g2o::SparseOptimizer * MapG2OReflector::g2oInit(){
// graph construction
typedef BlockSolver< BlockSolverTraits<-1, -1> > SlamBlockSolver;
typedef LinearSolverCSparse<SlamBlockSolver::PoseMatrixType> SlamLinearSolver;
SlamLinearSolver* linearSolver = new SlamLinearSolver();
linearSolver->setBlockOrdering(false);
SlamBlockSolver* blockSolver = new SlamBlockSolver(linearSolver);
OptimizationAlgorithmLevenberg* solverGauss = new OptimizationAlgorithmLevenberg(blockSolver);
//OptimizationAlgorithmGaussNewton* solverGauss = new OptimizationAlgorithmGaussNewton(blockSolver);
SparseOptimizer * graph = new SparseOptimizer();
graph->setAlgorithm(solverGauss);
g2o::ParameterSE3Offset* imuOffset = new ParameterSE3Offset();
imuOffset->setOffset(Eigen::Isometry3d::Identity());
imuOffset->setId(0);
graph->addParameter(imuOffset);
return graph;
}
int main()
{
double euc_noise = 0.01; // noise in position, m
// double outlier_ratio = 0.1;
SparseOptimizer optimizer;
optimizer.setVerbose(false);
// variable-size block solver
BlockSolverX::LinearSolverType * linearSolver = new LinearSolverDense<g2o::BlockSolverX::PoseMatrixType>();
BlockSolverX * solver_ptr = new BlockSolverX(linearSolver);
g2o::OptimizationAlgorithmLevenberg* solver = new g2o::OptimizationAlgorithmLevenberg(solver_ptr);
optimizer.setAlgorithm(solver);
vector<Vector3d> true_points;
for (size_t i=0;i<1000; ++i)
{
true_points.push_back(Vector3d((Sample::uniform()-0.5)*3,
Sample::uniform()-0.5,
Sample::uniform()+10));
}
// set up two poses
int vertex_id = 0;
for (size_t i=0; i<2; ++i)
{
// set up rotation and translation for this node
Vector3d t(0,0,i);
Quaterniond q;
q.setIdentity();
Eigen::Isometry3d cam; // camera pose
cam = q;
cam.translation() = t;
// set up node
VertexSE3 *vc = new VertexSE3();
vc->setEstimate(cam);
vc->setId(vertex_id); // vertex id
cerr << t.transpose() << " | " << q.coeffs().transpose() << endl;
// set first cam pose fixed
if (i==0)
vc->setFixed(true);
// add to optimizer
optimizer.addVertex(vc);
vertex_id++;
}
// set up point matches
for (size_t i=0; i<true_points.size(); ++i)
{
// get two poses
VertexSE3* vp0 =
dynamic_cast<VertexSE3*>(optimizer.vertices().find(0)->second);
VertexSE3* vp1 =
dynamic_cast<VertexSE3*>(optimizer.vertices().find(1)->second);
// calculate the relative 3D position of the point
Vector3d pt0,pt1;
pt0 = vp0->estimate().inverse() * true_points[i];
pt1 = vp1->estimate().inverse() * true_points[i];
// add in noise
pt0 += Vector3d(Sample::gaussian(euc_noise ),
Sample::gaussian(euc_noise ),
Sample::gaussian(euc_noise ));
pt1 += Vector3d(Sample::gaussian(euc_noise ),
Sample::gaussian(euc_noise ),
Sample::gaussian(euc_noise ));
// form edge, with normals in varioius positions
Vector3d nm0, nm1;
nm0 << 0, i, 1;
nm1 << 0, i, 1;
nm0.normalize();
nm1.normalize();
Edge_V_V_GICP * e // new edge with correct cohort for caching
= new Edge_V_V_GICP();
e->setVertex(0, vp0); // first viewpoint
e->setVertex(1, vp1); // second viewpoint
EdgeGICP meas;
meas.pos0 = pt0;
meas.pos1 = pt1;
meas.normal0 = nm0;
meas.normal1 = nm1;
e->setMeasurement(meas);
// e->inverseMeasurement().pos() = -kp;
meas = e->measurement();
// use this for point-plane
e->information() = meas.prec0(0.01);
// use this for point-point
// e->information().setIdentity();
// e->setRobustKernel(true);
//e->setHuberWidth(0.01);
optimizer.addEdge(e);
}
// move second cam off of its true position
VertexSE3* vc =
dynamic_cast<VertexSE3*>(optimizer.vertices().find(1)->second);
Eigen::Isometry3d cam = vc->estimate();
cam.translation() = Vector3d(0,0,0.2);
vc->setEstimate(cam);
optimizer.initializeOptimization();
optimizer.computeActiveErrors();
cout << "Initial chi2 = " << FIXED(optimizer.chi2()) << endl;
optimizer.setVerbose(true);
optimizer.optimize(5);
cout << endl << "Second vertex should be near 0,0,1" << endl;
cout << dynamic_cast<VertexSE3*>(optimizer.vertices().find(0)->second)
->estimate().translation().transpose() << endl;
cout << dynamic_cast<VertexSE3*>(optimizer.vertices().find(1)->second)
->estimate().translation().transpose() << endl;
}
int main(int argc, char **argv) {
/************************************************************************
* Input handling *
************************************************************************/
float rows, cols, gain, square_size;
float resolution, max_range, usable_range, angle, threshold;
string g2oFilename, mapFilename;
g2o::CommandArgs arg;
arg.param("resolution", resolution, 0.05f, "resolution of the map (how much is in meters a pixel)");
arg.param("threshold", threshold, -1.0f, "threshold to apply to the frequency map (values under the threshold are discarded)");
arg.param("rows", rows, 0, "impose the resulting map to have this number of rows");
arg.param("cols", cols, 0, "impose the resulting map to have this number of columns");
arg.param("max_range", max_range, -1.0f, "max laser range to consider for map building");
arg.param("usable_range", usable_range, -1.0f, "usable laser range for map building");
arg.param("gain", gain, 1, "gain to impose to the pixels of the map");
arg.param("square_size", square_size, 1, "square size of the region where increment the hits");
