12345qiupeng
/
gtsam
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

correct formatting

release/4.3a0
lcarlone 2020-11-27 22:54:51 -05:00
parent c4644a0d61
commit 7699f04820
1 changed files with 155 additions and 120 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

125
tests/testGncOptimizer.cpp
View File

@ -51,10 +51,13 @@ public:
using BaseOptimizer = GaussNewtonOptimizer; // BaseOptimizerParameters::OptimizerType; using BaseOptimizer = GaussNewtonOptimizer; // BaseOptimizerParameters::OptimizerType;
GncParams(const BaseOptimizerParameters& baseOptimizerParams) : GncParams(const BaseOptimizerParameters& baseOptimizerParams) :
baseOptimizerParams(baseOptimizerParams) {} baseOptimizerParams(baseOptimizerParams) {
}
// default constructor // default constructor
GncParams(): baseOptimizerParams() {} GncParams() :
baseOptimizerParams() {
}
BaseOptimizerParameters baseOptimizerParams; BaseOptimizerParameters baseOptimizerParams;
/// any other specific GNC parameters: /// any other specific GNC parameters:
@ -65,16 +68,24 @@ public:
VerbosityGNC verbosityGNC = SILENT; /* verbosity level */ VerbosityGNC verbosityGNC = SILENT; /* verbosity level */
std::vector<size_t> knownInliers = std::vector<size_t>(); /* slots in the factor graph corresponding to measurements that we know are inliers */ std::vector<size_t> knownInliers = std::vector<size_t>(); /* slots in the factor graph corresponding to measurements that we know are inliers */
void setLossType(const RobustLossType type){ lossType = type; } void setLossType(const RobustLossType type) {
lossType = type;
}
void setMaxIterations(const size_t maxIter) { void setMaxIterations(const size_t maxIter) {
std::cout std::cout
<< "setMaxIterations: changing the max nr of iters might lead to less accurate solutions and is not recommended! " << "setMaxIterations: changing the max nr of iters might lead to less accurate solutions and is not recommended! "
<< std::endl; << std::endl;
maxIterations = maxIter; maxIterations = maxIter;
} }
void setInlierThreshold(const double inth){ barcSq = inth; } void setInlierThreshold(const double inth) {
void setMuStep(const double step){ muStep = step; } barcSq = inth;
void setVerbosityGNC(const VerbosityGNC verbosity) { verbosityGNC = verbosity; } }
void setMuStep(const double step) {
muStep = step;
}
void setVerbosityGNC(const VerbosityGNC verbosity) {
verbosityGNC = verbosity;
}
void setKnownInliers(const std::vector<size_t> knownIn) { void setKnownInliers(const std::vector<size_t> knownIn) {
for (size_t i = 0; i < knownIn.size(); i++) for (size_t i = 0; i < knownIn.size(); i++)
knownInliers.push_back(knownIn[i]); knownInliers.push_back(knownIn[i]);
@ -83,8 +94,7 @@ public:
/// equals /// equals
bool equals(const GncParams& other, double tol = 1e-9) const { bool equals(const GncParams& other, double tol = 1e-9) const {
return baseOptimizerParams.equals(other.baseOptimizerParams) return baseOptimizerParams.equals(other.baseOptimizerParams)
&& lossType == other.lossType && lossType == other.lossType && maxIterations == other.maxIterations
&& maxIterations == other.maxIterations
&& std::fabs(barcSq - other.barcSq) <= tol && std::fabs(barcSq - other.barcSq) <= tol
&& std::fabs(muStep - other.muStep) <= tol && std::fabs(muStep - other.muStep) <= tol
&& verbosityGNC == other.verbosityGNC && verbosityGNC == other.verbosityGNC
@ -95,10 +105,11 @@ public:
void print(const std::string& str) const { void print(const std::string& str) const {
std::cout << str << "\n"; std::cout << str << "\n";
switch (lossType) { switch (lossType) {
case GM: std::cout << "lossType: Geman McClure" << "\n"; break; case GM:
std::cout << "lossType: Geman McClure" << "\n";
break;
default: default:
throw std::runtime_error( throw std::runtime_error("GncParams::print: unknown loss type.");
"GncParams::print: unknown loss type.");
} }
std::cout << "maxIterations: " << maxIterations << "\n"; std::cout << "maxIterations: " << maxIterations << "\n";
