), S(-0.5)}, // v2
{S(0.5), S(0.5), S(-0.5)}, // v3
{S(-0.5), S(-0.5), S(0.5)}, // v4
{S(0.5), S(-0.5), S(0.5)}, // v5
{S(-0.5), S(0.5), S(0.5)}, // v6
{S(0.5), S(0.5), S(0.5)}}; // v7
for (const auto& v : points_C) {
this->add_vertex(scale * v);
}
// Now add the polygons
this->add_face({1, 3, 7, 5}); // +x
this->add_face({0, 4, 6, 2}); // -x
this->add_face({4, 5, 7, 6}); // +y
this->add_face({0, 2, 3, 1}); // -y
this->add_face({6, 7, 3, 2}); // +z
this->add_face({0, 1, 5, 4}); // -z
this->confirm_data();
}
// Polytope properties
int face_count() const final { return 6; }
int vertex_count() const final { return 8; }
virtual S volume() const final {
S s = this->scale();
return s * s * s;
}
virtual Vector3<S> com() const final { return Vector3<S>::Zero(); }
virtual Matrix3<S> principal_inertia_tensor() const {
S scale_sqd = this->scale() * this->scale();
// This assumes unit mass.
return Eigen::DiagonalMatrix<S, 3>(1. / 6., 1. / 6., 1. / 6.) * scale_sqd;
};
std::string description() const final {
return "Cube with scale: " + std::to_string(this->scale());
}
};
void testConvexConstruction() {
Cube<double> cube{1};
// Set the cube at some other location to make sure that the interior point
// test/ doesn't pass just because it initialized to zero.
Vector3<double> p_WB(1, 2, 3);
cube.SetPose(Transform3<double>(Eigen::Translation3d(p_WB)));
Convex<double> convex = cube.MakeConvex();
// This doesn't depend on the correct logic in the constructor. But this is
// as convenient a time as any to test that it reports the right node type.
EXPECT_EQ(convex.getNodeType(), GEOM_CONVEX);
// The constructor computes the interior point.
EXPECT_TRUE(CompareMatrices(convex.getInteriorPoint(), p_WB));
}
template <template <typename> class Shape, typename S>
void testAABBComputation(const Shape<S>& model, const Transform3<S>& X_WS) {
Shape<S> shape(model);
shape.SetPose(X_WS);
Convex<S> convex = shape.MakeConvex();
convex.computeLocalAABB();
typename constants<S>::Real eps = constants<S>::eps();
EXPECT_NEAR(shape.aabb_radius(), convex.aabb_radius, eps);
EXPECT_TRUE(CompareMatrices(shape.aabb_center(), convex.aabb_center, eps));
EXPECT_TRUE(CompareMatrices(shape.min_point(), convex.aabb_local.min_, eps));
EXPECT_TRUE(CompareMatrices(shape.max_point(), convex.aabb_local.max_, eps));
}
template <template <typename> class Shape, typename S>
void testVolume(const Shape<S>& model, const Transform3<S>& X_WS,
int bits_lost) {
// If we're losing more than 10 bits, then we have a major problem.
GTEST_ASSERT_LE(bits_lost, 10);
Shape<S> shape(model);
shape.SetPose(X_WS);
Convex<S> convex = shape.MakeConvex();
// We want the basic tolerance to be near machine precision. The invocation
// of this function indicates how many bits of precision are expected to be
// lost and the machine epsilon is modified to account for this.
typename constants<S>::Real eps = (1 << bits_lost) * constants<S>::eps();
// We want to do a *relative* comparison. We scale our eps by the volume so
// that large volumes have tolerances proportional to the actual true value.
S scale = max(shape.volume(), S(1));
EXPECT_NEAR(shape.volume(), convex.computeVolume(), eps * scale)
<< shape.description() << " at\n" << X_WS.matrix()
<< "\nusing scalar: " << ScalarString<S>::value();
}
template <template <typename> class Shape, typename S>
void testCenterOfMass(const Shape<S>& model, const Transform3<S>& X_WS,
int bits_lost) {
// If we're losing more than 10 bits, then we have a major problem.