arg.param("angle", angle, 0, "rotate the map of x degrees");
arg.paramLeftOver("input_graph.g2o", g2oFilename, "", "input g2o graph to use to build the map", false);
arg.paramLeftOver("output_map", mapFilename, "", "output filename where to save the map (without extension)", false);
arg.parseArgs(argc, argv);
angle = angle*M_PI/180.0;
/************************************************************************
* Loading Graph *
************************************************************************/
// Load graph
typedef BlockSolver< BlockSolverTraits<-1, -1> > SlamBlockSolver;
typedef LinearSolverCSparse<SlamBlockSolver::PoseMatrixType> SlamLinearSolver;
SlamLinearSolver *linearSolver = new SlamLinearSolver();
linearSolver->setBlockOrdering(false);
SlamBlockSolver *blockSolver = new SlamBlockSolver(linearSolver);
OptimizationAlgorithmGaussNewton *solverGauss = new OptimizationAlgorithmGaussNewton(blockSolver);
SparseOptimizer *graph = new SparseOptimizer();
graph->setAlgorithm(solverGauss);
graph->load(g2oFilename.c_str());
// Sort verteces
vector<int> vertexIds(graph->vertices().size());
int k = 0;
for(OptimizableGraph::VertexIDMap::iterator it = graph->vertices().begin(); it != graph->vertices().end(); ++it) {
vertexIds[k++] = (it->first);
}
sort(vertexIds.begin(), vertexIds.end());
/************************************************************************
* Compute map size *
************************************************************************/
// Check the entire graph to find map bounding box
Eigen::Matrix2d boundingBox = Eigen::Matrix2d::Zero();
std::vector<RobotLaser*> robotLasers;
std::vector<SE2> robotPoses;
double xmin=std::numeric_limits<double>::max();
double xmax=std::numeric_limits<double>::min();
double ymin=std::numeric_limits<double>::max();
double ymax=std::numeric_limits<double>::min();
SE2 baseTransform(0,0,angle);
for(size_t i = 0; i < vertexIds.size(); ++i) {
OptimizableGraph::Vertex *_v = graph->vertex(vertexIds[i]);
VertexSE2 *v = dynamic_cast<VertexSE2*>(_v);
if(!v) { continue; }
v->setEstimate(baseTransform*v->estimate());
OptimizableGraph::Data *d = v->userData();
while(d) {
RobotLaser *robotLaser = dynamic_cast<RobotLaser*>(d);
if(!robotLaser) {
d = d->next();
continue;
}
robotLasers.push_back(robotLaser);
robotPoses.push_back(v->estimate());
double x = v->estimate().translation().x();
double y = v->estimate().translation().y();
xmax = xmax > x+usable_range ? xmax : x+usable_range;
ymax = ymax > y+usable_range ? ymax : y+usable_range;
xmin = xmin < x-usable_range ? xmin : x-usable_range;
ymin = ymin < y-usable_range ? ymin : y-usable_range;
d = d->next();
}
}
boundingBox(0,0)=xmin;
boundingBox(0,1)=xmax;
boundingBox(1,0)=ymin;
boundingBox(1,1)=ymax;
std::cout << "Found " << robotLasers.size() << " laser scans"<< std::endl;
std::cout << "Bounding box: " << std::endl << boundingBox << std::endl;
if(robotLasers.size() == 0) {
std::cout << "No laser scans found ... quitting!" << std::endl;
return 0;
}
/************************************************************************
* Compute the map *
************************************************************************/
// Create the map
Eigen::Vector2i size;
if(rows != 0 && cols != 0) { size = Eigen::Vector2i(rows, cols); }
else {
size = Eigen::Vector2i((boundingBox(0, 1) - boundingBox(0, 0))/ resolution,
(boundingBox(1, 1) - boundingBox(1, 0))/ resolution);
}
std::cout << "Map size: " << size.transpose() << std::endl;
if(size.x() == 0 || size.y() == 0) {
std::cout << "Zero map size ... quitting!" << std::endl;
return 0;
}
//Eigen::Vector2f offset(-size.x() * resolution / 2.0f, -size.y() * resolution / 2.0f);
Eigen::Vector2f offset(boundingBox(0, 0),boundingBox(1, 0));
FrequencyMapCell unknownCell;
FrequencyMap map = FrequencyMap(resolution, offset, size, unknownCell);
for(size_t i = 0; i < vertexIds.size(); ++i) {
OptimizableGraph::Vertex *_v = graph->vertex(vertexIds[i]);
VertexSE2 *v = dynamic_cast<VertexSE2*>(_v);
if(!v) { continue; }
OptimizableGraph::Data *d = v->userData();
SE2 robotPose = v->estimate();
while(d) {
RobotLaser *robotLaser = dynamic_cast<RobotLaser*>(d);
if(!robotLaser) {
d = d->next();
continue;
}
map.integrateScan(robotLaser, robotPose, max_range, usable_range, gain, square_size);
d = d->next();
}
}
/************************************************************************
* Save map image *
************************************************************************/
cv::Mat mapImage(map.rows(), map.cols(), CV_8UC1);
mapImage.setTo(cv::Scalar(0));
for(int c = 0; c < map.cols(); c++) {
for(int r = 0; r < map.rows(); r++) {
if(map(r, c).misses() == 0 && map(r, c).hits() == 0) {
mapImage.at<unsigned char>(r, c) = 127;
} else {
float fraction = (float)map(r, c).hits()/(float)(map(r, c).hits()+map(r, c).misses());
if (threshold > 0 && fraction > threshold)
mapImage.at<unsigned char>(r, c) = 0;
else if (threshold > 0 && fraction <= threshold)
mapImage.at<unsigned char>(r, c) = 255;
else {
float val = 255*(1-fraction);
mapImage.at<unsigned char>(r, c) = (unsigned char)val;
}
}
// else if(map(r, c).hits() > threshold) {
// mapImage.at<unsigned char>(r, c) = 255;
// }
// else {
// mapImage.at<unsigned char>(r, c) = 0;
// }