std::cout << "barcSq: " << barcSq << "\n"; std::cout << "barcSq: " << barcSq << "\n";
@ -123,19 +134,22 @@ private:
Vector weights_; // this could be a local variable in optimize, but it is useful to make it accessible from outside Vector weights_; // this could be a local variable in optimize, but it is useful to make it accessible from outside
public: public:
GncOptimizer(const NonlinearFactorGraph& graph, GncOptimizer(const NonlinearFactorGraph& graph, const Values& initialValues,
const Values& initialValues, const GncParameters& params = GncParameters()) : const GncParameters& params = GncParameters()) :
state_(initialValues), params_(params) { state_(initialValues), params_(params) {
// make sure all noiseModels are Gaussian or convert to Gaussian // make sure all noiseModels are Gaussian or convert to Gaussian
nfg_.resize(graph.size()); nfg_.resize(graph.size());
for (size_t i = 0; i < graph.size(); i++) { for (size_t i = 0; i < graph.size(); i++) {
if (graph[i]) { if (graph[i]) {
NoiseModelFactor::shared_ptr factor = boost::dynamic_pointer_cast<NoiseModelFactor>(graph[i]); NoiseModelFactor::shared_ptr factor = boost::dynamic_pointer_cast<
noiseModel::Robust::shared_ptr robust = boost::dynamic_pointer_cast<noiseModel::Robust>(factor->noiseModel()); NoiseModelFactor>(graph[i]);
noiseModel::Robust::shared_ptr robust = boost::dynamic_pointer_cast<
noiseModel::Robust>(factor->noiseModel());
if (robust) { // if the factor has a robust loss, we have to change it: if (robust) { // if the factor has a robust loss, we have to change it:
SharedNoiseModel gaussianNoise = robust->noise(); SharedNoiseModel gaussianNoise = robust->noise();
NoiseModelFactor::shared_ptr gaussianFactor = factor->cloneWithNewNoiseModel(gaussianNoise); NoiseModelFactor::shared_ptr gaussianFactor =
factor->cloneWithNewNoiseModel(gaussianNoise);
nfg_[i] = gaussianFactor; nfg_[i] = gaussianFactor;
} else { // else we directly push it back } else { // else we directly push it back
nfg_[i] = factor; nfg_[i] = factor;
@ -145,10 +159,18 @@ public:
} }
/// getter functions /// getter functions
NonlinearFactorGraph getFactors() const { return NonlinearFactorGraph(nfg_); } NonlinearFactorGraph getFactors() const {
Values getState() const { return Values(state_); } return NonlinearFactorGraph(nfg_);
GncParameters getParams() const { return GncParameters(params_); } }
Vector getWeights() const {return weights_;} Values getState() const {
return Values(state_);
}
GncParameters getParams() const {
return GncParameters(params_);
}
Vector getWeights() const {
return weights_;
}
/// implement GNC main loop, including graduating nonconvexity with mu /// implement GNC main loop, including graduating nonconvexity with mu
Values optimize() { Values optimize() {
@ -238,11 +260,15 @@ public:
newGraph.resize(nfg_.size()); newGraph.resize(nfg_.size());
for (size_t i = 0; i < nfg_.size(); i++) { for (size_t i = 0; i < nfg_.size(); i++) {
if (nfg_[i]) { if (nfg_[i]) {
NoiseModelFactor::shared_ptr factor = boost::dynamic_pointer_cast<NoiseModelFactor>(nfg_[i]); NoiseModelFactor::shared_ptr factor = boost::dynamic_pointer_cast<
noiseModel::Gaussian::shared_ptr noiseModel = boost::dynamic_pointer_cast<noiseModel::Gaussian>(factor->noiseModel()); NoiseModelFactor>(nfg_[i]);
noiseModel::Gaussian::shared_ptr noiseModel =
boost::dynamic_pointer_cast<noiseModel::Gaussian>(
factor->noiseModel());
if (noiseModel) { if (noiseModel) {
Matrix newInfo = weights[i] * noiseModel->information(); Matrix newInfo = weights[i] * noiseModel->information();
SharedNoiseModel newNoiseModel = noiseModel::Gaussian::Information(newInfo); SharedNoiseModel newNoiseModel = noiseModel::Gaussian::Information(
newInfo);
newGraph[i] = factor->cloneWithNewNoiseModel(newNoiseModel); newGraph[i] = factor->cloneWithNewNoiseModel(newNoiseModel);
} else { } else {
throw std::runtime_error( throw std::runtime_error(
@ -259,7 +285,9 @@ public:
