gtsam/gtsam_unstable/nonlinear/BatchFixedLagSmoother.cpp

411 lines
16 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 BatchFixedLagSmoother.cpp
* @brief An LM-based fixed-lag smoother.
*
* @author Michael Kaess, Stephen Williams
* @date Oct 14, 2012
*/
#include <gtsam_unstable/nonlinear/BatchFixedLagSmoother.h>
#include <gtsam_unstable/nonlinear/LinearizedFactor.h>
#include <gtsam/linear/GaussianFactorGraph.h>
#include <gtsam/linear/GaussianFactor.h>
#include <gtsam/inference/inference.h>
#include <gtsam/base/debug.h>
namespace gtsam {
/* ************************************************************************* */
void BatchFixedLagSmoother::print(const std::string& s, const KeyFormatter& keyFormatter) const {
FixedLagSmoother::print(s, keyFormatter);
// TODO: What else to print?
}
/* ************************************************************************* */
bool BatchFixedLagSmoother::equals(const FixedLagSmoother& rhs, double tol) const {
const BatchFixedLagSmoother* e = dynamic_cast<const BatchFixedLagSmoother*> (&rhs);
return e != NULL
&& FixedLagSmoother::equals(*e, tol)
&& factors_.equals(e->factors_, tol)
&& theta_.equals(e->theta_, tol);
}
/* ************************************************************************* */
FixedLagSmoother::Result BatchFixedLagSmoother::update(const NonlinearFactorGraph& newFactors, const Values& newTheta, const KeyTimestampMap& timestamps) {
const bool debug = ISDEBUG("BatchFixedLagSmoother update");
if(debug) {
std::cout << "BatchFixedLagSmoother::update() START" << std::endl;
}
// Add the new factors
updateFactors(newFactors);
// Add the new variables
theta_.insert(newTheta);
// Update the Timestamps associated with the factor keys
updateKeyTimestampMap(timestamps);
// Get current timestamp
double current_timestamp = getCurrentTimestamp();
if(debug) std::cout << "Current Timestamp: " << current_timestamp << std::endl;
// Find the set of variables to be marginalized out
std::set<Key> marginalizableKeys = findKeysBefore(current_timestamp - smootherLag_);
if(debug) {
std::cout << "Marginalizable Keys: ";
BOOST_FOREACH(Key key, marginalizableKeys) {
std::cout << DefaultKeyFormatter(key) << " ";
}
std::cout << std::endl;
}
// Marginalize out these variables.
// This removes any factors that touch marginalized variables and adds new linear(ized) factors to the graph
marginalizeKeys(marginalizableKeys);
// Create the optimizer
Values linpoint;
linpoint.insert(theta_);
if(enforceConsistency_ && linearizedKeys_.size() > 0) {
linpoint.update(linearizedKeys_);
}
LevenbergMarquardtOptimizer optimizer(factors_, linpoint, parameters_);
// Use a custom optimization loop so the linearization points can be controlled
double currentError;
do {
// Do next iteration
currentError = optimizer.error();
optimizer.iterate();
// Force variables associated with linearized factors to keep the same linearization point
if(enforceConsistency_ && linearizedKeys_.size() > 0) {
// Put the old values of the linearized keys back into the optimizer state
optimizer.state().values.update(linearizedKeys_);
optimizer.state().error = factors_.error(optimizer.state().values);
}
// Maybe show output
if(parameters_.verbosity >= NonlinearOptimizerParams::VALUES) optimizer.values().print("newValues");
if(parameters_.verbosity >= NonlinearOptimizerParams::ERROR) std::cout << "newError: " << optimizer.error() << std::endl;
} while(optimizer.iterations() < parameters_.maxIterations &&
!checkConvergence(parameters_.relativeErrorTol, parameters_.absoluteErrorTol,
parameters_.errorTol, currentError, optimizer.error(), parameters_.verbosity));
// Update the Values from the optimizer
theta_ = optimizer.values();
// Create result structure
Result result;
result.iterations = optimizer.state().iterations;