}
}
cv::imwrite(mapFilename + ".png", mapImage);
/************************************************************************
* Write yaml file *
************************************************************************/
std::ofstream ofs(string(mapFilename + ".yaml").c_str());
Eigen::Vector3f origin(0.0f, 0.0f, 0.0f);
ofs << "image: " << mapFilename << ".png" << std::endl
<< "resolution: " << resolution << std::endl
<< "origin: [" << origin.x() << ", " << origin.y() << ", " << origin.z() << "]" << std::endl
<< "negate: 0" << std::endl
<< "occupied_thresh: " << 0.65f << std::endl
<< "free_thresh: " << 0.2f << std::endl;
return 0;
}
int main(int argc, char** argv)
{
OptimizableGraph::initMultiThreading();
int maxIterations;
bool verbose;
string inputFilename;
string gnudump;
string outputfilename;
string solverProperties;
string strSolver;
string loadLookup;
bool initialGuess;
bool initialGuessOdometry;
bool marginalize;
bool listTypes;
bool listSolvers;
bool listRobustKernels;
bool incremental;
bool guiOut;
int gaugeId;
string robustKernel;
bool computeMarginals;
bool printSolverProperties;
double huberWidth;
double gain;
int maxIterationsWithGain;
//double lambdaInit;
int updateGraphEachN = 10;
string statsFile;
string summaryFile;
bool nonSequential;
// command line parsing
std::vector<int> gaugeList;
CommandArgs arg;
arg.param("i", maxIterations, 5, "perform n iterations, if negative consider the gain");
arg.param("gain", gain, 1e-6, "the gain used to stop optimization (default = 1e-6)");
arg.param("ig",maxIterationsWithGain, std::numeric_limits<int>::max(), "Maximum number of iterations with gain enabled (default: inf)");
arg.param("v", verbose, false, "verbose output of the optimization process");
arg.param("guess", initialGuess, false, "initial guess based on spanning tree");
arg.param("guessOdometry", initialGuessOdometry, false, "initial guess based on odometry");
arg.param("inc", incremental, false, "run incremetally");
arg.param("update", updateGraphEachN, 10, "updates after x odometry nodes");
arg.param("guiout", guiOut, false, "gui output while running incrementally");
arg.param("marginalize", marginalize, false, "on or off");
arg.param("printSolverProperties", printSolverProperties, false, "print the properties of the solver");
arg.param("solverProperties", solverProperties, "", "set the internal properties of a solver,\n\te.g., initialLambda=0.0001,maxTrialsAfterFailure=2");
arg.param("gnudump", gnudump, "", "dump to gnuplot data file");
arg.param("robustKernel", robustKernel, "", "use this robust error function");
arg.param("robustKernelWidth", huberWidth, -1., "width for the robust Kernel (only if robustKernel)");
arg.param("computeMarginals", computeMarginals, false, "computes the marginal covariances of something. FOR TESTING ONLY");
arg.param("gaugeId", gaugeId, -1, "force the gauge");
arg.param("o", outputfilename, "", "output final version of the graph");
arg.param("solver", strSolver, "gn_var", "specify which solver to use underneat\n\t {gn_var, lm_fix3_2, gn_fix6_3, lm_fix7_3}");
#ifndef G2O_DISABLE_DYNAMIC_LOADING_OF_LIBRARIES
string dummy;
arg.param("solverlib", dummy, "", "specify a solver library which will be loaded");
arg.param("typeslib", dummy, "", "specify a types library which will be loaded");
#endif
arg.param("stats", statsFile, "", "specify a file for the statistics");
arg.param("listTypes", listTypes, false, "list the registered types");
arg.param("listRobustKernels", listRobustKernels, false, "list the registered robust kernels");
arg.param("listSolvers", listSolvers, false, "list the available solvers");
arg.param("renameTypes", loadLookup, "", "create a lookup for loading types into other types,\n\t TAG_IN_FILE=INTERNAL_TAG_FOR_TYPE,TAG2=INTERNAL2\n\t e.g., VERTEX_CAM=VERTEX_SE3:EXPMAP");
arg.param("gaugeList", gaugeList, std::vector<int>(), "set the list of gauges separated by commas without spaces \n e.g: 1,2,3,4,5 ");
arg.param("summary", summaryFile, "", "append a summary of this optimization run to the summary file passed as argument");
arg.paramLeftOver("graph-input", inputFilename, "", "graph file which will be processed", true);
arg.param("nonSequential", nonSequential, false, "apply the robust kernel only on loop closures and not odometries");
arg.parseArgs(argc, argv);
if (verbose) {
cout << "# Used Compiler: " << G2O_CXX_COMPILER << endl;
}
#ifndef G2O_DISABLE_DYNAMIC_LOADING_OF_LIBRARIES
// registering all the types from the libraries
DlWrapper dlTypesWrapper;
loadStandardTypes(dlTypesWrapper, argc, argv);
// register all the solvers
DlWrapper dlSolverWrapper;
loadStandardSolver(dlSolverWrapper, argc, argv);
#else
if (verbose)
cout << "# linked version of g2o" << endl;
#endif
OptimizationAlgorithmFactory* solverFactory = OptimizationAlgorithmFactory::instance();
if (listSolvers) {
solverFactory->listSolvers(cout);
}
if (listTypes) {
Factory::instance()->printRegisteredTypes(cout, true);
}
if (listRobustKernels) {
std::vector<std::string> kernels;
RobustKernelFactory::instance()->fillKnownKernels(kernels);
cout << "Robust Kernels:" << endl;
for (size_t i = 0; i < kernels.size(); ++i) {
cout << kernels[i] << endl;
}
}
SparseOptimizer optimizer;