// do not update the weights that the user has decided are known inliers // do not update the weights that the user has decided are known inliers
std::vector<size_t> allWeights; std::vector<size_t> allWeights;
for (size_t k = 0; k < nfg_.size(); k++) {allWeights.push_back(k);} for (size_t k = 0; k < nfg_.size(); k++) {
allWeights.push_back(k);
}
std::vector<size_t> unknownWeights; std::vector<size_t> unknownWeights;
std::set_difference(allWeights.begin(), allWeights.end(), std::set_difference(allWeights.begin(), allWeights.end(),
params_.knownInliers.begin(), params_.knownInliers.end(), params_.knownInliers.begin(), params_.knownInliers.end(),
@ -271,7 +299,8 @@ public:
for (size_t k : unknownWeights) { for (size_t k : unknownWeights) {
if (nfg_[k]) { if (nfg_[k]) {
double u2_k = nfg_[k]->error(currentEstimate); // squared (and whitened) residual double u2_k = nfg_[k]->error(currentEstimate); // squared (and whitened) residual
weights[k] = std::pow( ( mu*params_.barcSq )/( u2_k + mu*params_.barcSq ) , 2); weights[k] = std::pow(
(mu * params_.barcSq) / (u2_k + mu * params_.barcSq), 2);
} }
} }
return weights; return weights;
@ -284,7 +313,6 @@ public:
/* ************************************************************************* */ /* ************************************************************************* */
TEST(GncOptimizer, gncParamsConstructor) { TEST(GncOptimizer, gncParamsConstructor) {
//check params are correctly parsed //check params are correctly parsed
LevenbergMarquardtParams lmParams; LevenbergMarquardtParams lmParams;
GncParams<LevenbergMarquardtParams> gncParams1(lmParams); GncParams<LevenbergMarquardtParams> gncParams1(lmParams);
@ -324,7 +352,8 @@ TEST(GncOptimizer, gncConstructor) {
LevenbergMarquardtParams lmParams; LevenbergMarquardtParams lmParams;
GncParams<LevenbergMarquardtParams> gncParams(lmParams); GncParams<LevenbergMarquardtParams> gncParams(lmParams);
auto gnc = GncOptimizer<GncParams<LevenbergMarquardtParams>>(fg, initial, gncParams); auto gnc = GncOptimizer<GncParams<LevenbergMarquardtParams>>(fg, initial,
gncParams);
CHECK(gnc.getFactors().equals(fg)); CHECK(gnc.getFactors().equals(fg));
CHECK(gnc.getState().equals(initial)); CHECK(gnc.getState().equals(initial));
@ -333,7 +362,6 @@ TEST(GncOptimizer, gncConstructor) {
/* ************************************************************************* */ /* ************************************************************************* */
TEST(GncOptimizer, gncConstructorWithRobustGraphAsInput) { TEST(GncOptimizer, gncConstructorWithRobustGraphAsInput) {
// simple graph with Gaussian noise model
auto fg = example::sharedNonRobustFactorGraphWithOutliers(); auto fg = example::sharedNonRobustFactorGraphWithOutliers();
// same graph with robust noise model // same graph with robust noise model
auto fg_robust = example::sharedRobustFactorGraphWithOutliers(); auto fg_robust = example::sharedRobustFactorGraphWithOutliers();
@ -344,7 +372,8 @@ TEST(GncOptimizer, gncConstructorWithRobustGraphAsInput) {
LevenbergMarquardtParams lmParams; LevenbergMarquardtParams lmParams;
GncParams<LevenbergMarquardtParams> gncParams(lmParams); GncParams<LevenbergMarquardtParams> gncParams(lmParams);
auto gnc = GncOptimizer<GncParams<LevenbergMarquardtParams>>(fg_robust, initial, gncParams); auto gnc = GncOptimizer<GncParams<LevenbergMarquardtParams>>(fg_robust,
initial, gncParams);
// make sure that when parsing the graph is transformed into one without robust loss // make sure that when parsing the graph is transformed into one without robust loss
CHECK(fg.equals(gnc.getFactors())); CHECK(fg.equals(gnc.getFactors()));
@ -352,7 +381,6 @@ TEST(GncOptimizer, gncConstructorWithRobustGraphAsInput) {
/* ************************************************************************* */ /* ************************************************************************* */
TEST(GncOptimizer, initializeMu) { TEST(GncOptimizer, initializeMu) {
// has to have Gaussian noise models !