result.linearVariables = linearizedKeys_.size();
result.nonlinearVariables = theta_.size() - linearizedKeys_.size();
result.error = optimizer.state().error;
if(debug) {
std::cout << "BatchFixedLagSmoother::update() FINISH" << std::endl;
}
return result;
}
/* ************************************************************************* */
void BatchFixedLagSmoother::updateFactors(const NonlinearFactorGraph& newFactors) {
BOOST_FOREACH(const NonlinearFactor::shared_ptr& factor, newFactors) {
Index index;
// Insert the factor into an existing hole in the factor graph, if possible
if(availableSlots_.size() > 0) {
index = availableSlots_.front();
availableSlots_.pop();
factors_.replace(index, factor);
} else {
index = factors_.size();
factors_.push_back(factor);
}
// Update the FactorIndex
BOOST_FOREACH(Key key, *factor) {
factorIndex_[key].insert(index);
}
}
}
/* ************************************************************************* */
void BatchFixedLagSmoother::removeFactors(const std::set<size_t>& deleteFactors) {
BOOST_FOREACH(size_t slot, deleteFactors) {
if(factors_.at(slot)) {
// Remove references to this factor from the FactorIndex
BOOST_FOREACH(Key key, *(factors_.at(slot))) {
factorIndex_[key].erase(slot);
}
// Remove the factor from the factor graph
factors_.remove(slot);
// Add the factor's old slot to the list of available slots
availableSlots_.push(slot);
} else {
// TODO: Throw an error??
std::cout << "Attempting to remove a factor from slot " << slot << ", but it is already NULL." << std::endl;
}
}
}
/* ************************************************************************* */
void BatchFixedLagSmoother::eraseKeys(const std::set<Key>& keys) {
BOOST_FOREACH(Key key, keys) {
// Erase the key from the values
theta_.erase(key);
// Erase the key from the factor index
factorIndex_.erase(key);
// Erase the key from the set of linearized keys
if(linearizedKeys_.exists(key)) {
linearizedKeys_.erase(key);
}
}
eraseKeyTimestampMap(keys);
}
/* ************************************************************************* */
struct FactorTree {
std::set<Index> factors;
std::set<Key> keys;
FactorTree(const std::set<Index>& factors, const NonlinearFactorGraph& allFactors) : factors(factors) {
BOOST_FOREACH(const Index& factor, factors) {
BOOST_FOREACH(Key key, *(allFactors.at(factor))) {
keys.insert(key);
}
}
};
void push_back(const FactorTree& factorTree) {
factors.insert(factorTree.factors.begin(), factorTree.factors.end());
keys.insert(factorTree.keys.begin(), factorTree.keys.end());
}
bool hasCommonKeys(Index factor, const NonlinearFactorGraph& allFactors) {
const NonlinearFactor::shared_ptr& f = allFactors.at(factor);
std::set<Key>::const_iterator iter = std::find_first_of(keys.begin(), keys.end(), f->begin(), f->end());
return iter != keys.end();
}
template <class ForwardIterator>
bool hasCommonKeys(ForwardIterator first, ForwardIterator last, const NonlinearFactorGraph& allFactors) {
for(ForwardIterator factor = first; factor != last; ++factor) {
if(hasCommonKeys(*factor, allFactors))
return true;
}
return false;
}
};
/* ************************************************************************* */
void BatchFixedLagSmoother::marginalizeKeys(const std::set<Key>& marginalizableKeys) {
const bool debug = ISDEBUG("BatchFixedLagSmoother update");
if(debug) std::cout << "BatchFixedLagSmoother::marginalizeKeys() START" << std::endl;
// In order to marginalize out the selected variables, the factors involved in those variables
// must be identified and removed from iSAM2. Also, the effect of those removed factors on the
// remaining variables needs to be accounted for. This will be done with linear(ized) factors from
// a partial clique marginalization (or from the iSAM2 cached factor if the entire clique is removed).