optimizer.setVerbose(verbose);
optimizer.setForceStopFlag(&hasToStop);
SparseOptimizerTerminateAction* terminateAction = 0;
if (maxIterations < 0) {
cerr << "# setup termination criterion based on the gain of the iteration" << endl;
maxIterations = maxIterationsWithGain;
terminateAction = new SparseOptimizerTerminateAction;
terminateAction->setGainThreshold(gain);
terminateAction->setMaxIterations(maxIterationsWithGain);
optimizer.addPostIterationAction(terminateAction);
}
// allocating the desired solver + testing whether the solver is okay
OptimizationAlgorithmProperty solverProperty;
optimizer.setAlgorithm(solverFactory->construct(strSolver, solverProperty));
if (! optimizer.solver()) {
cerr << "Error allocating solver. Allocating \"" << strSolver << "\" failed!" << endl;
return 0;
}
if (solverProperties.size() > 0) {
bool updateStatus = optimizer.solver()->updatePropertiesFromString(solverProperties);
if (! updateStatus) {
cerr << "Failure while updating the solver properties from the given string" << endl;
}
}
if (solverProperties.size() > 0 || printSolverProperties) {
optimizer.solver()->printProperties(cerr);
}
// Loading the input data
if (loadLookup.size() > 0) {
optimizer.setRenamedTypesFromString(loadLookup);
}
if (inputFilename.size() == 0) {
cerr << "No input data specified" << endl;
return 0;
} else if (inputFilename == "-") {
cerr << "Read input from stdin" << endl;
if (!optimizer.load(cin)) {
cerr << "Error loading graph" << endl;
return 2;
}
} else {
cerr << "Read input from " << inputFilename << endl;
ifstream ifs(inputFilename.c_str());
if (!ifs) {
cerr << "Failed to open file" << endl;
return 1;
}
if (!optimizer.load(ifs)) {
cerr << "Error loading graph" << endl;
return 2;
}
}
cerr << "Loaded " << optimizer.vertices().size() << " vertices" << endl;
cerr << "Loaded " << optimizer.edges().size() << " edges" << endl;
if (optimizer.vertices().size() == 0) {
cerr << "Graph contains no vertices" << endl;
return 1;
}
set<int> vertexDimensions = optimizer.dimensions();
if (! optimizer.isSolverSuitable(solverProperty, vertexDimensions)) {
cerr << "The selected solver is not suitable for optimizing the given graph" << endl;
return 3;
}
assert (optimizer.solver());
//optimizer.setMethod(str2method(strMethod));
//optimizer.setUserLambdaInit(lambdaInit);
// check for vertices to fix to remove DoF
bool gaugeFreedom = optimizer.gaugeFreedom();
OptimizableGraph::Vertex* gauge=0;
if (gaugeList.size()){
cerr << "Fixing gauges: ";
for (size_t i=0; i<gaugeList.size(); i++){
int id=gaugeList[i];
OptimizableGraph::Vertex* v=optimizer.vertex(id);
if (!v){
cerr << "fatal, not found the vertex of id " << id << " in the gaugeList. Aborting";
return -1;
} else {
if (i==0)
gauge = v;
cerr << v->id() << " ";
v->setFixed(1);
}
}
cerr << endl;
gaugeFreedom = false;
} else {
gauge=optimizer.findGauge();
}
if (gaugeFreedom) {
if (! gauge) {
cerr << "# cannot find a vertex to fix in this thing" << endl;
return 2;
} else {
cerr << "# graph is fixed by node " << gauge->id() << endl;
gauge->setFixed(true);
}
} else {
cerr << "# graph is fixed by priors or already fixed vertex" << endl;
}
// if schur, we wanna marginalize the landmarks...
if (marginalize || solverProperty.requiresMarginalize) {
int maxDim = *vertexDimensions.rbegin();
int minDim = *vertexDimensions.begin();
if (maxDim != minDim) {
cerr << "# Preparing Marginalization of the Landmarks ... ";
for (HyperGraph::VertexIDMap::iterator it=optimizer.vertices().begin(); it!=optimizer.vertices().end(); it++){
OptimizableGraph::Vertex* v=static_cast<OptimizableGraph::Vertex*>(it->second);
if (v->dimension() != maxDim) {
v->setMarginalized(true);
}
}
cerr << "done." << endl;
}
}
if (robustKernel.size() > 0) {
AbstractRobustKernelCreator* creator = RobustKernelFactory::instance()->creator(robustKernel);
cerr << "# Preparing robust error function ... ";
if (creator) {
if (nonSequential) {
for (SparseOptimizer::EdgeSet::iterator it = optimizer.edges().begin(); it != optimizer.edges().end(); ++it) {
SparseOptimizer::Edge* e = dynamic_cast<SparseOptimizer::Edge*>(*it);
if (e->vertices().size() >= 2 && std::abs(e->vertex(0)->id() - e->vertex(1)->id()) != 1) {
e->setRobustKernel(creator->construct());
if (huberWidth > 0)
e->robustKernel()->setDelta(huberWidth);
}
}
} else {
for (SparseOptimizer::EdgeSet::iterator it = optimizer.edges().begin(); it != optimizer.edges().end(); ++it) {
SparseOptimizer::Edge* e = dynamic_cast<SparseOptimizer::Edge*>(*it);
e->setRobustKernel(creator->construct());
if (huberWidth > 0)
e->robustKernel()->setDelta(huberWidth);
}
}
cerr << "done." << endl;
} else {
cerr << "Unknown Robust Kernel: " << robustKernel << endl;
}
}
// sanity check
HyperDijkstra d(&optimizer);
UniformCostFunction f;
d.shortestPaths(gauge,&f);
//cerr << PVAR(d.visited().size()) << endl;
if (d.visited().size()!=optimizer.vertices().size()) {
cerr << CL_RED("Warning: d.visited().size() != optimizer.vertices().size()") << endl;
cerr << "visited: " << d.visited().size() << endl;
cerr << "vertices: " << optimizer.vertices().size() << endl;
}
if (incremental) {
cerr << CL_RED("# Note: this variant performs batch steps in each time step") << endl;