auto fg = example::createReallyNonlinearFactorGraph(); auto fg = example::createReallyNonlinearFactorGraph();
Point2 p0(3, 3); Point2 p0(3, 3);
@ -361,8 +389,10 @@ TEST(GncOptimizer, initializeMu) {
LevenbergMarquardtParams lmParams; LevenbergMarquardtParams lmParams;
GncParams<LevenbergMarquardtParams> gncParams(lmParams); GncParams<LevenbergMarquardtParams> gncParams(lmParams);
gncParams.setLossType(GncParams<LevenbergMarquardtParams>::RobustLossType::GM); gncParams.setLossType(
auto gnc = GncOptimizer<GncParams<LevenbergMarquardtParams>>(fg, initial, gncParams); GncParams<LevenbergMarquardtParams>::RobustLossType::GM);
auto gnc = GncOptimizer<GncParams<LevenbergMarquardtParams>>(fg, initial,
gncParams);
EXPECT_DOUBLES_EQUAL(gnc.initializeMu(), 2 * 198.999, 1e-3); // according to rmk 5 in the gnc paper: m0 = 2 rmax^2 / barcSq (barcSq=1 in this example) EXPECT_DOUBLES_EQUAL(gnc.initializeMu(), 2 * 198.999, 1e-3); // according to rmk 5 in the gnc paper: m0 = 2 rmax^2 / barcSq (barcSq=1 in this example)
} }
@ -377,8 +407,10 @@ TEST(GncOptimizer, updateMu) {
LevenbergMarquardtParams lmParams; LevenbergMarquardtParams lmParams;
GncParams<LevenbergMarquardtParams> gncParams(lmParams); GncParams<LevenbergMarquardtParams> gncParams(lmParams);
gncParams.setLossType(GncParams<LevenbergMarquardtParams>::RobustLossType::GM); gncParams.setLossType(
auto gnc = GncOptimizer<GncParams<LevenbergMarquardtParams>>(fg, initial, gncParams); GncParams<LevenbergMarquardtParams>::RobustLossType::GM);
auto gnc = GncOptimizer<GncParams<LevenbergMarquardtParams>>(fg, initial,
gncParams);
double mu = 5.0; double mu = 5.0;
EXPECT_DOUBLES_EQUAL(gnc.updateMu(mu), mu / 1.4, tol); EXPECT_DOUBLES_EQUAL(gnc.updateMu(mu), mu / 1.4, tol);
@ -399,8 +431,10 @@ TEST(GncOptimizer, checkMuConvergence) {
LevenbergMarquardtParams lmParams; LevenbergMarquardtParams lmParams;
GncParams<LevenbergMarquardtParams> gncParams(lmParams); GncParams<LevenbergMarquardtParams> gncParams(lmParams);
gncParams.setLossType(GncParams<LevenbergMarquardtParams>::RobustLossType::GM); gncParams.setLossType(
auto gnc = GncOptimizer<GncParams<LevenbergMarquardtParams>>(fg, initial, gncParams); GncParams<LevenbergMarquardtParams>::RobustLossType::GM);
auto gnc = GncOptimizer<GncParams<LevenbergMarquardtParams>>(fg, initial,
gncParams);
double mu = 1.0; double mu = 1.0;
CHECK(gnc.checkMuConvergence(mu)); CHECK(gnc.checkMuConvergence(mu));
@ -408,7 +442,6 @@ TEST(GncOptimizer, checkMuConvergence) {
/* ************************************************************************* */ /* ************************************************************************* */
TEST(GncOptimizer, calculateWeights) { TEST(GncOptimizer, calculateWeights) {
// has to have Gaussian noise models !