// This function finds the set of factors to be removed and generates the linearized factors that
// must be added.
if(marginalizableKeys.size() > 0) {
if(debug) PrintKeySet(marginalizableKeys, "Marginalizable Keys:");
// Find all of the factors associated with marginalizable variables. This set of factors may form a forest.
typedef std::list<FactorTree> FactorForest;
FactorForest factorForest;
BOOST_FOREACH(Key key, marginalizableKeys) {
if(debug) std::cout << "Looking for factors involving key " << DefaultKeyFormatter(key) << std::endl;
// Get the factors associated with this variable
const std::set<size_t>& factors = factorIndex_.at(key);
if(debug) { std::cout << "Found the following factors:" << std::endl; BOOST_FOREACH(size_t i, factors) { std::cout << " "; PrintSymbolicFactor(factors_.at(i)); } }
// Loop over existing factor trees, looking for common keys
std::vector<FactorForest::iterator> commonTrees;
for(FactorForest::iterator tree = factorForest.begin(); tree != factorForest.end(); ++tree) {
if(tree->hasCommonKeys(factors.begin(), factors.end(), factors_)) {
commonTrees.push_back(tree);
}
}
if(debug) std::cout << "Found " << commonTrees.size() << " common trees." << std::endl;
if(commonTrees.size() == 0) {
// No common trees were found. Create a new one.
factorForest.push_back(FactorTree(factors, factors_));
if(debug) std::cout << "Created a new tree." << std::endl;
} else {
// Extract the last common tree
FactorForest::iterator commonTree = commonTrees.back();
commonTrees.pop_back();
// Merge the current factors into this tree
commonTree->push_back(FactorTree(factors, factors_));
// Merge all other common trees into this one, deleting the other trees from the forest.
BOOST_FOREACH(FactorForest::iterator& tree, commonTrees) {
commonTree->push_back(*tree);
factorForest.erase(tree);
}
}
}
if(debug) std::cout << "Found " << factorForest.size() << " factor trees in the set of removed factors." << std::endl;
// For each tree in the forest:
// (0) construct an ordering for the tree
// (1) construct a linear factor graph
// (2) solve for the marginal factors
// (3) convert the marginal factors into Linearized Factors
// (4) remove the marginalized factors from the graph
// (5) add the factors in this tree to the graph
BOOST_FOREACH(const FactorTree& factorTree, factorForest) {
// (0) construct an ordering for this tree
// The ordering should place the marginalizable keys first, then the remaining keys
Ordering ordering;
std::set<Key> marginalizableTreeKeys;
std::set_intersection(factorTree.keys.begin(), factorTree.keys.end(),
marginalizableKeys.begin(), marginalizableKeys.end(),
std::inserter(marginalizableTreeKeys, marginalizableTreeKeys.end()));
std::set<Key> remainingTreeKeys;
std::set_difference(factorTree.keys.begin(), factorTree.keys.end(),
marginalizableTreeKeys.begin(), marginalizableTreeKeys.end(),
std::inserter(remainingTreeKeys, remainingTreeKeys.end()));
// TODO: It may be worthwhile to use CCOLAMD here. (but maybe not???)