cerr << CL_RED("# For a variant which updates the Cholesky factor use the binary g2o_incremental") << endl;
int incIterations = maxIterations;
if (! arg.parsedParam("i")) {
cerr << "# Setting default number of iterations" << endl;
incIterations = 1;
}
int updateDisplayEveryN = updateGraphEachN;
int maxDim = 0;
cerr << "# incremental settings" << endl;
cerr << "#\t solve every " << updateGraphEachN << endl;
cerr << "#\t iterations " << incIterations << endl;
SparseOptimizer::VertexIDMap vertices = optimizer.vertices();
for (SparseOptimizer::VertexIDMap::const_iterator it = vertices.begin(); it != vertices.end(); ++it) {
const SparseOptimizer::Vertex* v = static_cast<const SparseOptimizer::Vertex*>(it->second);
maxDim = max(maxDim, v->dimension());
}
vector<SparseOptimizer::Edge*> edges;
for (SparseOptimizer::EdgeSet::iterator it = optimizer.edges().begin(); it != optimizer.edges().end(); ++it) {
SparseOptimizer::Edge* e = dynamic_cast<SparseOptimizer::Edge*>(*it);
edges.push_back(e);
}
optimizer.edges().clear();
optimizer.vertices().clear();
optimizer.setVerbose(false);
// sort the edges in a way that inserting them makes sense
sort(edges.begin(), edges.end(), IncrementalEdgesCompare());
double cumTime = 0.;
int vertexCount=0;
int lastOptimizedVertexCount = 0;
int lastVisUpdateVertexCount = 0;
bool freshlyOptimized=false;
bool firstRound = true;
HyperGraph::VertexSet verticesAdded;
HyperGraph::EdgeSet edgesAdded;
for (vector<SparseOptimizer::Edge*>::iterator it = edges.begin(); it != edges.end(); ++it) {
SparseOptimizer::Edge* e = *it;
int doInit = 0;
SparseOptimizer::Vertex* v1 = optimizer.vertex(e->vertices()[0]->id());
SparseOptimizer::Vertex* v2 = optimizer.vertex(e->vertices()[1]->id());
if (! v1) {
SparseOptimizer::Vertex* v = v1 = dynamic_cast<SparseOptimizer::Vertex*>(e->vertices()[0]);
bool v1Added = optimizer.addVertex(v);
//cerr << "adding" << v->id() << "(" << v->dimension() << ")" << endl;
assert(v1Added);
if (! v1Added)
cerr << "Error adding vertex " << v->id() << endl;
else
verticesAdded.insert(v);
doInit = 1;
if (v->dimension() == maxDim)
vertexCount++;
}
if (! v2) {
SparseOptimizer::Vertex* v = v2 = dynamic_cast<SparseOptimizer::Vertex*>(e->vertices()[1]);
bool v2Added = optimizer.addVertex(v);
//cerr << "adding" << v->id() << "(" << v->dimension() << ")" << endl;
assert(v2Added);
if (! v2Added)
cerr << "Error adding vertex " << v->id() << endl;
else
verticesAdded.insert(v);
doInit = 2;
if (v->dimension() == maxDim)
vertexCount++;
}
// adding the edge and initialization of the vertices
{
//cerr << " adding edge " << e->vertices()[0]->id() << " " << e->vertices()[1]->id() << endl;
if (! optimizer.addEdge(e)) {
cerr << "Unable to add edge " << e->vertices()[0]->id() << " -> " << e->vertices()[1]->id() << endl;
} else {
edgesAdded.insert(e);
}
if (doInit) {
OptimizableGraph::Vertex* from = static_cast<OptimizableGraph::Vertex*>(e->vertices()[0]);
OptimizableGraph::Vertex* to = static_cast<OptimizableGraph::Vertex*>(e->vertices()[1]);
switch (doInit){
case 1: // initialize v1 from v2
{
HyperGraph::VertexSet toSet;
toSet.insert(to);
if (e->initialEstimatePossible(toSet, from) > 0.) {
//cerr << "init: "
//<< to->id() << "(" << to->dimension() << ") -> "
//<< from->id() << "(" << from->dimension() << ") " << endl;
e->initialEstimate(toSet, from);
} else {
assert(0 && "Added unitialized variable to the graph");
}
break;
}
case 2:
{
HyperGraph::VertexSet fromSet;
fromSet.insert(from);
if (e->initialEstimatePossible(fromSet, to) > 0.) {
//cerr << "init: "
//<< from->id() << "(" << from->dimension() << ") -> "
//<< to->id() << "(" << to->dimension() << ") " << endl;
e->initialEstimate(fromSet, to);
} else {
assert(0 && "Added unitialized variable to the graph");
}
break;
}
default: cerr << "doInit wrong value\n";
}
}
}
freshlyOptimized=false;
{
//cerr << "Optimize" << endl;
if (vertexCount - lastOptimizedVertexCount >= updateGraphEachN) {
if (firstRound) {
if (!optimizer.initializeOptimization()){
cerr << "initialization failed" << endl;
return 0;
}
} else {
if (! optimizer.updateInitialization(verticesAdded, edgesAdded)) {
cerr << "updating initialization failed" << endl;
return 0;
}
}
verticesAdded.clear();
edgesAdded.clear();
double ts = get_monotonic_time();
int currentIt=optimizer.optimize(incIterations, !firstRound);
double dts = get_monotonic_time() - ts;
cumTime += dts;
firstRound = false;
//optimizer->setOptimizationTime(cumTime);
if (verbose) {
double chi2 = optimizer.chi2();
cerr << "nodes= " << optimizer.vertices().size() << "\t edges= " << optimizer.edges().size() << "\t chi2= " << chi2
<< "\t time= " << dts << "\t iterations= " << currentIt << "\t cumTime= " << cumTime << endl;
}
lastOptimizedVertexCount = vertexCount;
freshlyOptimized = true;
if (guiOut) {
if (vertexCount - lastVisUpdateVertexCount >= updateDisplayEveryN) {
dumpEdges(cout, optimizer);
lastVisUpdateVertexCount = vertexCount;
}
}
}
if (! verbose)
cerr << ".";
}
} // for all edges
if (! freshlyOptimized) {
double ts = get_monotonic_time();
int currentIt=optimizer.optimize(incIterations, !firstRound);
double dts = get_monotonic_time() - ts;
cumTime += dts;
//optimizer->setOptimizationTime(cumTime);
if (verbose) {