auto fg = example::sharedNonRobustFactorGraphWithOutliers(); auto fg = example::sharedNonRobustFactorGraphWithOutliers();
Point2 p0(0, 0); Point2 p0(0, 0);
@ -433,7 +466,8 @@ TEST(GncOptimizer, calculateWeights) {
double barcSq = 5.0; double barcSq = 5.0;
weights_expected[3] = std::pow(mu * barcSq / (50.0 + mu * barcSq), 2); // outlier, error = 50 weights_expected[3] = std::pow(mu * barcSq / (50.0 + mu * barcSq), 2); // outlier, error = 50
gncParams.setInlierThreshold(barcSq); gncParams.setInlierThreshold(barcSq);
auto gnc2 = GncOptimizer<GncParams<GaussNewtonParams>>(fg, initial, gncParams); auto gnc2 = GncOptimizer<GncParams<GaussNewtonParams>>(fg, initial,
gncParams);
weights_actual = gnc2.calculateWeights(initial, mu); weights_actual = gnc2.calculateWeights(initial, mu);
CHECK(assert_equal(weights_expected, weights_actual, tol)); CHECK(assert_equal(weights_expected, weights_actual, tol));
} }
@ -442,11 +476,13 @@ TEST(GncOptimizer, calculateWeights) {
TEST(GncOptimizer, makeWeightedGraph) { TEST(GncOptimizer, makeWeightedGraph) {
// create original factor // create original factor
double sigma1 = 0.1; double sigma1 = 0.1;
NonlinearFactorGraph nfg = example::nonlinearFactorGraphWithGivenSigma(sigma1); NonlinearFactorGraph nfg = example::nonlinearFactorGraphWithGivenSigma(
sigma1);
// create expected // create expected
double sigma2 = 10; double sigma2 = 10;
NonlinearFactorGraph expected = example::nonlinearFactorGraphWithGivenSigma(sigma2); NonlinearFactorGraph expected = example::nonlinearFactorGraphWithGivenSigma(
sigma2);
// create weights // create weights
Vector weights = Vector::Ones(1); // original info:1/0.1^2 = 100. New info: 1/10^2 = 0.01. Ratio is 10-4 Vector weights = Vector::Ones(1); // original info:1/0.1^2 = 100. New info: 1/10^2 = 0.01. Ratio is 10-4
@ -459,7 +495,8 @@ TEST(GncOptimizer, makeWeightedGraph) {
LevenbergMarquardtParams lmParams; LevenbergMarquardtParams lmParams;
GncParams<LevenbergMarquardtParams> gncParams(lmParams); GncParams<LevenbergMarquardtParams> gncParams(lmParams);
auto gnc = GncOptimizer<GncParams<LevenbergMarquardtParams>>(nfg, initial, gncParams); auto gnc = GncOptimizer<GncParams<LevenbergMarquardtParams>>(nfg, initial,
gncParams);
NonlinearFactorGraph actual = gnc.makeWeightedGraph(weights); NonlinearFactorGraph actual = gnc.makeWeightedGraph(weights);
// check it's all good // check it's all good
@ -468,7 +505,6 @@ TEST(GncOptimizer, makeWeightedGraph) {
/* ************************************************************************* */ /* ************************************************************************* */
TEST(GncOptimizer, optimizeSimple) { TEST(GncOptimizer, optimizeSimple) {
// has to have Gaussian noise models !
auto fg = example::createReallyNonlinearFactorGraph(); auto fg = example::createReallyNonlinearFactorGraph();
Point2 p0(3, 3); Point2 p0(3, 3);
@ -477,7 +513,8 @@ TEST(GncOptimizer, optimizeSimple) {
LevenbergMarquardtParams lmParams; LevenbergMarquardtParams lmParams;
GncParams<LevenbergMarquardtParams> gncParams(lmParams); GncParams<LevenbergMarquardtParams> gncParams(lmParams);
auto gnc = GncOptimizer<GncParams<LevenbergMarquardtParams>>(fg, initial, gncParams); auto gnc = GncOptimizer<GncParams<LevenbergMarquardtParams>>(fg, initial,
gncParams);
Values actual = gnc.optimize(); Values actual = gnc.optimize();
DOUBLES_EQUAL(0, fg.error(actual), tol); DOUBLES_EQUAL(0, fg.error(actual), tol);
@ -485,7 +522,6 @@ TEST(GncOptimizer, optimizeSimple) {
/* ************************************************************************* */ /* ************************************************************************* */
TEST(GncOptimizer, optimize) { TEST(GncOptimizer, optimize) {
// has to have Gaussian noise models !
auto fg = example::sharedNonRobustFactorGraphWithOutliers(); auto fg = example::sharedNonRobustFactorGraphWithOutliers();
Point2 p0(1, 0); Point2 p0(1, 0);
@ -516,7 +552,6 @@ TEST(GncOptimizer, optimize) {
/* ************************************************************************* */ /* ************************************************************************* */
TEST(GncOptimizer, optimizeWithKnownInliers) { TEST(GncOptimizer, optimizeWithKnownInliers) {
// has to have Gaussian noise models !
auto fg = example::sharedNonRobustFactorGraphWithOutliers(); auto fg = example::sharedNonRobustFactorGraphWithOutliers();
Point2 p0(1, 0); Point2 p0(1, 0);