gtsam/gtsam_unstable/base/tests/testBAD.cpp

406 lines
12 KiB
C++

/* ----------------------------------------------------------------------------
* GTSAM Copyright 2010, Georgia Tech Research Corporation,
* Atlanta, Georgia 30332-0415
* All Rights Reserved
* Authors: Frank Dellaert, et al. (see THANKS for the full author list)
* See LICENSE for the license information
* -------------------------------------------------------------------------- */
/**
* @file testBAD.cpp
* @date September 18, 2014
* @author Frank Dellaert
* @brief unit tests for Block Automatic Differentiation
*/
#include <gtsam/nonlinear/NonlinearFactor.h>
#include <gtsam/geometry/Pose3.h>
#include <gtsam/geometry/Cal3_S2.h>
#include <gtsam/slam/GeneralSFMFactor.h>
#include <gtsam/inference/Key.h>
#include <gtsam/base/Testable.h>
#include <boost/make_shared.hpp>
#include <boost/foreach.hpp>
#include <CppUnitLite/TestHarness.h>
namespace gtsam {
///-----------------------------------------------------------------------------
/// Expression node. The superclass for objects that do the heavy lifting
/// An Expression<T> has a pointer to an ExpressionNode<T> underneath
/// allowing Expressions to have polymorphic behaviour even though they
/// are passed by value. This is the same way boost::function works.
/// http://loki-lib.sourceforge.net/html/a00652.html
template<class T>
class ExpressionNode {
public:
ExpressionNode(){}
virtual ~ExpressionNode(){}
virtual void getKeys(std::set<Key>& keys) const = 0;
virtual T value(const Values& values,
boost::optional<std::map<Key, Matrix>&> = boost::none) const = 0;
virtual ExpressionNode<T>* clone() const = 0;
};
/// Constant Expression
template<class T>
class ConstantExpression : public ExpressionNode<T> {
T value_;
public:
typedef T type;
/// Constructor with a value, yielding a constant
ConstantExpression(const T& value) :
value_(value) {
}
virtual ~ConstantExpression(){}
virtual void getKeys(std::set<Key>& /* keys */) const {}
virtual T value(const Values& values,
boost::optional<std::map<Key, Matrix>&> jacobians = boost::none) const {
return value_;
}
virtual ExpressionNode<T>* clone() const { return new ConstantExpression(*this); }
};
//-----------------------------------------------------------------------------
/// Leaf Expression
template<class T>
class LeafExpression : public ExpressionNode<T> {
Key key_;
public:
typedef T type;
/// Constructor with a single key
LeafExpression(Key key) :
key_(key) {
}
virtual ~LeafExpression(){}
virtual void getKeys(std::set<Key>& keys) const { keys.insert(key_); }
virtual T value(const Values& values,
boost::optional<std::map<Key, Matrix>&> jacobians = boost::none) const {
const T& value = values.at<T>(key_);
if( jacobians ) {
std::map<Key, Matrix>::iterator it = jacobians->find(key_);
if(it != jacobians->end()) {
it->second += Eigen::MatrixXd::Identity(value.dim(), value.dim());
} else {
(*jacobians)[key_] = Eigen::MatrixXd::Identity(value.dim(), value.dim());
}
}
return value;
}
virtual ExpressionNode<T>* clone() const { return new LeafExpression(*this); }
};
//-----------------------------------------------------------------------------
/// Unary Expression
template<class T, class E>
class UnaryExpression : public ExpressionNode<T> {
public:
typedef T (*function)(const E&, boost::optional<Matrix&>);
private:
boost::shared_ptr< ExpressionNode<E> > expression_;
function f_;
public:
typedef T type;
/// Constructor with a single key
UnaryExpression(function f, const ExpressionNode<E>& expression) :
expression_(expression.clone()), f_(f) {
}
virtual ~UnaryExpression(){}
virtual void getKeys(std::set<Key>& keys) const{ expression_->getKeys(keys); }
virtual T value(const Values& values,
boost::optional<std::map<Key, Matrix>&> jacobians = boost::none) const {
T value;
if(jacobians) {
Eigen::MatrixXd H;
value = f_(expression_->value(values, jacobians), H);
std::map<Key, Matrix>::iterator it = jacobians->begin();
for( ; it != jacobians->end(); ++it) {
it->second = H * it->second;
}
} else {
value = f_(expression_->value(values), boost::none);
}
return value;
}
virtual ExpressionNode<T>* clone() const { return new UnaryExpression(*this); }
};
