linear_solver_eigen.h

Go to the documentation of this file.

// g2o - General Graph Optimization
// Copyright (C) 2011 R. Kuemmerle, G. Grisetti, W. Burgard
// 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.
//
// 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
// HOLDER 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.
 
#ifndef G2O_LINEAR_SOLVER_EIGEN_H
#define G2O_LINEAR_SOLVER_EIGEN_H
 
#include <Eigen/Sparse>
#include <Eigen/SparseCholesky>
#include <cassert>
#include <iostream>
#include <vector>
 
#include "g2o/core/batch_stats.h"
#include "g2o/core/linear_solver.h"
#include "g2o/core/marginal_covariance_cholesky.h"
#include "g2o/stuff/logger.h"
#include "g2o/stuff/timeutil.h"
 
namespace g2o {
 
template <typename MatrixType>
class LinearSolverEigen : public LinearSolverCCS<MatrixType> {
 public:
  typedef Eigen::SparseMatrix<double, Eigen::ColMajor> SparseMatrix;
  typedef Eigen::Triplet<double> Triplet;
  typedef Eigen::PermutationMatrix<Eigen::Dynamic, Eigen::Dynamic>
      PermutationMatrix;
 
  using CholeskyDecompositionBase =
      Eigen::SimplicialLLT<SparseMatrix, Eigen::Upper>;
 
  class CholeskyDecomposition : public CholeskyDecompositionBase {
   public:
    CholeskyDecomposition() : CholeskyDecompositionBase() {}
 
    void analyzePatternWithPermutation(SparseMatrix& a,
                                       const PermutationMatrix& permutation) {
      m_Pinv = permutation;
      m_P = permutation.inverse();
      int size = a.cols();
      SparseMatrix ap(size, size);
      ap.selfadjointView<Eigen::Upper>() =
          a.selfadjointView<UpLo>().twistedBy(m_P);
      analyzePattern_preordered(ap, false);
    }
 
   protected:
    using CholeskyDecompositionBase::analyzePattern_preordered;
  };
 
 public:
  EIGEN_MAKE_ALIGNED_OPERATOR_NEW;
  LinearSolverEigen() : LinearSolverCCS<MatrixType>(), _init(true) {}
 
  virtual bool init() {
    _init = true;
    return true;
  }
 
  bool solve(const SparseBlockMatrix<MatrixType>& A, double* x, double* b) {
    double t;
    bool cholState = computeCholesky(A, t);
    if (!cholState) return false;
 
    // Solving the system
    VectorX::MapType xx(x, _sparseMatrix.cols());
    VectorX::ConstMapType bb(b, _sparseMatrix.cols());
    xx = _cholesky.solve(bb);
    G2OBatchStatistics* globalStats = G2OBatchStatistics::globalStats();
    if (globalStats) {
      globalStats->timeNumericDecomposition = get_monotonic_time() - t;
      globalStats->choleskyNNZ =
          _cholesky.matrixL().nestedExpression().nonZeros();
    }
    return true;
  }
 
 protected:
  bool _init;
  SparseMatrix _sparseMatrix;
  CholeskyDecomposition _cholesky;
 
  // compute the cholesky factor
  bool computeCholesky(const SparseBlockMatrix<MatrixType>& A, double& t) {
    // perform some operations only once upon init
    if (_init) _sparseMatrix.resize(A.rows(), A.cols());
    fillSparseMatrix(A, !_init);
    if (_init) computeSymbolicDecomposition(A);
    _init = false;
 
    t = get_monotonic_time();
    _cholesky.factorize(_sparseMatrix);
    if (_cholesky.info() !=
        Eigen::Success) {  // the matrix is not positive definite
      if (this->writeDebug()) {
        G2O_ERROR(
            "Cholesky failure, writing debug.txt (Hessian loadable by Octave)");
        A.writeOctave("debug.txt");
      } else {
        G2O_DEBUG("Cholesky failure");
      }
      return false;
    }
    return true;
  }
 
  void computeSymbolicDecomposition(const SparseBlockMatrix<MatrixType>& A) {
    double t = get_monotonic_time();
    if (!this->blockOrdering()) {
      _cholesky.analyzePattern(_sparseMatrix);
    } else {
      assert(A.rows() == A.cols() && "Matrix A is not square");
      // block ordering with the Eigen Interface
      Eigen::PermutationMatrix<Eigen::Dynamic, Eigen::Dynamic> blockP;
      {
        SparseMatrix auxBlockMatrix(A.blockCols().size(), A.blockCols().size());
        auxBlockMatrix.resizeNonZeros(A.nonZeroBlocks());
        // fill the CCS structure of the Eigen SparseMatrix
        A.fillBlockStructure(auxBlockMatrix.outerIndexPtr(),
                             auxBlockMatrix.innerIndexPtr());
        // determine ordering by AMD
        using Ordering = Eigen::AMDOrdering<SparseMatrix::StorageIndex>;
        Ordering ordering;
        ordering(auxBlockMatrix, blockP);
      }
 
      // Adapt the block permutation to the scalar matrix
      PermutationMatrix scalarP(A.rows());
      this->blockToScalarPermutation(A, blockP.indices(), scalarP.indices());
      // analyze with the scalar permutation
      _cholesky.analyzePatternWithPermutation(_sparseMatrix, scalarP);
    }
    G2OBatchStatistics* globalStats = G2OBatchStatistics::globalStats();
    if (globalStats)
      globalStats->timeSymbolicDecomposition = get_monotonic_time() - t;
  }
 
  void fillSparseMatrix(const SparseBlockMatrix<MatrixType>& A,
                        bool onlyValues) {
    if (onlyValues) {
      this->_ccsMatrix->fillCCS(_sparseMatrix.valuePtr(), true);
      return;
    }
    this->initMatrixStructure(A);
    _sparseMatrix.resizeNonZeros(A.nonZeros());
    int nz = this->_ccsMatrix->fillCCS(_sparseMatrix.outerIndexPtr(),
                                       _sparseMatrix.innerIndexPtr(),
                                       _sparseMatrix.valuePtr(), true);
    (void)nz;
    assert(nz <= static_cast<int>(_sparseMatrix.data().size()));
  }
 
  bool solveBlocks_impl(
      const SparseBlockMatrix<MatrixType>& A,
      std::function<void(MarginalCovarianceCholesky&)> compute) {
    // compute the cholesky factor
    double t;
    bool cholState = computeCholesky(A, t);
    if (!cholState) return false;
    // compute the inverse blocks
    MarginalCovarianceCholesky mcc;
    mcc.setCholeskyFactor(
        _cholesky.matrixL().rows(),
        const_cast<int*>(
            _cholesky.matrixL().nestedExpression().outerIndexPtr()),
        const_cast<int*>(
            _cholesky.matrixL().nestedExpression().innerIndexPtr()),
        const_cast<double*>(_cholesky.matrixL().nestedExpression().valuePtr()),
        const_cast<int*>(_cholesky.permutationP().indices().data()));
    compute(mcc);  // call the desired computation on the mcc object
    // book keeping statistics
    G2OBatchStatistics* globalStats = G2OBatchStatistics::globalStats();
    if (globalStats) {
      globalStats->choleskyNNZ =
          _cholesky.matrixL().nestedExpression().nonZeros();
    }
    return true;
  }
};
 
}  // namespace g2o
 
#endif