double chi2 = optimizer.chi2();
cerr << "nodes= " << optimizer.vertices().size() << "\t edges= " << optimizer.edges().size() << "\t chi2= " << chi2
<< "\t time= " << dts << "\t iterations= " << currentIt << "\t cumTime= " << cumTime << endl;
}
}
} else {
// BATCH optimization
if (statsFile!=""){
// allocate buffer for statistics;
optimizer.setComputeBatchStatistics(true);
}
optimizer.initializeOptimization();
optimizer.computeActiveErrors();
double loadChi = optimizer.chi2();
cerr << "Initial chi2 = " << FIXED(loadChi) << endl;
if (initialGuess) {
optimizer.computeInitialGuess();
} else if (initialGuessOdometry) {
EstimatePropagatorCostOdometry costFunction(&optimizer);
optimizer.computeInitialGuess(costFunction);
}
double initChi = optimizer.chi2();
signal(SIGINT, sigquit_handler);
int result=optimizer.optimize(maxIterations);
if (maxIterations > 0 && result==OptimizationAlgorithm::Fail){
cerr << "Cholesky failed, result might be invalid" << endl;
} else if (computeMarginals){
std::vector<std::pair<int, int> > blockIndices;
for (size_t i=0; i<optimizer.activeVertices().size(); i++){
OptimizableGraph::Vertex* v=optimizer.activeVertices()[i];
if (v->hessianIndex()>=0){
blockIndices.push_back(make_pair(v->hessianIndex(), v->hessianIndex()));
}
if (v->hessianIndex()>0){
blockIndices.push_back(make_pair(v->hessianIndex()-1, v->hessianIndex()));
}
}
SparseBlockMatrix<MatrixXd> spinv;
if (optimizer.computeMarginals(spinv, blockIndices)) {
for (size_t i=0; i<optimizer.activeVertices().size(); i++){
OptimizableGraph::Vertex* v=optimizer.activeVertices()[i];
cerr << "Vertex id:" << v->id() << endl;
if (v->hessianIndex()>=0){
cerr << "inv block :" << v->hessianIndex() << ", " << v->hessianIndex()<< endl;
cerr << *(spinv.block(v->hessianIndex(), v->hessianIndex()));
cerr << endl;
}
if (v->hessianIndex()>0){
cerr << "inv block :" << v->hessianIndex()-1 << ", " << v->hessianIndex()<< endl;
cerr << *(spinv.block(v->hessianIndex()-1, v->hessianIndex()));
cerr << endl;
}
}
}
}
optimizer.computeActiveErrors();
double finalChi=optimizer.chi2();
if (summaryFile!="") {
PropertyMap summary;
summary.makeProperty<StringProperty>("filename", inputFilename);
summary.makeProperty<IntProperty>("n_vertices", optimizer.vertices().size());
summary.makeProperty<IntProperty>("n_edges", optimizer.edges().size());
int nLandmarks=0;
int nPoses=0;
int maxDim = *vertexDimensions.rbegin();
for (HyperGraph::VertexIDMap::iterator it=optimizer.vertices().begin(); it!=optimizer.vertices().end(); it++){
OptimizableGraph::Vertex* v=static_cast<OptimizableGraph::Vertex*>(it->second);
if (v->dimension() != maxDim) {
nLandmarks++;
} else
nPoses++;
}
set<string> edgeTypes;
for (HyperGraph::EdgeSet::iterator it=optimizer.edges().begin(); it!=optimizer.edges().end(); it++){
edgeTypes.insert(Factory::instance()->tag(*it));
}
stringstream edgeTypesString;
for (std::set<string>::iterator it=edgeTypes.begin(); it!=edgeTypes.end(); it++){
edgeTypesString << *it << " ";
}
summary.makeProperty<IntProperty>("n_poses", nPoses);
summary.makeProperty<IntProperty>("n_landmarks", nLandmarks);
summary.makeProperty<StringProperty>("edge_types", edgeTypesString.str());
summary.makeProperty<DoubleProperty>("load_chi", loadChi);
summary.makeProperty<StringProperty>("solver", strSolver);
summary.makeProperty<BoolProperty>("robustKernel", robustKernel.size() > 0);
summary.makeProperty<DoubleProperty>("init_chi", initChi);
summary.makeProperty<DoubleProperty>("final_chi", finalChi);
summary.makeProperty<IntProperty>("maxIterations", maxIterations);
summary.makeProperty<IntProperty>("realIterations", result);
ofstream os;
os.open(summaryFile.c_str(), ios::app);
summary.writeToCSV(os);
}
if (statsFile!=""){
cerr << "writing stats to file \"" << statsFile << "\" ... ";
ofstream os(statsFile.c_str());
const BatchStatisticsContainer& bsc = optimizer.batchStatistics();
for (int i=0; i<maxIterations; i++) {
os << bsc[i] << endl;
}
cerr << "done." << endl;
}
}
// saving again
if (gnudump.size() > 0) {
bool gnuPlotStatus = saveGnuplot(gnudump, optimizer);
if (! gnuPlotStatus) {
cerr << "Error while writing gnuplot files" << endl;
}
}
if (outputfilename.size() > 0) {
if (outputfilename == "-") {
cerr << "saving to stdout";
optimizer.save(cout);
} else {
cerr << "saving " << outputfilename << " ... ";
optimizer.save(outputfilename.c_str());
}
cerr << "done." << endl;
}
// destroy all the singletons
//Factory::destroy();
//OptimizationAlgorithmFactory::destroy();
//HyperGraphActionLibrary::destroy();
return 0;
}
int main(int argc, char **argv)
{
int num_points = 0;
// check for arg, # of points to use in projection SBA
if (argc > 1)
num_points = atoi(argv[1]);
double euc_noise = 0.1; // noise in position, m
double pix_noise = 1.0; // pixel noise
// double outlier_ratio = 0.1;
SparseOptimizer optimizer;
optimizer.setVerbose(false);
// variable-size block solver
BlockSolverX::LinearSolverType * linearSolver
= new LinearSolverCSparse<g2o
::BlockSolverX::PoseMatrixType>();
BlockSolverX * solver_ptr
= new BlockSolverX(linearSolver);
g2o::OptimizationAlgorithmLevenberg* solver = new g2o::OptimizationAlgorithmLevenberg(solver_ptr);
optimizer.setAlgorithm(solver);