//-----------------------------------------------------------------------------
/// Binary Expression
template<class T, class E1, class E2>
class BinaryExpression : public ExpressionNode<T> {
public:
typedef T (*function)(const E1&, const E2&,
boost::optional<Matrix&>, boost::optional<Matrix&>);
private:
boost::shared_ptr< ExpressionNode<E1> > expression1_;
boost::shared_ptr< ExpressionNode<E2> > expression2_;
function f_;
public:
typedef T type;
/// Constructor with a single key
BinaryExpression(function f, const ExpressionNode<E1>& expression1, const ExpressionNode<E2>& expression2) :
expression1_(expression1.clone()), expression2_(expression2.clone()), f_(f) {
}
virtual ~BinaryExpression(){}
virtual void getKeys(std::set<Key>& keys) const{
expression1_->getKeys(keys);
expression2_->getKeys(keys);
}
virtual T value(const Values& values,
boost::optional<std::map<Key, Matrix>&> jacobians = boost::none) const {
T val;
if(jacobians) {
std::map<Key, Matrix> terms1;
std::map<Key, Matrix> terms2;
Matrix H1, H2;
val = f_(expression1_->value(values, terms1), expression2_->value(values, terms2), H1, H2);
// TODO: both Jacobians and terms are sorted. There must be a simple
// but fast algorithm that does this.
typedef std::pair<Key, Matrix> Pair;
BOOST_FOREACH(const Pair& term, terms1) {
std::map<Key, Matrix>::iterator it = jacobians->find(term.first);
if(it != jacobians->end()) {
it->second += H1 * term.second;
} else {
(*jacobians)[term.first] = H1 * term.second;
}
}
BOOST_FOREACH(const Pair& term, terms2) {
std::map<Key, Matrix>::iterator it = jacobians->find(term.first);
if(it != jacobians->end()) {
it->second += H2 * term.second;
} else {
(*jacobians)[term.first] = H2 * term.second;
}
}
} else {
val = f_(expression1_->value(values), expression2_->value(values),
boost::none, boost::none);
}
return val;
}
virtual ExpressionNode<T>* clone() const { return new BinaryExpression(*this); }
};
template<typename T>
class Expression {
public:
Expression(const ExpressionNode<T>& root) {
root_.reset(root.clone());
}
// Initialize a constant expression
Expression(const T& value) :
root_(new ConstantExpression<T>(value)){ }
// Initialize a leaf expression
Expression(const Key& key) :
root_(new LeafExpression<T>(key)) {}
/// Initialize a unary expression
template<typename E>
Expression(typename UnaryExpression<T,E>::function f,
const Expression<E>& expression) {
// TODO Assert that root of expression is not null.
root_.reset(new UnaryExpression<T,E>(f, *expression.root()));
}
/// Initialize a binary expression
template<typename E1, typename E2>
Expression(typename BinaryExpression<T,E1,E2>::function f,
const Expression<E1>& expression1,
const Expression<E2>& expression2) {
// TODO Assert that root of expressions 1 and 2 are not null.
root_.reset(new BinaryExpression<T,E1,E2>(f, *expression1.root(),
*expression2.root()));
}
void getKeys(std::set<Key>& keys) const { root_->getKeys(keys); }
T value(const Values& values,
boost::optional<std::map<Key, Matrix>&> jacobians = boost::none) const {
return root_->value(values, jacobians);
}
const boost::shared_ptr<ExpressionNode<T> >& root() const{ return root_; }
private:
boost::shared_ptr<ExpressionNode<T> > root_;
};
//-----------------------------------------------------------------------------
void printPair(std::pair<Key, Matrix> pair) {
std::cout << pair.first << ": " << pair.second << std::endl;
}
// usage: std::for_each(terms.begin(), terms.end(), printPair);
//-----------------------------------------------------------------------------
/// AD Factor
template<class T>
class BADFactor: NonlinearFactor {
const T measurement_;
const Expression<T> expression_;
/// get value from expression and calculate error with respect to measurement
Vector unwhitenedError(const Values& values) const {
const T& value = expression_.value(values);
return value.localCoordinates(measurement_);
}
public:
/// Constructor
BADFactor(const T& measurement, const Expression<T>& expression) :
measurement_(measurement), expression_(expression) {
}
/// Constructor
BADFactor(const T& measurement, const ExpressionNode<T>& expression) :
measurement_(measurement), expression_(expression) {
}
/**
* Calculate the error of the factor.
