ndt_mcl.hpp
Go to the documentation of this file.
00001 #ifndef NDT_MCL_HPP_
00002 #define NDT_MCL_HPP_
00003 #include <pcl/ros/conversions.h>
00004 #include <pcl/point_cloud.h>
00005 #include <pcl/io/pcd_io.h>
00006 #include <pcl/point_types.h>
00007 #include <ndt_map.h>
00008 #include <ndt_cell.h>
00009 #include <pointcloud_utils.h>
00010 #include <tf_conversions/tf_eigen.h>
00011 #include "CParticleFilter.h"
00012 #include <pointcloud_utils.h>
00013 #include <cstdio>
00014 #include <Eigen/Eigen>
00015 #include <Eigen/Geometry>
00016 #include <fstream>
00017 
00021 template <typename PointT>
00022 class NDTMCL{
00023         public:
00024                 lslgeneric::NDTMap<PointT> map;                 
00025                 mcl::CParticleFilter pf;                                                
00026                 double resolution;
00027                 int counter;
00028                 double zfilt_min;
00029                 bool forceSIR;
00033                 NDTMCL(double map_resolution, lslgeneric::NDTMap<PointT> &nd_map, double zfilter):
00034                         map(new lslgeneric::LazyGrid<PointT>(map_resolution))
00035                 {
00036                                 isInit = false;
00037                                 forceSIR = false;
00038                                 resolution=map_resolution;
00039                                 counter = 0;
00040                                 zfilt_min = zfilter;
00041                                 sinceSIR = 0;
00044                                 double cx,cy,cz;
00045                                 if(!nd_map.getCentroid(cx, cy, cz)){
00046                                         fprintf(stderr,"Centroid NOT Given-abort!\n");
00047                                         exit(1);
00048                                 }else{
00049                                         fprintf(stderr,"Centroid(%lf,%lf,%lf)\n",cx,cy,cz);
00050                                 }
00051                                 
00052                                 double wx,wy,wz;
00053                                 
00054                                 if(!nd_map.getGridSizeInMeters(wx, wy, wz)){
00055                                         fprintf(stderr,"Grid size NOT Given-abort!\n");
00056                                         exit(1);
00057                                 }else{
00058                                         fprintf(stderr,"GridSize(%lf,%lf,%lf)\n",wx,wy,wz);
00059                                 }
00060                                 
00061                                 map.initialize(cx,cy,cz,wx,wy,wz);
00062                                 
00063                                 std::vector<lslgeneric::NDTCell<PointT>*> ndts;
00064                                 ndts = nd_map.getAllCells();
00065                                 fprintf(stderr,"NDT MAP with %d components",ndts.size());
00066                                 for(unsigned int i=0;i<ndts.size();i++){
00067                                         Eigen::Vector3d m = ndts[i]->getMean(); 
00068                                         if(m[2]>zfilter){
00069                                                 Eigen::Matrix3d cov = ndts[i]->getCov();
00070                                                 unsigned int nump = ndts[i]->getN();
00071                                                 //fprintf(stderr,"-%d-",nump);
00072                                                 map.addDistributionToCell(cov, m,nump);
00073                                         }
00074                                 }
00075                 }
00076                 
00081                 void initializeFilter(double x, double y, double yaw,double x_e, double y_e, double yaw_e, int numParticles){
00082                         pf.initializeNormalRandom(mcl::pose(x,y,yaw), mcl::pose(x_e,y_e,yaw_e),numParticles);
00083                 }
00084                 
00085                 void updateAndPredict(Eigen::Affine3d Tmotion, pcl::PointCloud<PointT> &cloud){
00086                         Eigen::Vector3d tr = Tmotion.translation();
00087                         Eigen::Vector3d rot = Tmotion.rotation().eulerAngles(0,1,2);
00088                         
00089                         
00099                         pf.predict(mcl::pose(tr[0],tr[1],rot[2]), mcl::pose(tr[0]*0.1 + 0.005,tr[1]*0.1+ 0.005,rot[2]*0.1+0.001));
00100                         
00101                         lslgeneric::NDTMap<PointT> local_map(new lslgeneric::LazyGrid<PointT>(resolution));
00102                         
00103                         /*
00104                         pcl::PointCloud<PointT> cl_f;
00105                         pcl::PointCloud<PointT> cl_z;
00106                         for(int i=0;i<cloud.size();i++){
00107                                 if(cloud.points[i].z > 1.95 && cloud.points[i].z <2.05 ) cl_z.push_back(cloud.points[i]);
00108                                 if(cloud.points[i].z > zfilt_min ) cl_f.push_back(cloud.points[i]);
00109                         }
00110                         
00111                         
00112                         fprintf(stderr,"2D scan = %d Points\n",cl_z.size());
00113                         
00114                         */
00115                         //fprintf(stderr,"Could = %d Points\n",cloud.size());
00116                         local_map.addPointCloudSimple(cloud);
00117                         //local_map.computeNDTCells();
00118                         local_map.computeNDTCells(CELL_UPDATE_MODE_STUDENT_T);
00119                  int Nn = 0;
00120 //                      #pragma omp parallel for
