84 lines
		
	
	
		
			2.3 KiB
		
	
	
	
		
			C++
		
	
	
			
		
		
	
	
			84 lines
		
	
	
		
			2.3 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    timeVirtual.cpp
 | |
|  * @brief   Time the overhead of using virtual destructors and methods
 | |
|  * @author  Richard Roberts
 | |
|  * @date    Dec 3, 2010
 | |
|  */
 | |
| 
 | |
| #include <gtsam/slam/dataset.h>
 | |
| #include <gtsam/slam/PriorFactor.h>
 | |
| #include <gtsam/slam/lago.h>
 | |
| #include <gtsam/geometry/Pose2.h>
 | |
| #include <gtsam/nonlinear/GaussNewtonOptimizer.h>
 | |
| #include <gtsam/linear/Sampler.h>
 | |
| #include <gtsam/base/timing.h>
 | |
| 
 | |
| #include <iostream>
 | |
| 
 | |
| using namespace std;
 | |
| using namespace gtsam;
 | |
| 
 | |
| int main(int argc, char *argv[]) {
 | |
| 
 | |
|   size_t trials = 1;
 | |
| 
 | |
|   // read graph
 | |
|   Values::shared_ptr solution;
 | |
|   NonlinearFactorGraph::shared_ptr g;
 | |
|   string inputFile = findExampleDataFile("w10000");
 | |
|   SharedDiagonal model = noiseModel::Diagonal::Sigmas((Vector(3) << 0.05, 0.05, 5.0 * M_PI / 180.0).finished());
 | |
|   boost::tie(g, solution) = load2D(inputFile, model);
 | |
| 
 | |
|   // add noise to create initial estimate
 | |
|   Values initial;
 | |
|   Sampler sampler(42u);
 | |
|   Values::ConstFiltered<Pose2> poses = solution->filter<Pose2>();
 | |
|   SharedDiagonal noise = noiseModel::Diagonal::Sigmas((Vector(3) << 0.5, 0.5, 15.0 * M_PI / 180.0).finished());
 | |
|   for(const Values::ConstFiltered<Pose2>::KeyValuePair& it: poses)
 | |
|     initial.insert(it.key, it.value.retract(sampler.sampleNewModel(noise)));
 | |
| 
 | |
|   // Add prior on the pose having index (key) = 0
 | |
|   noiseModel::Diagonal::shared_ptr priorModel = //
 | |
|       noiseModel::Diagonal::Sigmas(Vector3(1e-6, 1e-6, 1e-8));
 | |
|   g->add(PriorFactor<Pose2>(0, Pose2(), priorModel));
 | |
| 
 | |
|   // LAGO
 | |
|   for (size_t i = 0; i < trials; i++) {
 | |
|     {
 | |
|       gttic_(lago);
 | |
| 
 | |
|       gttic_(init);
 | |
|       Values lagoInitial = lago::initialize(*g);
 | |
|       gttoc_(init);
 | |
| 
 | |
|       gttic_(refine);
 | |
|       GaussNewtonOptimizer optimizer(*g, lagoInitial);
 | |
|       Values result = optimizer.optimize();
 | |
|       gttoc_(refine);
 | |
|     }
 | |
| 
 | |
|     {
 | |
|       gttic_(optimize);
 | |
|       GaussNewtonOptimizer optimizer(*g, initial);
 | |
|       Values result = optimizer.optimize();
 | |
|     }
 | |
| 
 | |
|     tictoc_finishedIteration_();
 | |
|   }
 | |
| 
 | |
|   tictoc_print_();
 | |
| 
 | |
|   return 0;
 | |
| }
 |