BOOST_FOREACH(Key key, marginalizableTreeKeys) {
ordering.push_back(key);
}
BOOST_FOREACH(Key key, remainingTreeKeys) {
ordering.push_back(key);
}
// (1) construct a linear factor graph
GaussianFactorGraph graph;
BOOST_FOREACH(size_t factor, factorTree.factors) {
graph.push_back( factors_.at(factor)->linearize(theta_, ordering) );
}
if(debug) PrintSymbolicGraph(graph, ordering, "Factor Tree:");
// (2) solve for the marginal factors
// Perform partial elimination, resulting in a conditional probability ( P(MarginalizedVariable | RemainingVariables)
// and factors on the remaining variables ( f(RemainingVariables) ). These are the factors we need to add to iSAM2
std::vector<Index> variables;
BOOST_FOREACH(Key key, marginalizableTreeKeys) {
variables.push_back(ordering.at(key));
}
std::pair<GaussianFactorGraph::sharedConditional, GaussianFactorGraph> result = graph.eliminate(variables);
graph = result.second;
if(debug) PrintSymbolicGraph(graph, ordering, "Factors on Remaining Variables:");
// (3) convert the marginal factors into Linearized Factors
NonlinearFactorGraph newFactors;
BOOST_FOREACH(const GaussianFactor::shared_ptr& gaussianFactor, graph) {
// These factors are all generated from BayesNet conditionals. They should all be Jacobians.
JacobianFactor::shared_ptr jacobianFactor = boost::dynamic_pointer_cast<JacobianFactor>(gaussianFactor);
assert(jacobianFactor);
LinearizedGaussianFactor::shared_ptr factor = LinearizedJacobianFactor::shared_ptr(new LinearizedJacobianFactor(jacobianFactor, ordering, theta_));
// add it to the new factor set
newFactors.push_back(factor);
}
// (4) remove the marginalized factors from the graph
removeFactors(factorTree.factors);
// (5) add the factors in this tree to the main set of factors
updateFactors(newFactors);
// (6) add the keys involved in the linear(ized) factors to the linearizedKey list
FastSet<Key> linearizedKeys = newFactors.keys();
BOOST_FOREACH(Key key, linearizedKeys) {
if(!linearizedKeys_.exists(key)) {
linearizedKeys_.insert(key, theta_.at(key));
}
}
}
// Remove the marginalized keys from the smoother data structures
eraseKeys(marginalizableKeys);
}
if(debug) std::cout << "BatchFixedLagSmoother::marginalizeKeys() FINISH" << std::endl;
}
/* ************************************************************************* */
void BatchFixedLagSmoother::PrintKeySet(const std::set<Key>& keys, const std::string& label) {
std::cout << label;
BOOST_FOREACH(gtsam::Key key, keys) {
std::cout << " " << gtsam::DefaultKeyFormatter(key);
}
std::cout << std::endl;
}
/* ************************************************************************* */
void BatchFixedLagSmoother::PrintSymbolicFactor(const NonlinearFactor::shared_ptr& factor) {
std::cout << "f(";
if(factor) {
BOOST_FOREACH(Key key, factor->keys()) {
std::cout << " " << gtsam::DefaultKeyFormatter(key);
}
} else {
std::cout << " NULL";
}
std::cout << " )" << std::endl;
}
/* ************************************************************************* */
void BatchFixedLagSmoother::PrintSymbolicFactor(const GaussianFactor::shared_ptr& factor, const Ordering& ordering) {
std::cout << "f(";
BOOST_FOREACH(Index index, factor->keys()) {
std::cout << " " << index << "[" << gtsam::DefaultKeyFormatter(ordering.key(index)) << "]";
}
std::cout << " )" << std::endl;
}
/* ************************************************************************* */
void BatchFixedLagSmoother::PrintSymbolicGraph(const NonlinearFactorGraph& graph, const std::string& label) {
std::cout << label << std::endl;
BOOST_FOREACH(const NonlinearFactor::shared_ptr& factor, graph) {
PrintSymbolicFactor(factor);
}
}
/* ************************************************************************* */
void BatchFixedLagSmoother::PrintSymbolicGraph(const GaussianFactorGraph& graph, const Ordering& ordering, const std::string& label) {
std::cout << label << std::endl;
BOOST_FOREACH(const GaussianFactor::shared_ptr& factor, graph) {
PrintSymbolicFactor(factor, ordering);
}
}
/* ************************************************************************* */
} /// namespace gtsam