vector<Vector3d> true_points;
for (size_t i=0;i<1000; ++i)
{
true_points.push_back(Vector3d((Sample::uniform()-0.5)*3,
Sample::uniform()-0.5,
Sample::uniform()+10));
}
// set up camera params
Vector2d focal_length(500,500); // pixels
Vector2d principal_point(320,240); // 640x480 image
double baseline = 0.075; // 7.5 cm baseline
// set up camera params and projection matrices on vertices
g2o::VertexSCam::setKcam(focal_length[0],focal_length[1],
principal_point[0],principal_point[1],
baseline);
// set up two poses
int vertex_id = 0;
for (size_t i=0; i<2; ++i)
{
// set up rotation and translation for this node
Vector3d t(0,0,i);
Quaterniond q;
q.setIdentity();
Eigen::Isometry3d cam; // camera pose
cam = q;
cam.translation() = t;
// set up node
VertexSCam *vc = new VertexSCam();
vc->setEstimate(cam);
vc->setId(vertex_id); // vertex id
cerr << t.transpose() << " | " << q.coeffs().transpose() << endl;
// set first cam pose fixed
if (i==0)
vc->setFixed(true);
// make sure projection matrices are set
vc->setAll();
// add to optimizer
optimizer.addVertex(vc);
vertex_id++;
}
// set up point matches for GICP
for (size_t i=0; i<true_points.size(); ++i)
{
// get two poses
VertexSE3* vp0 =
dynamic_cast<VertexSE3*>(optimizer.vertices().find(0)->second);
VertexSE3* vp1 =
dynamic_cast<VertexSE3*>(optimizer.vertices().find(1)->second);
// calculate the relative 3D position of the point
Vector3d pt0,pt1;
pt0 = vp0->estimate().inverse() * true_points[i];
pt1 = vp1->estimate().inverse() * true_points[i];
// add in noise
pt0 += Vector3d(Sample::gaussian(euc_noise ),
Sample::gaussian(euc_noise ),
Sample::gaussian(euc_noise ));
pt1 += Vector3d(Sample::gaussian(euc_noise ),
Sample::gaussian(euc_noise ),
Sample::gaussian(euc_noise ));
// form edge, with normals in varioius positions
Vector3d nm0, nm1;
nm0 << 0, i, 1;
nm1 << 0, i, 1;
nm0.normalize();
nm1.normalize();
Edge_V_V_GICP * e // new edge with correct cohort for caching
= new Edge_V_V_GICP();
e->vertices()[0] // first viewpoint
= dynamic_cast<OptimizableGraph::Vertex*>(vp0);
e->vertices()[1] // second viewpoint
= dynamic_cast<OptimizableGraph::Vertex*>(vp1);
EdgeGICP meas;
meas.pos0 = pt0;
meas.pos1 = pt1;
meas.normal0 = nm0;
meas.normal1 = nm1;
e->setMeasurement(meas);
meas = e->measurement();
// e->inverseMeasurement().pos() = -kp;
// use this for point-plane
e->information() = meas.prec0(0.01);
// use this for point-point
// e->information().setIdentity();
// e->setRobustKernel(true);
//e->setHuberWidth(0.01);
optimizer.addEdge(e);
}
// set up SBA projections with some number of points
true_points.clear();
for (int i=0;i<num_points; ++i)
{
true_points.push_back(Vector3d((Sample::uniform()-0.5)*3,
Sample::uniform()-0.5,
Sample::uniform()+10));
}
// add point projections to this vertex
for (size_t i=0; i<true_points.size(); ++i)
{
g2o::VertexSBAPointXYZ * v_p
= new g2o::VertexSBAPointXYZ();
v_p->setId(vertex_id++);
v_p->setMarginalized(true);
v_p->setEstimate(true_points.at(i)
+ Vector3d(Sample::gaussian(1),
Sample::gaussian(1),
Sample::gaussian(1)));
optimizer.addVertex(v_p);
for (size_t j=0; j<2; ++j)
{
Vector3d z;
dynamic_cast<g2o::VertexSCam*>
(optimizer.vertices().find(j)->second)
->mapPoint(z,true_points.at(i));
if (z[0]>=0 && z[1]>=0 && z[0]<640 && z[1]<480)
{
z += Vector3d(Sample::gaussian(pix_noise),
Sample::gaussian(pix_noise),
Sample::gaussian(pix_noise/16.0));
g2o::Edge_XYZ_VSC * e
= new g2o::Edge_XYZ_VSC();
e->vertices()[0]
= dynamic_cast<g2o::OptimizableGraph::Vertex*>(v_p);
e->vertices()[1]
= dynamic_cast<g2o::OptimizableGraph::Vertex*>
(optimizer.vertices().find(j)->second);
e->setMeasurement(z);
//e->inverseMeasurement() = -z;
e->information() = Matrix3d::Identity();
//e->setRobustKernel(false);
//e->setHuberWidth(1);
optimizer.addEdge(e);
}
}
} // done with adding projection points
// move second cam off of its true position
VertexSE3* vc =
dynamic_cast<VertexSE3*>(optimizer.vertices().find(1)->second);
Eigen::Isometry3d cam = vc->estimate();
cam.translation() = Vector3d(-0.1,0.1,0.2);
vc->setEstimate(cam);
optimizer.initializeOptimization();
optimizer.computeActiveErrors();
cout << "Initial chi2 = " << FIXED(optimizer.chi2()) << endl;
optimizer.setVerbose(true);
optimizer.optimize(20);
cout << endl << "Second vertex should be near 0,0,1" << endl;
cout << dynamic_cast<VertexSE3*>(optimizer.vertices().find(0)->second)
->estimate().translation().transpose() << endl;
cout << dynamic_cast<VertexSE3*>(optimizer.vertices().find(1)->second)
->estimate().translation().transpose() << endl;
}
void Optimizer::optimizeUseG2O()
{
// create the linear solver
BlockSolverX::LinearSolverType * linearSolver = new LinearSolverCSparse<BlockSolverX::PoseMatrixType>();
// create the block solver on top of the linear solver
BlockSolverX* blockSolver = new BlockSolverX(linearSolver);
// create the algorithm to carry out the optimization
//OptimizationAlgorithmGaussNewton* optimizationAlgorithm = new OptimizationAlgorithmGaussNewton(blockSolver);
OptimizationAlgorithmLevenberg* optimizationAlgorithm = new OptimizationAlgorithmLevenberg(blockSolver);
// NOTE: We skip to fix a variable here, either this is stored in the file
// itself or Levenberg will handle it.