GTEST_ASSERT_LE(bits_lost, 10);
Shape<S> shape(model);
shape.SetPose(X_WS);
Convex<S> convex = shape.MakeConvex();
// We want the basic tolerance to be near machine precision. The invocation
// of this function indicates how many bits of precision are expected to be
// lost and the machine epsilon is modified to account for this.
typename constants<S>::Real eps = (1 << bits_lost) * constants<S>::eps();
// We want to do a *relative* comparison. The center-of-mass calculation is a
// volume-weighted calculation. So, the relative tolerance should scale with
// volume.
S scale = max(shape.volume(), S(1));
EXPECT_TRUE(
CompareMatrices(X_WS * shape.com(), convex.computeCOM(), eps * scale))
<< shape.description() << " at\n" << X_WS.matrix()
<< "\nusing scalar: " << ScalarString<S>::value();
}
template <template <typename> class Shape, typename S>
void testMomentOfInertia(const Shape<S>& model, const Transform3<S>& X_WS,
int bits_lost) {
// If we're losing more than 10 bits, then we have a major problem.
GTEST_ASSERT_LE(bits_lost, 10);
Shape<S> shape(model);
shape.SetPose(X_WS);
Convex<S> convex = shape.MakeConvex();
// We want the basic tolerance to be near machine precision. The invocation
// of this function indicates how many bits of precision are expected to be
// lost and the machine epsilon is modified to account for this.
typename constants<S>::Real eps = (1 << bits_lost) * constants<S>::eps();
// We want to do a *relative* comparison. The inertia calculation is a
// volume-weighted calculation. So, the relative tolerance should scale with
// volume.
S scale = max(shape.volume(), S(1));
EXPECT_TRUE(
CompareMatrices(X_WS.linear().transpose() *
shape.principal_inertia_tensor() * X_WS.linear(),
convex.computeMomentofInertiaRelatedToCOM(), eps * scale))
<< shape.description() << " at\n" << X_WS.matrix()
<< "\nusing scalar: " << ScalarString<S>::value();
}
template <typename S>
PoseVector<S> GetPoses() {
PoseVector<S> poses;
// Identity.
poses.push_back(Transform3<S>::Identity());
Transform3<S> X_WS;
// 90-degree rotation around each axis, in turn.
for (int i = 0; i < 3; ++i) {
X_WS = Transform3<S>::Identity();
X_WS.linear() = AngleAxis<S>(constants<S>::pi() / 2,
Vector3<S>::Unit(i)).matrix();
poses.push_back(X_WS);
}
// Small angle away from identity.
X_WS.linear() = AngleAxis<S>(S(1e-5), Vector3<S>{1, 2, 3}.normalized())
.matrix();
poses.push_back(X_WS);
// 45-degree angle to move away from axis-aligned as much as possible.
X_WS.linear() = AngleAxis<S>(constants<S>::pi() / 4,
Vector3<S>{1, 2, 3}.normalized()).matrix();
poses.push_back(X_WS);
// We don't test translation because that would imply the geometry is
// defined away from its own frame's origin. And that's just a recklessly
// stupid thing to do. Given the *current* algorithms, this will degrade
// the answers based on the *distance* to the origin.
// TODO(SeanCurtis-TRI): When the algorithms are no longer sensitive to vertex
// position relative to the origin, add tests that show that.