* This is the log-likelihood, e.g. \f$ 0.5(h(x)-z)^2/\sigma^2 \f$ in case of Gaussian.
* In this class, we take the raw prediction error \f$ h(x)-z \f$, ask the noise model
* to transform it to \f$ (h(x)-z)^2/\sigma^2 \f$, and then multiply by 0.5.
*/
virtual double error(const Values& values) const {
if (this->active(values)) {
const Vector e = unwhitenedError(values);
return 0.5 * e.squaredNorm();
} else {
return 0.0;
}
}
/// get the dimension of the factor (number of rows on linearization)
size_t dim() const {
return 0;
}
/// linearize to a GaussianFactor
boost::shared_ptr<GaussianFactor> linearize(const Values& values) const {
// We will construct an n-ary factor below, where terms is a container whose
// value type is std::pair<Key, Matrix>, specifying the
// collection of keys and matrices making up the factor.
std::map<Key, Matrix> terms;
expression_.value(values, terms);
Vector b = unwhitenedError(values);
SharedDiagonal model = SharedDiagonal();
return boost::shared_ptr<JacobianFactor>(
new JacobianFactor(terms, b, model));
}
};
}
using namespace std;
using namespace gtsam;
/* ************************************************************************* */
Point3 transformTo(const Pose3& x, const Point3& p,
boost::optional<Matrix&> Dpose, boost::optional<Matrix&> Dpoint) {
return x.transform_to(p, Dpose, Dpoint);
}
Point2 project(const Point3& p, boost::optional<Matrix&> Dpoint) {
return PinholeCamera<Cal3_S2>::project_to_camera(p, Dpoint);
}
template<class CAL>
Point2 uncalibrate(const CAL& K, const Point2& p, boost::optional<Matrix&> Dcal,
boost::optional<Matrix&> Dp) {
return K.uncalibrate(p, Dcal, Dp);
}
/* ************************************************************************* */
TEST(BAD, test) {
// Create some values
Values values;
values.insert(1, Pose3());
values.insert(2, Point3(0, 0, 1));
values.insert(3, Cal3_S2());
// Create old-style factor to create expected value and derivatives
Point2 measured(-17, 30);
SharedNoiseModel model = noiseModel::Unit::Create(2);
GeneralSFMFactor2<Cal3_S2> old(measured, model, 1, 2, 3);
double expected_error = old.error(values);
GaussianFactor::shared_ptr expected = old.linearize(values);
// Create leaves
Expression<Pose3> x(1);
Expression<Point3> p(2);
Expression<Cal3_S2> K(3);
// Create expression tree
Expression<Point3> p_cam(transformTo, x, p);
Expression<Point2> projection(project, p_cam);
Expression<Point2> uv_hat(uncalibrate, K, projection);
// Check getKeys
std::set<Key> keys;
uv_hat.getKeys(keys);
EXPECT_LONGS_EQUAL(3, keys.size());
// Create factor
BADFactor<Point2> f(measured, uv_hat);
// Check value
EXPECT_DOUBLES_EQUAL(expected_error, f.error(values), 1e-9);
// Check dimension
EXPECT_LONGS_EQUAL(0, f.dim());
// Check linearization
boost::shared_ptr<GaussianFactor> gf = f.linearize(values);
EXPECT( assert_equal(*expected, *gf, 1e-9));
}
/* ************************************************************************* */
int main() {
TestResult tr;
return TestRegistry::runAllTests(tr);
}
/* ************************************************************************* */