00121                         for(int i=0;i<pf.NumOfParticles;i++){
00122                                 Eigen::Affine3d T = getAsAffine(i);
00123                                 
00124                                 std::vector<lslgeneric::NDTCell<PointT>*> ndts;
00125                                 ndts = local_map.pseudoTransformNDT(T);
00126                                 double score=1;
00127                                 
00128                                 if(ndts.size()==0) fprintf(stderr,"ERROR no gaussians in measurement!!!\n");
00129                                 Nn = ndts.size();
00130                                 for(int n=0;n<ndts.size();n++){
00131                                         Eigen::Vector3d m = ndts[n]->getMean(); 
00132                                         if(m[2]<zfilt_min) continue;
00133                                 
00134                                         lslgeneric::NDTCell<PointT> *cell;
00135                                         PointT p;
00136                                         p.x = m[0];p.y=m[1];p.z=m[2];
00137                                         
00138                                         //if(map.getCellAtPoint(p,cell)){
00139                                         if(map.getCellForPoint(p,cell)){
00140                                                 if(cell == NULL) continue;
00141                                                 if(cell->hasGaussian_){
00142                                                         Eigen::Matrix3d covCombined = cell->getCov() + ndts[n]->getCov();
00143                                                         Eigen::Matrix3d icov;
00144                                                         bool exists;
00145                                                         double det = 0;
00146                                                         covCombined.computeInverseAndDetWithCheck(icov,det,exists);
00147                                                         if(!exists) continue;
00148                                                         double l = (cell->getMean() - m).dot(icov*(cell->getMean() - m));
00149                                                         if(l*0 != 0) continue;
00150                                                         //if(l > 120) continue;
00151                                                         score += 0.1 + 0.9 * exp(-0.05*l/2.0);
00152                                                 }else{
00153                                                 }
00154                                         }
00155                                 }
00156                                 
00157                                  /*  -lfd1*(exp(-lfd2*l/2));*/
00158                                 
00159                                 pf.Particles[i].lik = score;
00160                                 for(unsigned int j=0;j<ndts.size();j++){
00161                                         delete ndts[j];
00162                                 }
00163                                 
00164                         }
00165                         
00166                         pf.normalize();
00167                         
00168 
00169                         if(forceSIR){
00170                                 
00171                                 fprintf(stderr, "forceSIR(%d) ",forceSIR);
00172                                 pf.SIRUpdate();
00173                         }else{
00174                         
00175                                 double varP=0;
00176                                 for(int i = 0; i<pf.NumOfParticles;i++){
00177                                         varP += (pf.Particles[i].p - 1.0/pf.NumOfParticles)*(pf.Particles[i].p - 1.0/pf.NumOfParticles);
00178                                 }
00179                                 varP /= pf.NumOfParticles;
00180                                 varP = sqrt(varP); 
00181                                 fprintf(stderr,"Var P=%lf (Npf=%d, Nm=%d)",varP,pf.NumOfParticles, Nn);
00182                                 if(varP > 0.006 || sinceSIR >25){
00183                                         fprintf(stderr,"-SIR- ");
00184                                         sinceSIR = 0;
00185                                         pf.SIRUpdate();
00186                                 }else{
00187                                         sinceSIR++;
00188                                 }
00189                                 
00190                         }
00191                 }
00192                 
00193                 Eigen::Vector3d getMean(){
00194                         mcl::pose m = pf.getDistributionMean(true);
00195                         Eigen::Vector3d mm; mm<<m.x,m.y,m.a;
00196                         return mm;
00197                 }
00198                 
00199                 
00200         private:
00201                 bool isInit;
00202                 int sinceSIR;
00203                 Eigen::Affine3d getAsAffine(int i){
00204                         Eigen::Matrix3d m;
00205                         m = Eigen::AngleAxisd(0, Eigen::Vector3d::UnitX())
00206                                         * Eigen::AngleAxisd(0, Eigen::Vector3d::UnitY())
00207                                         * Eigen::AngleAxisd(pf.Particles[i].a, Eigen::Vector3d::UnitZ());
00208                         Eigen::Translation3d v(pf.Particles[i].x,pf.Particles[i].y,0);
00209                         Eigen::Affine3d T = v*m;
00210                         return T;
00211                 }
00212                 
00213 };
00214 
00215 #endif
00216 


ndt_mcl
Author(s): jari
autogenerated on Mon Jan 6 2014 11:32:07