// create the optimizer to load the data and carry out the optimization
SparseOptimizer optimizer;
SparseOptimizer::initMultiThreading();
optimizer.setVerbose(true);
optimizer.setAlgorithm(optimizationAlgorithm);
{
pcl::ScopeTime time("G2O setup Graph vertices");
for (size_t cloud_count = 0; cloud_count < m_pointClouds.size(); ++cloud_count)
{
VertexSE3 *vertex = new VertexSE3;
vertex->setId(cloud_count);
Isometry3D affine = Isometry3D::Identity();
affine.linear() = m_pointClouds[cloud_count]->sensor_orientation_.toRotationMatrix().cast<Isometry3D::Scalar>();
affine.translation() = m_pointClouds[cloud_count]->sensor_origin_.block<3, 1>(0, 0).cast<Isometry3D::Scalar>();
vertex->setEstimate(affine);
optimizer.addVertex(vertex);
}
optimizer.vertex(0)->setFixed(true);
}
{
pcl::ScopeTime time("G2O setup Graph edges");
double trans_noise = 0.5, rot_noise = 0.5235;
EdgeSE3::InformationType infomation = EdgeSE3::InformationType::Zero();
infomation.block<3, 3>(0, 0) << trans_noise * trans_noise, 0, 0,
0, trans_noise * trans_noise, 0,
0, 0, trans_noise * trans_noise;
infomation.block<3, 3>(3, 3) << rot_noise * rot_noise, 0, 0,
0, rot_noise * rot_noise, 0,
0, 0, rot_noise * rot_noise;
for (size_t pair_count = 0; pair_count < m_cloudPairs.size(); ++pair_count)
{
CloudPair pair = m_cloudPairs[pair_count];
int from = pair.corresIdx.first;
int to = pair.corresIdx.second;
EdgeSE3 *edge = new EdgeSE3;
edge->vertices()[0] = optimizer.vertex(from);
edge->vertices()[1] = optimizer.vertex(to);
Eigen::Matrix<double, 6, 6> ATA = Eigen::Matrix<double, 6, 6>::Zero();
Eigen::Matrix<double, 6, 1> ATb = Eigen::Matrix<double, 6, 1>::Zero();
#pragma unroll 8
for (size_t point_count = 0; point_count < pair.corresPointIdx.size(); ++point_count) {
int point_p = pair.corresPointIdx[point_count].first;
int point_q = pair.corresPointIdx[point_count].second;
PointType P = m_pointClouds[from]->points[point_p];
PointType Q = m_pointClouds[to]->points[point_q];
Eigen::Vector3d p = P.getVector3fMap().cast<double>();
Eigen::Vector3d q = Q.getVector3fMap().cast<double>();
Eigen::Vector3d Np = P.getNormalVector3fMap().cast<double>();
double b = (p - q).dot(Np);
Eigen::Matrix<double, 6, 1> A_p;
A_p.block<3, 1>(0, 0) = p.cross(Np);
A_p.block<3, 1>(3, 0) = Np;
ATA += A_p * A_p.transpose();
ATb += A_p * b;
}
Eigen::Matrix<double, 6, 1> X = ATA.ldlt().solve(ATb);
Isometry3D measure = Isometry3D::Identity();
float beta = X[0];
float gammar = X[1];
float alpha = X[2];
measure.linear() = (Eigen::Matrix3d)Eigen::AngleAxisd(alpha, Eigen::Vector3d::UnitZ()) *
Eigen::AngleAxisd(gammar, Eigen::Vector3d::UnitY()) *
Eigen::AngleAxisd(beta, Eigen::Vector3d::UnitX());
measure.translation() = X.block<3, 1>(3, 0);
edge->setMeasurement(measure);
edge->setInformation(infomation);
optimizer.addEdge(edge);
}
}
optimizer.save("debug_preOpt.g2o");
{
pcl::ScopeTime time("g2o optimizing");
optimizer.initializeOptimization();
optimizer.optimize(30);
}
optimizer.save("debug_postOpt.g2o");
for (size_t cloud_count = 0; cloud_count < m_pointClouds.size(); ++cloud_count)
{
CloudTypePtr tmp(new CloudType);
Isometry3D trans = ((VertexSE3 *)optimizer.vertices()[cloud_count])->estimate();
Eigen::Affine3d affine;
affine.linear() = trans.rotation();
affine.translation() = trans.translation();
pcl::transformPointCloudWithNormals(*m_pointClouds[cloud_count], *tmp, affine.cast<float>());
pcl::copyPointCloud(*tmp, *m_pointClouds[cloud_count]);
}
PCL_WARN("Opitimization DONE!!!!\n");
if (m_params.saveDirectory.length()) {
if (boost::filesystem::exists(m_params.saveDirectory) && !boost::filesystem::is_directory(m_params.saveDirectory)) {
boost::filesystem::remove(m_params.saveDirectory);
}
boost::filesystem::create_directories(m_params.saveDirectory);
char filename[1024] = { 0 };
for (size_t i = 0; i < m_pointClouds.size(); ++i) {
sprintf(filename, "%s/cloud_%d.ply", m_params.saveDirectory.c_str(), i);
pcl::io::savePLYFileBinary(filename, *m_pointClouds[i]);
}
}
}