gtsam/tests/testIterative.cpp

/* ----------------------------------------------------------------------------

 * 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   testIterative.cpp
 *  @brief  Unit tests for iterative methods
 *  @author Frank Dellaert
 **/

#include <tests/smallExample.h>
#include <gtsam/slam/PriorFactor.h>
#include <gtsam/slam/BetweenFactor.h>
#include <gtsam/nonlinear/NonlinearEquality.h>
#include <gtsam/nonlinear/Ordering.h>
#include <gtsam/nonlinear/Symbol.h>
#include <gtsam/linear/GaussianSequentialSolver.h>
#include <gtsam/linear/iterative.h>
#include <gtsam/geometry/Pose2.h>

#include <CppUnitLite/TestHarness.h>

using namespace std;
using namespace gtsam;
using namespace example;
using symbol_shorthand::X; // to create pose keys
using symbol_shorthand::L; // to create landmark keys

static ConjugateGradientParameters parameters;
// add following below to add printing:
// parameters.verbosity_ = ConjugateGradientParameters::COMPLEXITY;

/* ************************************************************************* */
TEST( Iterative, steepestDescent )
{
  // Create factor graph
  Ordering ordering;
  ordering += L(1), X(1), X(2);
  GaussianFactorGraph fg = createGaussianFactorGraph(ordering);

  // eliminate and solve
  VectorValues expected = *GaussianSequentialSolver(fg).optimize();

  // Do gradient descent
  VectorValues zero = VectorValues::Zero(expected); // TODO, how do we do this normally?
  VectorValues actual = steepestDescent(fg, zero, parameters);
  CHECK(assert_equal(expected,actual,1e-2));
}

/* ************************************************************************* */
TEST( Iterative, conjugateGradientDescent )
{
  // Create factor graph
  Ordering ordering;
  ordering += L(1), X(1), X(2);
  GaussianFactorGraph fg = createGaussianFactorGraph(ordering);

  // eliminate and solve
  VectorValues expected = *GaussianSequentialSolver(fg).optimize();

  // get matrices
  Matrix A;
  Vector b;
  Vector x0 = gtsam::zero(6);
  boost::tie(A, b) = fg.jacobian();
  Vector expectedX = Vector_(6, -0.1, 0.1, -0.1, -0.1, 0.1, -0.2);

  // Do conjugate gradient descent, System version
  System Ab(A, b);
  Vector actualX = conjugateGradientDescent(Ab, x0, parameters);
  CHECK(assert_equal(expectedX,actualX,1e-9));

  // Do conjugate gradient descent, Matrix version
  Vector actualX2 = conjugateGradientDescent(A, b, x0, parameters);
  CHECK(assert_equal(expectedX,actualX2,1e-9));

  // Do conjugate gradient descent on factor graph
  VectorValues zero = VectorValues::Zero(expected);
  VectorValues actual = conjugateGradientDescent(fg, zero, parameters);
  CHECK(assert_equal(expected,actual,1e-2));
}

/* ************************************************************************* */
TEST( Iterative, conjugateGradientDescent_hard_constraint )
{
  Values config;
  Pose2 pose1 = Pose2(0.,0.,0.);
  config.insert(X(1), pose1);
  config.insert(X(2), Pose2(1.5,0.,0.));

  NonlinearFactorGraph graph;
  graph.add(NonlinearEquality<Pose2>(X(1), pose1));
  graph.add(BetweenFactor<Pose2>(X(1),X(2), Pose2(1.,0.,0.), noiseModel::Isotropic::Sigma(3, 1)));

  Ordering ordering;
  ordering += X(1), X(2);
  boost::shared_ptr<GaussianFactorGraph> fg = graph.linearize(config,ordering);

  VectorValues zeros = VectorValues::Zero(2, 3);

  ConjugateGradientParameters parameters;
  parameters.setEpsilon_abs(1e-3);
  parameters.setEpsilon_rel(1e-5);
  parameters.setMaxIterations(100);
  VectorValues actual = conjugateGradientDescent(*fg, zeros, parameters);

  VectorValues expected;
  expected.insert(0, zero(3));
  expected.insert(1, Vector_(3,-0.5,0.,0.));
  CHECK(assert_equal(expected, actual));
}

/* ************************************************************************* */
TEST( Iterative, conjugateGradientDescent_soft_constraint )
{
  Values config;
  config.insert(X(1), Pose2(0.,0.,0.));
  config.insert(X(2), Pose2(1.5,0.,0.));

  NonlinearFactorGraph graph;
  graph.add(PriorFactor<Pose2>(X(1), Pose2(0.,0.,0.), noiseModel::Isotropic::Sigma(3, 1e-10)));
  graph.add(BetweenFactor<Pose2>(X(1),X(2), Pose2(1.,0.,0.), noiseModel::Isotropic::Sigma(3, 1)));

  Ordering ordering;
  ordering += X(1), X(2);
  boost::shared_ptr<GaussianFactorGraph> fg = graph.linearize(config,ordering);

  VectorValues zeros = VectorValues::Zero(2, 3);

  ConjugateGradientParameters parameters;
  parameters.setEpsilon_abs(1e-3);
  parameters.setEpsilon_rel(1e-5);
  parameters.setMaxIterations(100);
  VectorValues actual = conjugateGradientDescent(*fg, zeros, parameters);

  VectorValues expected;
  expected.insert(0, zero(3));
  expected.insert(1, Vector_(3,-0.5,0.,0.));
  CHECK(assert_equal(expected, actual));
}

/* ************************************************************************* */
int main() {
  TestResult tr;
  return TestRegistry::runAllTests(tr);
}
/* ************************************************************************* */