return poses;
}
std::vector<double> get_test_scales() {
return std::vector<double>{0.001, 1, 1000.};
}
template <template <typename> class Shape, typename S>
void testLocalAABBComputation(const Shape<S>& shape) {
for (const auto& X_WP : GetPoses<S>()) {
testAABBComputation<Shape>(shape, X_WP);
}
}
template <template <typename> class Shape, typename S>
void testVolumeComputation(const Shape<S>& shape, int bits_lost) {
for (const auto& X_WP : GetPoses<S>()) {
testVolume<Shape>(shape, X_WP, bits_lost);
}
}
template <template <typename> class Shape, typename S>
void testCenterOfMassComputation(const Shape<S>& shape, int bits_lost) {
for (const auto& X_WP : GetPoses<S>()) {
testCenterOfMass<Shape>(shape, X_WP, bits_lost);
}
}
template <template <typename> class Shape, typename S>
void testMomentOfInertiaComputation(const Shape<S>& shape, int bits_lost) {
for (const auto& X_WP : GetPoses<S>()) {
testMomentOfInertia<Shape>(shape, X_WP, bits_lost);
}
}
GTEST_TEST(ConvexGeometry, Constructor) {
testConvexConstruction();
}
GTEST_TEST(ConvexGeometry, LocalAABBComputation_Cube) {
for (double scale : get_test_scales()) {
Cube<double> cube_d(scale);
testLocalAABBComputation(cube_d);
Cube<float> cube_f(static_cast<float>(scale));
testLocalAABBComputation(cube_f);
}
}
GTEST_TEST(ConvexGeometry, Volume_Cube) {
for (double scale : get_test_scales()) {
Cube<double> cube_d(scale);
testVolumeComputation(cube_d, 0);
Cube<float> cube_f(static_cast<float>(scale));
// Apparently, no bits of precision are lost (relative to machine precision)
// on the cube volume *except* for the *large* cube in single precision.
// The reason for this isn't obvious, but probably a coincidental artifact
// of the particular configuration.
const int bits_lost = scale > 1 ? 2 : 0;
testVolumeComputation(cube_f, bits_lost);
}
}
GTEST_TEST(ConvexGeometry, CenterOfMass_Cube) {
for (double scale : get_test_scales()) {
Cube<double> cube_d(scale);
testCenterOfMassComputation(cube_d, 0);
Cube<float> cube_f(static_cast<float>(scale));
testCenterOfMassComputation(cube_f, 0);
}
}
GTEST_TEST(ConvexGeometry, MomentOfInertia_Cube) {
for (double scale : get_test_scales()) {
Cube<double> cube_d(scale);
testMomentOfInertiaComputation(cube_d, 0);
Cube<float> cube_f(static_cast<float>(scale));
testMomentOfInertiaComputation(cube_f, 0);
}
}
GTEST_TEST(ConvexGeometry, LocalAABBComputation_Tetrahedron) {
for (double scale : get_test_scales()) {
EquilateralTetrahedron<double> tet_d(scale);
testLocalAABBComputation(tet_d);
EquilateralTetrahedron<float> tet_f(static_cast<float>(scale));
testLocalAABBComputation(tet_f);
}
}
GTEST_TEST(ConvexGeometry, Volume_Tetrahedron) {
for (double scale : get_test_scales()) {
EquilateralTetrahedron<double> tet_d(scale);
// Apparently, no bits of precision are lost (relative to machine precision)
// on the tet volume *except* for the *large* test in double precision.
// The reason for this isn't obvious, but probably a coincidental artifact
// of the particular configuration.
const int bits_lost = scale > 1 ? 1 : 0;
testVolumeComputation(tet_d, bits_lost);
EquilateralTetrahedron<float> tet_f(static_cast<float>(scale));
testVolumeComputation(tet_f, 0);
}
}
GTEST_TEST(ConvexGeometry, CenterOfMass_Tetrahedron) {
for (double scale : get_test_scales()) {
EquilateralTetrahedron<double> tet_d(scale);
testCenterOfMassComputation(tet_d, 0);
EquilateralTetrahedron<float> tet_f(static_cast<float>(scale));
testCenterOfMassComputation(tet_f, 0);
}
}
// TODO(SeanCurtis-TRI): Add Tetrahedron inertia unit test.
// TODO(SeanCurtis-TRI): Extend the moment of inertia test.
// Tesselate smooth geometries (sphere, ellipsoid, cone, etc) which have
// well-known closed-form values for the tensor product. Confirm that as
// the tesselation gets finer, that the answer converges to the reference
// solution.
} // namespace
} // namespace fcl
//==============================================================================
int main(int argc, char *argv[]) {
::testing::InitGoogleTest(&argc, argv);
return RUN_ALL_TESTS();
}